forked from FFmpeg/FFmpeg
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdvenc.c
1229 lines (1081 loc) · 44.8 KB
/
dvenc.c
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
/*
* DV encoder
* Copyright (c) 2003 Roman Shaposhnik
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*
* quant_deadzone code and fixes sponsored by NOA GmbH
*/
/**
* @file
* DV encoder
*/
#include "config.h"
#include "libavutil/attributes.h"
#include "libavutil/internal.h"
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "avcodec.h"
#include "dv.h"
#include "dv_profile_internal.h"
#include "dv_tablegen.h"
#include "fdctdsp.h"
#include "internal.h"
#include "mathops.h"
#include "me_cmp.h"
#include "pixblockdsp.h"
#include "put_bits.h"
static av_cold int dvvideo_encode_init(AVCodecContext *avctx)
{
DVVideoContext *s = avctx->priv_data;
FDCTDSPContext fdsp;
MECmpContext mecc;
PixblockDSPContext pdsp;
int ret;
s->sys = av_dv_codec_profile2(avctx->width, avctx->height, avctx->pix_fmt, avctx->time_base);
if (!s->sys) {
av_log(avctx, AV_LOG_ERROR, "Found no DV profile for %ix%i %s video. "
"Valid DV profiles are:\n",
avctx->width, avctx->height, av_get_pix_fmt_name(avctx->pix_fmt));
ff_dv_print_profiles(avctx, AV_LOG_ERROR);
return AVERROR(EINVAL);
}
ret = ff_dv_init_dynamic_tables(s, s->sys);
if (ret < 0) {
av_log(avctx, AV_LOG_ERROR, "Error initializing work tables.\n");
return ret;
}
dv_vlc_map_tableinit();
memset(&fdsp,0, sizeof(fdsp));
memset(&mecc,0, sizeof(mecc));
memset(&pdsp,0, sizeof(pdsp));
ff_fdctdsp_init(&fdsp, avctx);
ff_me_cmp_init(&mecc, avctx);
ff_pixblockdsp_init(&pdsp, avctx);
ff_set_cmp(&mecc, mecc.ildct_cmp, avctx->ildct_cmp);
s->get_pixels = pdsp.get_pixels;
s->ildct_cmp = mecc.ildct_cmp[5];
s->fdct[0] = fdsp.fdct;
s->fdct[1] = fdsp.fdct248;
return ff_dvvideo_init(avctx);
}
/* bit budget for AC only in 5 MBs */
static const int vs_total_ac_bits_hd = (68 * 6 + 52*2) * 5;
static const int vs_total_ac_bits = (100 * 4 + 68 * 2) * 5;
static const int mb_area_start[5] = { 1, 6, 21, 43, 64 };
#if CONFIG_SMALL
/* Convert run and level (where level != 0) pair into VLC, returning bit size */
static av_always_inline int dv_rl2vlc(int run, int level, int sign,
uint32_t *vlc)
{
int size;
if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) {
*vlc = dv_vlc_map[run][level].vlc | sign;
size = dv_vlc_map[run][level].size;
} else {
if (level < DV_VLC_MAP_LEV_SIZE) {
*vlc = dv_vlc_map[0][level].vlc | sign;
size = dv_vlc_map[0][level].size;
} else {
*vlc = 0xfe00 | (level << 1) | sign;
size = 16;
}
if (run) {
*vlc |= ((run < 16) ? dv_vlc_map[run - 1][0].vlc :
(0x1f80 | (run - 1))) << size;
size += (run < 16) ? dv_vlc_map[run - 1][0].size : 13;
}
}
return size;
}
static av_always_inline int dv_rl2vlc_size(int run, int level)
{
int size;
if (run < DV_VLC_MAP_RUN_SIZE && level < DV_VLC_MAP_LEV_SIZE) {
size = dv_vlc_map[run][level].size;
} else {
size = (level < DV_VLC_MAP_LEV_SIZE) ? dv_vlc_map[0][level].size : 16;
if (run)
size += (run < 16) ? dv_vlc_map[run - 1][0].size : 13;
}
return size;
}
#else
static av_always_inline int dv_rl2vlc(int run, int l, int sign, uint32_t *vlc)
{
*vlc = dv_vlc_map[run][l].vlc | sign;
return dv_vlc_map[run][l].size;
}
static av_always_inline int dv_rl2vlc_size(int run, int l)
{
return dv_vlc_map[run][l].size;
}
#endif
typedef struct EncBlockInfo {
int area_q[4];
int bit_size[4];
int prev[5];
int cur_ac;
int cno;
int dct_mode;
int16_t mb[64];
uint8_t next[64];
uint8_t sign[64];
uint8_t partial_bit_count;
uint32_t partial_bit_buffer; /* we can't use uint16_t here */
/* used by DV100 only: a copy of the weighted and classified but
not-yet-quantized AC coefficients. This is necessary for
re-quantizing at different steps. */
int16_t save[64];
int min_qlevel; /* DV100 only: minimum qlevel (for AC coefficients >255) */
} EncBlockInfo;
static av_always_inline PutBitContext *dv_encode_ac(EncBlockInfo *bi,
PutBitContext *pb_pool,
PutBitContext *pb_end)
{
int prev, bits_left;
PutBitContext *pb = pb_pool;
int size = bi->partial_bit_count;
uint32_t vlc = bi->partial_bit_buffer;
bi->partial_bit_count =
bi->partial_bit_buffer = 0;
for (;;) {
/* Find suitable storage space */
for (; size > (bits_left = put_bits_left(pb)); pb++) {
if (bits_left) {
size -= bits_left;
put_bits(pb, bits_left, vlc >> size);
vlc = av_mod_uintp2(vlc, size);
}
if (pb + 1 >= pb_end) {
bi->partial_bit_count = size;
bi->partial_bit_buffer = vlc;
return pb;
}
}
/* Store VLC */
put_bits(pb, size, vlc);
if (bi->cur_ac >= 64)
break;
/* Construct the next VLC */
prev = bi->cur_ac;
bi->cur_ac = bi->next[prev];
if (bi->cur_ac < 64) {
size = dv_rl2vlc(bi->cur_ac - prev - 1, bi->mb[bi->cur_ac],
bi->sign[bi->cur_ac], &vlc);
} else {
size = 4;
vlc = 6; /* End Of Block stamp */
}
}
return pb;
}
static av_always_inline int dv_guess_dct_mode(DVVideoContext *s, uint8_t *data,
ptrdiff_t linesize)
{
if (s->avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT) {
int ps = s->ildct_cmp(NULL, data, NULL, linesize, 8) - 400;
if (ps > 0) {
int is = s->ildct_cmp(NULL, data, NULL, linesize << 1, 4) +
s->ildct_cmp(NULL, data + linesize, NULL, linesize << 1, 4);
return ps > is;
}
}
return 0;
}
static const int dv_weight_bits = 18;
static const int dv_weight_88[64] = {
131072, 257107, 257107, 242189, 252167, 242189, 235923, 237536,
237536, 235923, 229376, 231390, 223754, 231390, 229376, 222935,
224969, 217965, 217965, 224969, 222935, 200636, 218652, 211916,
212325, 211916, 218652, 200636, 188995, 196781, 205965, 206433,
206433, 205965, 196781, 188995, 185364, 185364, 200636, 200704,
200636, 185364, 185364, 174609, 180568, 195068, 195068, 180568,
174609, 170091, 175557, 189591, 175557, 170091, 165371, 170627,
170627, 165371, 160727, 153560, 160727, 144651, 144651, 136258,
};
static const int dv_weight_248[64] = {
131072, 262144, 257107, 257107, 242189, 242189, 242189, 242189,
237536, 237536, 229376, 229376, 200636, 200636, 224973, 224973,
223754, 223754, 235923, 235923, 229376, 229376, 217965, 217965,
211916, 211916, 196781, 196781, 185364, 185364, 206433, 206433,
211916, 211916, 222935, 222935, 200636, 200636, 205964, 205964,
200704, 200704, 180568, 180568, 175557, 175557, 195068, 195068,
185364, 185364, 188995, 188995, 174606, 174606, 175557, 175557,
170627, 170627, 153560, 153560, 165371, 165371, 144651, 144651,
};
/* setting this to 1 results in a faster codec but
* somewhat lower image quality */
#define DV100_SACRIFICE_QUALITY_FOR_SPEED 1
#define DV100_ENABLE_FINER 1
/* pack combination of QNO and CNO into a single 8-bit value */
#define DV100_MAKE_QLEVEL(qno,cno) ((qno<<2) | (cno))
#define DV100_QLEVEL_QNO(qlevel) (qlevel>>2)
#define DV100_QLEVEL_CNO(qlevel) (qlevel&0x3)
#define DV100_NUM_QLEVELS 31
/* The quantization step is determined by a combination of QNO and
CNO. We refer to these combinations as "qlevels" (this term is our
own, it's not mentioned in the spec). We use CNO, a multiplier on
the quantization step, to "fill in the gaps" between quantization
steps associated with successive values of QNO. e.g. there is no
QNO for a quantization step of 10, but we can use QNO=5 CNO=1 to
get the same result. The table below encodes combinations of QNO
and CNO in order of increasing quantization coarseness. */
static const uint8_t dv100_qlevels[DV100_NUM_QLEVELS] = {
DV100_MAKE_QLEVEL( 1,0), // 1*1= 1
DV100_MAKE_QLEVEL( 1,0), // 1*1= 1
DV100_MAKE_QLEVEL( 2,0), // 2*1= 2
DV100_MAKE_QLEVEL( 3,0), // 3*1= 3
DV100_MAKE_QLEVEL( 4,0), // 4*1= 4
DV100_MAKE_QLEVEL( 5,0), // 5*1= 5
DV100_MAKE_QLEVEL( 6,0), // 6*1= 6
DV100_MAKE_QLEVEL( 7,0), // 7*1= 7
DV100_MAKE_QLEVEL( 8,0), // 8*1= 8
DV100_MAKE_QLEVEL( 5,1), // 5*2=10
DV100_MAKE_QLEVEL( 6,1), // 6*2=12
DV100_MAKE_QLEVEL( 7,1), // 7*2=14
DV100_MAKE_QLEVEL( 9,0), // 16*1=16
DV100_MAKE_QLEVEL(10,0), // 18*1=18
DV100_MAKE_QLEVEL(11,0), // 20*1=20
DV100_MAKE_QLEVEL(12,0), // 22*1=22
DV100_MAKE_QLEVEL(13,0), // 24*1=24
DV100_MAKE_QLEVEL(14,0), // 28*1=28
DV100_MAKE_QLEVEL( 9,1), // 16*2=32
DV100_MAKE_QLEVEL(10,1), // 18*2=36
DV100_MAKE_QLEVEL(11,1), // 20*2=40
DV100_MAKE_QLEVEL(12,1), // 22*2=44
DV100_MAKE_QLEVEL(13,1), // 24*2=48
DV100_MAKE_QLEVEL(15,0), // 52*1=52
DV100_MAKE_QLEVEL(14,1), // 28*2=56
DV100_MAKE_QLEVEL( 9,2), // 16*4=64
DV100_MAKE_QLEVEL(10,2), // 18*4=72
DV100_MAKE_QLEVEL(11,2), // 20*4=80
DV100_MAKE_QLEVEL(12,2), // 22*4=88
DV100_MAKE_QLEVEL(13,2), // 24*4=96
// ...
DV100_MAKE_QLEVEL(15,3), // 52*8=416
};
static const int dv100_min_bias = 0;
static const int dv100_chroma_bias = 0;
static const int dv100_starting_qno = 1;
#if DV100_SACRIFICE_QUALITY_FOR_SPEED
static const int dv100_qlevel_inc = 4;
#else
static const int dv100_qlevel_inc = 1;
#endif
// 1/qstep, shifted up by 16 bits
static const int dv100_qstep_bits = 16;
static const int dv100_qstep_inv[16] = {
65536, 65536, 32768, 21845, 16384, 13107, 10923, 9362, 8192, 4096, 3641, 3277, 2979, 2731, 2341, 1260,
};
/* DV100 weights are pre-zigzagged, inverted and multiplied by 2^(dv100_weight_shift)
(in DV100 the AC components are divided by the spec weights) */
static const int dv100_weight_shift = 16;
static const int dv_weight_1080[2][64] = {
{ 8192, 65536, 65536, 61681, 61681, 61681, 58254, 58254,
58254, 58254, 58254, 58254, 55188, 58254, 58254, 55188,
55188, 55188, 55188, 55188, 55188, 24966, 27594, 26214,
26214, 26214, 27594, 24966, 23831, 24385, 25575, 25575,
25575, 25575, 24385, 23831, 23302, 23302, 24966, 24966,
24966, 23302, 23302, 21845, 22795, 24385, 24385, 22795,
21845, 21400, 21845, 23831, 21845, 21400, 10382, 10700,
10700, 10382, 10082, 9620, 10082, 9039, 9039, 8525, },
{ 8192, 65536, 65536, 61681, 61681, 61681, 41943, 41943,
41943, 41943, 40330, 41943, 40330, 41943, 40330, 40330,
40330, 38836, 38836, 40330, 40330, 24966, 27594, 26214,
26214, 26214, 27594, 24966, 23831, 24385, 25575, 25575,
25575, 25575, 24385, 23831, 11523, 11523, 12483, 12483,
12483, 11523, 11523, 10923, 11275, 12193, 12193, 11275,
10923, 5323, 5490, 5924, 5490, 5323, 5165, 5323,
5323, 5165, 5017, 4788, 5017, 4520, 4520, 4263, }
};
static const int dv_weight_720[2][64] = {
{ 8192, 65536, 65536, 61681, 61681, 61681, 58254, 58254,
58254, 58254, 58254, 58254, 55188, 58254, 58254, 55188,
55188, 55188, 55188, 55188, 55188, 24966, 27594, 26214,
26214, 26214, 27594, 24966, 23831, 24385, 25575, 25575,
25575, 25575, 24385, 23831, 15420, 15420, 16644, 16644,
16644, 15420, 15420, 10923, 11398, 12193, 12193, 11398,
10923, 10700, 10923, 11916, 10923, 10700, 5191, 5350,
5350, 5191, 5041, 4810, 5041, 4520, 4520, 4263, },
{ 8192, 43691, 43691, 40330, 40330, 40330, 29127, 29127,
29127, 29127, 29127, 29127, 27594, 29127, 29127, 27594,
27594, 27594, 27594, 27594, 27594, 12483, 13797, 13107,
13107, 13107, 13797, 12483, 11916, 12193, 12788, 12788,
12788, 12788, 12193, 11916, 5761, 5761, 6242, 6242,
6242, 5761, 5761, 5461, 5638, 5461, 6096, 5638,
5461, 2661, 2745, 2962, 2745, 2661, 2583, 2661,
2661, 2583, 2509, 2394, 2509, 2260, 2260, 2131, }
};
static av_always_inline int dv_set_class_number_sd(DVVideoContext *s,
int16_t *blk, EncBlockInfo *bi,
const uint8_t *zigzag_scan,
const int *weight, int bias)
{
int i, area;
/* We offer two different methods for class number assignment: the
* method suggested in SMPTE 314M Table 22, and an improved
* method. The SMPTE method is very conservative; it assigns class
* 3 (i.e. severe quantization) to any block where the largest AC
* component is greater than 36. FFmpeg's DV encoder tracks AC bit
* consumption precisely, so there is no need to bias most blocks
* towards strongly lossy compression. Instead, we assign class 2
* to most blocks, and use class 3 only when strictly necessary
* (for blocks whose largest AC component exceeds 255). */
#if 0 /* SMPTE spec method */
static const int classes[] = { 12, 24, 36, 0xffff };
#else /* improved FFmpeg method */
static const int classes[] = { -1, -1, 255, 0xffff };
#endif
int max = classes[0];
int prev = 0;
const unsigned deadzone = s->quant_deadzone;
const unsigned threshold = 2 * deadzone;
bi->mb[0] = blk[0];
for (area = 0; area < 4; area++) {
bi->prev[area] = prev;
bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
for (i = mb_area_start[area]; i < mb_area_start[area + 1]; i++) {
int level = blk[zigzag_scan[i]];
if (level + deadzone > threshold) {
bi->sign[i] = (level >> 31) & 1;
/* Weight it and shift down into range, adding for rounding.
* The extra division by a factor of 2^4 reverses the 8x
* expansion of the DCT AND the 2x doubling of the weights. */
level = (FFABS(level) * weight[i] + (1 << (dv_weight_bits + 3))) >>
(dv_weight_bits + 4);
if (!level)
continue;
bi->mb[i] = level;
if (level > max)
max = level;
bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, level);
bi->next[prev] = i;
prev = i;
}
}
}
bi->next[prev] = i;
for (bi->cno = 0; max > classes[bi->cno]; bi->cno++)
;
bi->cno += bias;
if (bi->cno >= 3) {
bi->cno = 3;
prev = 0;
i = bi->next[prev];
for (area = 0; area < 4; area++) {
bi->prev[area] = prev;
bi->bit_size[area] = 1; // 4 areas 4 bits for EOB :)
for (; i < mb_area_start[area + 1]; i = bi->next[i]) {
bi->mb[i] >>= 1;
if (bi->mb[i]) {
bi->bit_size[area] += dv_rl2vlc_size(i - prev - 1, bi->mb[i]);
bi->next[prev] = i;
prev = i;
}
}
}
bi->next[prev] = i;
}
return bi->bit_size[0] + bi->bit_size[1] +
bi->bit_size[2] + bi->bit_size[3];
}
/* this function just copies the DCT coefficients and performs
the initial (non-)quantization. */
static inline void dv_set_class_number_hd(DVVideoContext *s,
int16_t *blk, EncBlockInfo *bi,
const uint8_t *zigzag_scan,
const int *weight, int bias)
{
int i, max = 0;
/* the first quantization (none at all) */
bi->area_q[0] = 1;
/* weigh AC components and store to save[] */
/* (i=0 is the DC component; we only include it to make the
number of loop iterations even, for future possible SIMD optimization) */
for (i = 0; i < 64; i += 2) {
int level0, level1;
/* get the AC component (in zig-zag order) */
level0 = blk[zigzag_scan[i+0]];
level1 = blk[zigzag_scan[i+1]];
/* extract sign and make it the lowest bit */
bi->sign[i+0] = (level0>>31)&1;
bi->sign[i+1] = (level1>>31)&1;
/* take absolute value of the level */
level0 = FFABS(level0);
level1 = FFABS(level1);
/* weigh it */
level0 = (level0*weight[i+0] + 4096 + (1<<17)) >> 18;
level1 = (level1*weight[i+1] + 4096 + (1<<17)) >> 18;
/* save unquantized value */
bi->save[i+0] = level0;
bi->save[i+1] = level1;
/* find max component */
if (bi->save[i+0] > max)
max = bi->save[i+0];
if (bi->save[i+1] > max)
max = bi->save[i+1];
}
/* copy DC component */
bi->mb[0] = blk[0];
/* the EOB code is 4 bits */
bi->bit_size[0] = 4;
bi->bit_size[1] = bi->bit_size[2] = bi->bit_size[3] = 0;
/* ensure that no AC coefficients are cut off */
bi->min_qlevel = ((max+256) >> 8);
bi->area_q[0] = 25; /* set to an "impossible" value */
bi->cno = 0;
}
static av_always_inline int dv_init_enc_block(EncBlockInfo* bi, uint8_t *data, int linesize,
DVVideoContext *s, int chroma)
{
LOCAL_ALIGNED_16(int16_t, blk, [64]);
bi->area_q[0] = bi->area_q[1] = bi->area_q[2] = bi->area_q[3] = 0;
bi->partial_bit_count = 0;
bi->partial_bit_buffer = 0;
bi->cur_ac = 0;
if (data) {
if (DV_PROFILE_IS_HD(s->sys)) {
s->get_pixels(blk, data, linesize << bi->dct_mode);
s->fdct[0](blk);
} else {
bi->dct_mode = dv_guess_dct_mode(s, data, linesize);
s->get_pixels(blk, data, linesize);
s->fdct[bi->dct_mode](blk);
}
} else {
/* We rely on the fact that encoding all zeros leads to an immediate EOB,
which is precisely what the spec calls for in the "dummy" blocks. */
memset(blk, 0, 64*sizeof(*blk));
bi->dct_mode = 0;
}
if (DV_PROFILE_IS_HD(s->sys)) {
const int *weights;
if (s->sys->height == 1080) {
weights = dv_weight_1080[chroma];
} else { /* 720p */
weights = dv_weight_720[chroma];
}
dv_set_class_number_hd(s, blk, bi,
ff_zigzag_direct,
weights,
dv100_min_bias+chroma*dv100_chroma_bias);
} else {
dv_set_class_number_sd(s, blk, bi,
bi->dct_mode ? ff_dv_zigzag248_direct : ff_zigzag_direct,
bi->dct_mode ? dv_weight_248 : dv_weight_88,
chroma);
}
return bi->bit_size[0] + bi->bit_size[1] + bi->bit_size[2] + bi->bit_size[3];
}
/* DV100 quantize
Perform quantization by divinding the AC component by the qstep.
As an optimization we use a fixed-point integer multiply instead
of a divide. */
static av_always_inline int dv100_quantize(int level, int qsinv)
{
/* this code is equivalent to */
/* return (level + qs/2) / qs; */
return (level * qsinv + 1024 + (1<<(dv100_qstep_bits-1))) >> dv100_qstep_bits;
/* the extra +1024 is needed to make the rounding come out right. */
/* I (DJM) have verified that the results are exactly the same as
division for level 0-2048 at all QNOs. */
}
static int dv100_actual_quantize(EncBlockInfo *b, int qlevel)
{
int prev, k, qsinv;
int qno = DV100_QLEVEL_QNO(dv100_qlevels[qlevel]);
int cno = DV100_QLEVEL_CNO(dv100_qlevels[qlevel]);
if (b->area_q[0] == qno && b->cno == cno)
return b->bit_size[0];
qsinv = dv100_qstep_inv[qno];
/* record the new qstep */
b->area_q[0] = qno;
b->cno = cno;
/* reset encoded size (EOB = 4 bits) */
b->bit_size[0] = 4;
/* visit nonzero components and quantize */
prev = 0;
for (k = 1; k < 64; k++) {
/* quantize */
int ac = dv100_quantize(b->save[k], qsinv) >> cno;
if (ac) {
if (ac > 255)
ac = 255;
b->mb[k] = ac;
b->bit_size[0] += dv_rl2vlc_size(k - prev - 1, ac);
b->next[prev] = k;
prev = k;
}
}
b->next[prev] = k;
return b->bit_size[0];
}
static inline void dv_guess_qnos_hd(EncBlockInfo *blks, int *qnos)
{
EncBlockInfo *b;
int min_qlevel[5];
int qlevels[5];
int size[5];
int i, j;
/* cache block sizes at hypothetical qlevels */
uint16_t size_cache[5*8][DV100_NUM_QLEVELS] = {{0}};
/* get minimum qlevels */
for (i = 0; i < 5; i++) {
min_qlevel[i] = 1;
for (j = 0; j < 8; j++) {
if (blks[8*i+j].min_qlevel > min_qlevel[i])
min_qlevel[i] = blks[8*i+j].min_qlevel;
}
}
/* initialize sizes */
for (i = 0; i < 5; i++) {
qlevels[i] = dv100_starting_qno;
if (qlevels[i] < min_qlevel[i])
qlevels[i] = min_qlevel[i];
qnos[i] = DV100_QLEVEL_QNO(dv100_qlevels[qlevels[i]]);
size[i] = 0;
for (j = 0; j < 8; j++) {
size_cache[8*i+j][qlevels[i]] = dv100_actual_quantize(&blks[8*i+j], qlevels[i]);
size[i] += size_cache[8*i+j][qlevels[i]];
}
}
/* must we go coarser? */
if (size[0]+size[1]+size[2]+size[3]+size[4] > vs_total_ac_bits_hd) {
int largest = size[0] % 5; /* 'random' number */
int qlevels_done = 0;
do {
/* find the macroblock with the lowest qlevel */
for (i = 0; i < 5; i++) {
if (qlevels[i] < qlevels[largest])
largest = i;
}
i = largest;
/* ensure that we don't enter infinite loop */
largest = (largest+1) % 5;
/* quantize a little bit more */
qlevels[i] += dv100_qlevel_inc;
if (qlevels[i] > DV100_NUM_QLEVELS-1) {
qlevels[i] = DV100_NUM_QLEVELS-1;
qlevels_done++;
}
qnos[i] = DV100_QLEVEL_QNO(dv100_qlevels[qlevels[i]]);
size[i] = 0;
/* for each block */
b = &blks[8*i];
for (j = 0; j < 8; j++, b++) {
/* accumulate block size into macroblock */
if(size_cache[8*i+j][qlevels[i]] == 0) {
/* it is safe to use actual_quantize() here because we only go from finer to coarser,
and it saves the final actual_quantize() down below */
size_cache[8*i+j][qlevels[i]] = dv100_actual_quantize(b, qlevels[i]);
}
size[i] += size_cache[8*i+j][qlevels[i]];
} /* for each block */
} while (vs_total_ac_bits_hd < size[0] + size[1] + size[2] + size[3] + size[4] && qlevels_done < 5);
// can we go finer?
} else if (DV100_ENABLE_FINER &&
size[0]+size[1]+size[2]+size[3]+size[4] < vs_total_ac_bits_hd) {
int save_qlevel;
int largest = size[0] % 5; /* 'random' number */
while (qlevels[0] > min_qlevel[0] ||
qlevels[1] > min_qlevel[1] ||
qlevels[2] > min_qlevel[2] ||
qlevels[3] > min_qlevel[3] ||
qlevels[4] > min_qlevel[4]) {
/* find the macroblock with the highest qlevel */
for (i = 0; i < 5; i++) {
if (qlevels[i] > min_qlevel[i] && qlevels[i] > qlevels[largest])
largest = i;
}
i = largest;
/* ensure that we don't enter infinite loop */
largest = (largest+1) % 5;
if (qlevels[i] <= min_qlevel[i]) {
/* can't unquantize any more */
continue;
}
/* quantize a little bit less */
save_qlevel = qlevels[i];
qlevels[i] -= dv100_qlevel_inc;
if (qlevels[i] < min_qlevel[i])
qlevels[i] = min_qlevel[i];
qnos[i] = DV100_QLEVEL_QNO(dv100_qlevels[qlevels[i]]);
size[i] = 0;
/* for each block */
b = &blks[8*i];
for (j = 0; j < 8; j++, b++) {
/* accumulate block size into macroblock */
if(size_cache[8*i+j][qlevels[i]] == 0) {
size_cache[8*i+j][qlevels[i]] = dv100_actual_quantize(b, qlevels[i]);
}
size[i] += size_cache[8*i+j][qlevels[i]];
} /* for each block */
/* did we bust the limit? */
if (vs_total_ac_bits_hd < size[0] + size[1] + size[2] + size[3] + size[4]) {
/* go back down and exit */
qlevels[i] = save_qlevel;
qnos[i] = DV100_QLEVEL_QNO(dv100_qlevels[qlevels[i]]);
break;
}
}
}
/* now do the actual quantization */
for (i = 0; i < 5; i++) {
/* for each block */
b = &blks[8*i];
size[i] = 0;
for (j = 0; j < 8; j++, b++) {
/* accumulate block size into macroblock */
size[i] += dv100_actual_quantize(b, qlevels[i]);
} /* for each block */
}
}
static inline void dv_guess_qnos(EncBlockInfo *blks, int *qnos)
{
int size[5];
int i, j, k, a, prev, a2;
EncBlockInfo *b;
size[0] =
size[1] =
size[2] =
size[3] =
size[4] = 1 << 24;
do {
b = blks;
for (i = 0; i < 5; i++) {
if (!qnos[i])
continue;
qnos[i]--;
size[i] = 0;
for (j = 0; j < 6; j++, b++) {
for (a = 0; a < 4; a++) {
if (b->area_q[a] != ff_dv_quant_shifts[qnos[i] + ff_dv_quant_offset[b->cno]][a]) {
b->bit_size[a] = 1; // 4 areas 4 bits for EOB :)
b->area_q[a]++;
prev = b->prev[a];
av_assert2(b->next[prev] >= mb_area_start[a + 1] || b->mb[prev]);
for (k = b->next[prev]; k < mb_area_start[a + 1]; k = b->next[k]) {
b->mb[k] >>= 1;
if (b->mb[k]) {
b->bit_size[a] += dv_rl2vlc_size(k - prev - 1, b->mb[k]);
prev = k;
} else {
if (b->next[k] >= mb_area_start[a + 1] && b->next[k] < 64) {
for (a2 = a + 1; b->next[k] >= mb_area_start[a2 + 1]; a2++)
b->prev[a2] = prev;
av_assert2(a2 < 4);
av_assert2(b->mb[b->next[k]]);
b->bit_size[a2] += dv_rl2vlc_size(b->next[k] - prev - 1, b->mb[b->next[k]]) -
dv_rl2vlc_size(b->next[k] - k - 1, b->mb[b->next[k]]);
av_assert2(b->prev[a2] == k && (a2 + 1 >= 4 || b->prev[a2 + 1] != k));
b->prev[a2] = prev;
}
b->next[prev] = b->next[k];
}
}
b->prev[a + 1] = prev;
}
size[i] += b->bit_size[a];
}
}
if (vs_total_ac_bits >= size[0] + size[1] + size[2] + size[3] + size[4])
return;
}
} while (qnos[0] | qnos[1] | qnos[2] | qnos[3] | qnos[4]);
for (a = 2; a == 2 || vs_total_ac_bits < size[0]; a += a) {
b = blks;
size[0] = 5 * 6 * 4; // EOB
for (j = 0; j < 6 * 5; j++, b++) {
prev = b->prev[0];
for (k = b->next[prev]; k < 64; k = b->next[k]) {
if (b->mb[k] < a && b->mb[k] > -a) {
b->next[prev] = b->next[k];
} else {
size[0] += dv_rl2vlc_size(k - prev - 1, b->mb[k]);
prev = k;
}
}
}
}
}
/* update all cno values into the blocks, over-writing the old values without
touching anything else. (only used for DV100) */
static inline void dv_revise_cnos(uint8_t *dif, EncBlockInfo *blk, const AVDVProfile *profile)
{
uint8_t *data;
int mb_index, i;
for (mb_index = 0; mb_index < 5; mb_index++) {
data = dif + mb_index*80 + 4;
for (i = 0; i < profile->bpm; i++) {
/* zero out the class number */
data[1] &= 0xCF;
/* add the new one */
data[1] |= blk[profile->bpm*mb_index+i].cno << 4;
data += profile->block_sizes[i] >> 3;
}
}
}
static int dv_encode_video_segment(AVCodecContext *avctx, void *arg)
{
DVVideoContext *s = avctx->priv_data;
DVwork_chunk *work_chunk = arg;
int mb_index, i, j;
int mb_x, mb_y, c_offset;
ptrdiff_t linesize, y_stride;
uint8_t *y_ptr;
uint8_t *dif, *p;
LOCAL_ALIGNED_8(uint8_t, scratch, [128]);
EncBlockInfo enc_blks[5 * DV_MAX_BPM];
PutBitContext pbs[5 * DV_MAX_BPM];
PutBitContext *pb;
EncBlockInfo *enc_blk;
int vs_bit_size = 0;
int qnos[5];
int *qnosp = &qnos[0];
p = dif = &s->buf[work_chunk->buf_offset * 80];
enc_blk = &enc_blks[0];
for (mb_index = 0; mb_index < 5; mb_index++) {
dv_calculate_mb_xy(s, work_chunk, mb_index, &mb_x, &mb_y);
qnos[mb_index] = DV_PROFILE_IS_HD(s->sys) ? 1 : 15;
y_ptr = s->frame->data[0] + ((mb_y * s->frame->linesize[0] + mb_x) << 3);
linesize = s->frame->linesize[0];
if (s->sys->height == 1080 && mb_y < 134)
enc_blk->dct_mode = dv_guess_dct_mode(s, y_ptr, linesize);
else
enc_blk->dct_mode = 0;
for (i = 1; i < 8; i++)
enc_blk[i].dct_mode = enc_blk->dct_mode;
/* initializing luminance blocks */
if ((s->sys->pix_fmt == AV_PIX_FMT_YUV420P) ||
(s->sys->pix_fmt == AV_PIX_FMT_YUV411P && mb_x >= (704 / 8)) ||
(s->sys->height >= 720 && mb_y != 134)) {
y_stride = s->frame->linesize[0] << (3*!enc_blk->dct_mode);
} else {
y_stride = 16;
}
y_ptr = s->frame->data[0] +
((mb_y * s->frame->linesize[0] + mb_x) << 3);
linesize = s->frame->linesize[0];
if (s->sys->video_stype == 4) { /* SD 422 */
vs_bit_size +=
dv_init_enc_block(enc_blk + 0, y_ptr, linesize, s, 0) +
dv_init_enc_block(enc_blk + 1, NULL, linesize, s, 0) +
dv_init_enc_block(enc_blk + 2, y_ptr + 8, linesize, s, 0) +
dv_init_enc_block(enc_blk + 3, NULL, linesize, s, 0);
} else {
vs_bit_size +=
dv_init_enc_block(enc_blk + 0, y_ptr, linesize, s, 0) +
dv_init_enc_block(enc_blk + 1, y_ptr + 8, linesize, s, 0) +
dv_init_enc_block(enc_blk + 2, y_ptr + y_stride, linesize, s, 0) +
dv_init_enc_block(enc_blk + 3, y_ptr + 8 + y_stride, linesize, s, 0);
}
enc_blk += 4;
/* initializing chrominance blocks */
c_offset = (((mb_y >> (s->sys->pix_fmt == AV_PIX_FMT_YUV420P)) * s->frame->linesize[1] +
(mb_x >> ((s->sys->pix_fmt == AV_PIX_FMT_YUV411P) ? 2 : 1))) << 3);
for (j = 2; j; j--) {
uint8_t *c_ptr = s->frame->data[j] + c_offset;
linesize = s->frame->linesize[j];
y_stride = (mb_y == 134) ? 8 : (s->frame->linesize[j] << (3*!enc_blk->dct_mode));
if (s->sys->pix_fmt == AV_PIX_FMT_YUV411P && mb_x >= (704 / 8)) {
uint8_t *d;
uint8_t *b = scratch;
for (i = 0; i < 8; i++) {
d = c_ptr + (linesize << 3);
b[0] = c_ptr[0];
b[1] = c_ptr[1];
b[2] = c_ptr[2];
b[3] = c_ptr[3];
b[4] = d[0];
b[5] = d[1];
b[6] = d[2];
b[7] = d[3];
c_ptr += linesize;
b += 16;
}
c_ptr = scratch;
linesize = 16;
}
vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr, linesize, s, 1);
if (s->sys->bpm == 8)
vs_bit_size += dv_init_enc_block(enc_blk++, c_ptr + y_stride,
linesize, s, 1);
}
}
if (DV_PROFILE_IS_HD(s->sys)) {
/* unconditional */
dv_guess_qnos_hd(&enc_blks[0], qnosp);
} else if (vs_total_ac_bits < vs_bit_size) {
dv_guess_qnos(&enc_blks[0], qnosp);
}
/* DIF encoding process */
for (j = 0; j < 5 * s->sys->bpm;) {
int start_mb = j;
p[3] = *qnosp++;
p += 4;
/* First pass over individual cells only */
for (i = 0; i < s->sys->bpm; i++, j++) {
int sz = s->sys->block_sizes[i] >> 3;
init_put_bits(&pbs[j], p, sz);
put_sbits(&pbs[j], 9, ((enc_blks[j].mb[0] >> 3) - 1024 + 2) >> 2);
put_bits(&pbs[j], 1, DV_PROFILE_IS_HD(s->sys) && i ? 1 : enc_blks[j].dct_mode);
put_bits(&pbs[j], 2, enc_blks[j].cno);
dv_encode_ac(&enc_blks[j], &pbs[j], &pbs[j + 1]);
p += sz;
}
/* Second pass over each MB space */
pb = &pbs[start_mb];
for (i = 0; i < s->sys->bpm; i++)
if (enc_blks[start_mb + i].partial_bit_count)
pb = dv_encode_ac(&enc_blks[start_mb + i], pb,
&pbs[start_mb + s->sys->bpm]);
}
/* Third and final pass over the whole video segment space */
pb = &pbs[0];
for (j = 0; j < 5 * s->sys->bpm; j++) {
if (enc_blks[j].partial_bit_count)
pb = dv_encode_ac(&enc_blks[j], pb, &pbs[s->sys->bpm * 5]);
if (enc_blks[j].partial_bit_count)
av_log(avctx, AV_LOG_ERROR, "ac bitstream overflow\n");
}
for (j = 0; j < 5 * s->sys->bpm; j++) {
int pos;
int size = pbs[j].size_in_bits >> 3;
flush_put_bits(&pbs[j]);
pos = put_bits_count(&pbs[j]) >> 3;
if (pos > size) {
av_log(avctx, AV_LOG_ERROR,
"bitstream written beyond buffer size\n");
return -1;
}
memset(pbs[j].buf + pos, 0xff, size - pos);
}
if (DV_PROFILE_IS_HD(s->sys))
dv_revise_cnos(dif, enc_blks, s->sys);
return 0;
}
static inline int dv_write_pack(enum dv_pack_type pack_id, DVVideoContext *c,
uint8_t *buf)
{
/*