aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/dcaenc.c
blob: ff7b0eb715285b2e9170d0faa69dea899422bff1 (plain) (blame)
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
/*
 * DCA encoder
 * Copyright (C) 2008-2012 Alexander E. Patrakov
 *               2010 Benjamin Larsson
 *               2011 Xiang Wang
 *
 * 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
 */

#include "libavutil/avassert.h"
#include "libavutil/channel_layout.h"
#include "libavutil/common.h"
#include "libavutil/ffmath.h"
#include "avcodec.h"
#include "dca.h"
#include "dcadata.h"
#include "dcaenc.h"
#include "internal.h"
#include "mathops.h"
#include "put_bits.h"

#define MAX_CHANNELS 6
#define DCA_MAX_FRAME_SIZE 16384
#define DCA_HEADER_SIZE 13
#define DCA_LFE_SAMPLES 8

#define DCAENC_SUBBANDS 32
#define SUBFRAMES 1
#define SUBSUBFRAMES 2
#define SUBBAND_SAMPLES (SUBFRAMES * SUBSUBFRAMES * 8)
#define AUBANDS 25

typedef struct DCAEncContext {
    PutBitContext pb;
    int frame_size;
    int frame_bits;
    int fullband_channels;
    int channels;
    int lfe_channel;
    int samplerate_index;
    int bitrate_index;
    int channel_config;
    const int32_t *band_interpolation;
    const int32_t *band_spectrum;
    int lfe_scale_factor;
    softfloat lfe_quant;
    int32_t lfe_peak_cb;
    const int8_t *channel_order_tab;  ///< channel reordering table, lfe and non lfe

    int32_t history[MAX_CHANNELS][512]; /* This is a circular buffer */
    int32_t subband[MAX_CHANNELS][DCAENC_SUBBANDS][SUBBAND_SAMPLES];
    int32_t quantized[MAX_CHANNELS][DCAENC_SUBBANDS][SUBBAND_SAMPLES];
    int32_t peak_cb[MAX_CHANNELS][DCAENC_SUBBANDS];
    int32_t downsampled_lfe[DCA_LFE_SAMPLES];
    int32_t masking_curve_cb[SUBSUBFRAMES][256];
    int abits[MAX_CHANNELS][DCAENC_SUBBANDS];
    int scale_factor[MAX_CHANNELS][DCAENC_SUBBANDS];
    softfloat quant[MAX_CHANNELS][DCAENC_SUBBANDS];
    int32_t quant_index_sel[MAX_CHANNELS][DCA_CODE_BOOKS];
    int32_t eff_masking_curve_cb[256];
    int32_t band_masking_cb[32];
    int32_t worst_quantization_noise;
    int32_t worst_noise_ever;
    int consumed_bits;
} DCAEncContext;

static int32_t cos_table[2048];
static int32_t band_interpolation[2][512];
static int32_t band_spectrum[2][8];
static int32_t auf[9][AUBANDS][256];
static int32_t cb_to_add[256];
static int32_t cb_to_level[2048];
static int32_t lfe_fir_64i[512];

/* Transfer function of outer and middle ear, Hz -> dB */
static double hom(double f)
{
    double f1 = f / 1000;

    return -3.64 * pow(f1, -0.8)
           + 6.8 * exp(-0.6 * (f1 - 3.4) * (f1 - 3.4))
           - 6.0 * exp(-0.15 * (f1 - 8.7) * (f1 - 8.7))
           - 0.0006 * (f1 * f1) * (f1 * f1);
}

static double gammafilter(int i, double f)
{
    double h = (f - fc[i]) / erb[i];

    h = 1 + h * h;
    h = 1 / (h * h);
    return 20 * log10(h);
}

static int encode_init(AVCodecContext *avctx)
{
    DCAEncContext *c = avctx->priv_data;
    uint64_t layout = avctx->channel_layout;
    int i, j, min_frame_bits;

    c->fullband_channels = c->channels = avctx->channels;
    c->lfe_channel = (avctx->channels == 3 || avctx->channels == 6);
    c->band_interpolation = band_interpolation[1];
    c->band_spectrum = band_spectrum[1];
    c->worst_quantization_noise = -2047;
    c->worst_noise_ever = -2047;

    if (!layout) {
        av_log(avctx, AV_LOG_WARNING, "No channel layout specified. The "
                                      "encoder will guess the layout, but it "
                                      "might be incorrect.\n");
        layout = av_get_default_channel_layout(avctx->channels);
    }
    switch (layout) {
    case AV_CH_LAYOUT_MONO:         c->channel_config = 0; break;
    case AV_CH_LAYOUT_STEREO:       c->channel_config = 2; break;
    case AV_CH_LAYOUT_2_2:          c->channel_config = 8; break;
    case AV_CH_LAYOUT_5POINT0:      c->channel_config = 9; break;
    case AV_CH_LAYOUT_5POINT1:      c->channel_config = 9; break;
    default:
        av_log(avctx, AV_LOG_ERROR, "Unsupported channel layout!\n");
        return AVERROR_PATCHWELCOME;
    }

    if (c->lfe_channel) {
        c->fullband_channels--;
        c->channel_order_tab = channel_reorder_lfe[c->channel_config];
    } else {
        c->channel_order_tab = channel_reorder_nolfe[c->channel_config];
    }

    for (i = 0; i < MAX_CHANNELS; i++) {
        for (j = 0; j < DCA_CODE_BOOKS; j++) {
            c->quant_index_sel[i][j] = ff_dca_quant_index_group_size[j];
        }
    }

    for (i = 0; i < 9; i++) {
        if (sample_rates[i] == avctx->sample_rate)
            break;
    }
    if (i == 9)
        return AVERROR(EINVAL);
    c->samplerate_index = i;

    if (avctx->bit_rate < 32000 || avctx->bit_rate > 3840000) {
        av_log(avctx, AV_LOG_ERROR, "Bit rate %"PRId64" not supported.", (int64_t)avctx->bit_rate);
        return AVERROR(EINVAL);
    }
    for (i = 0; ff_dca_bit_rates[i] < avctx->bit_rate; i++)
        ;
    c->bitrate_index = i;
    c->frame_bits = FFALIGN((avctx->bit_rate * 512 + avctx->sample_rate - 1) / avctx->sample_rate, 32);
    min_frame_bits = 132 + (493 + 28 * 32) * c->fullband_channels + c->lfe_channel * 72;
    if (c->frame_bits < min_frame_bits || c->frame_bits > (DCA_MAX_FRAME_SIZE << 3))
        return AVERROR(EINVAL);

    c->frame_size = (c->frame_bits + 7) / 8;

    avctx->frame_size = 32 * SUBBAND_SAMPLES;

    if (!cos_table[0]) {
        int j, k;

        cos_table[0] = 0x7fffffff;
        cos_table[512] = 0;
        cos_table[1024] = -cos_table[0];
        for (i = 1; i < 512; i++) {
            cos_table[i]   = (int32_t)(0x7fffffff * cos(M_PI * i / 1024));
            cos_table[1024-i] = -cos_table[i];
            cos_table[1024+i] = -cos_table[i];
            cos_table[2048-i] = cos_table[i];
        }
        for (i = 0; i < 2048; i++) {
            cb_to_level[i] = (int32_t)(0x7fffffff * ff_exp10(-0.005 * i));
        }

        for (k = 0; k < 32; k++) {
            for (j = 0; j < 8; j++) {
                lfe_fir_64i[64 * j + k] = (int32_t)(0xffffff800000ULL * ff_dca_lfe_fir_64[8 * k + j]);
                lfe_fir_64i[64 * (7-j) + (63 - k)] = (int32_t)(0xffffff800000ULL * ff_dca_lfe_fir_64[8 * k + j]);
            }
        }

        for (i = 0; i < 512; i++) {
            band_interpolation[0][i] = (int32_t)(0x1000000000ULL * ff_dca_fir_32bands_perfect[i]);
            band_interpolation[1][i] = (int32_t)(0x1000000000ULL * ff_dca_fir_32bands_nonperfect[i]);
        }

        for (i = 0; i < 9; i++) {
            for (j = 0; j < AUBANDS; j++) {
                for (k = 0; k < 256; k++) {
                    double freq = sample_rates[i] * (k + 0.5) / 512;

                    auf[i][j][k] = (int32_t)(10 * (hom(freq) + gammafilter(j, freq)));
                }
            }
        }

        for (i = 0; i < 256; i++) {
            double add = 1 + ff_exp10(-0.01 * i);
            cb_to_add[i] = (int32_t)(100 * log10(add));
        }
        for (j = 0; j < 8; j++) {
            double accum = 0;
            for (i = 0; i < 512; i++) {
                double reconst = ff_dca_fir_32bands_perfect[i] * ((i & 64) ? (-1) : 1);
                accum += reconst * cos(2 * M_PI * (i + 0.5 - 256) * (j + 0.5) / 512);
            }
            band_spectrum[0][j] = (int32_t)(200 * log10(accum));
        }
        for (j = 0; j < 8; j++) {
            double accum = 0;
            for (i = 0; i < 512; i++) {
                double reconst = ff_dca_fir_32bands_nonperfect[i] * ((i & 64) ? (-1) : 1);
                accum += reconst * cos(2 * M_PI * (i + 0.5 - 256) * (j + 0.5) / 512);
            }
            band_spectrum[1][j] = (int32_t)(200 * log10(accum));
        }
    }
    return 0;
}

static inline int32_t cos_t(int x)
{
    return cos_table[x & 2047];
}

static inline int32_t sin_t(int x)
{
    return cos_t(x - 512);
}

static inline int32_t half32(int32_t a)
{
    return (a + 1) >> 1;
}

static inline int32_t mul32(int32_t a, int32_t b)
{
    int64_t r = (int64_t)a * b + 0x80000000ULL;
    return r >> 32;
}

static void subband_transform(DCAEncContext *c, const int32_t *input)
{
    int ch, subs, i, k, j;

    for (ch = 0; ch < c->fullband_channels; ch++) {
        /* History is copied because it is also needed for PSY */
        int32_t hist[512];
        int hist_start = 0;
        const int chi = c->channel_order_tab[ch];

        memcpy(hist, &c->history[ch][0], 512 * sizeof(int32_t));

        for (subs = 0; subs < SUBBAND_SAMPLES; subs++) {
            int32_t accum[64];
            int32_t resp;
            int band;

            /* Calculate the convolutions at once */
            memset(accum, 0, 64 * sizeof(int32_t));

            for (k = 0, i = hist_start, j = 0;
                    i < 512; k = (k + 1) & 63, i++, j++)
                accum[k] += mul32(hist[i], c->band_interpolation[j]);
            for (i = 0; i < hist_start; k = (k + 1) & 63, i++, j++)
                accum[k] += mul32(hist[i], c->band_interpolation[j]);

            for (k = 16; k < 32; k++)
                accum[k] = accum[k] - accum[31 - k];
            for (k = 32; k < 48; k++)
                accum[k] = accum[k] + accum[95 - k];

            for (band = 0; band < 32; band++) {
                resp = 0;
                for (i = 16; i < 48; i++) {
                    int s = (2 * band + 1) * (2 * (i + 16) + 1);
                    resp += mul32(accum[i], cos_t(s << 3)) >> 3;
                }

                c->subband[ch][band][subs] = ((band + 1) & 2) ? -resp : resp;
            }

            /* Copy in 32 new samples from input */
            for (i = 0; i < 32; i++)
                hist[i + hist_start] = input[(subs * 32 + i) * c->channels + chi];

            hist_start = (hist_start + 32) & 511;
        }
    }
}

static void lfe_downsample(DCAEncContext *c, const int32_t *input)
{
    /* FIXME: make 128x LFE downsampling possible */
    const int lfech = lfe_index[c->channel_config];
    int i, j, lfes;
    int32_t hist[512];
    int32_t accum;
    int hist_start = 0;

    memcpy(hist, &c->history[c->channels - 1][0], 512 * sizeof(int32_t));

    for (lfes = 0; lfes < DCA_LFE_SAMPLES; lfes++) {
        /* Calculate the convolution */
        accum = 0;

        for (i = hist_start, j = 0; i < 512; i++, j++)
            accum += mul32(hist[i], lfe_fir_64i[j]);
        for (i = 0; i < hist_start; i++, j++)
            accum += mul32(hist[i], lfe_fir_64i[j]);

        c->downsampled_lfe[lfes] = accum;

        /* Copy in 64 new samples from input */
        for (i = 0; i < 64; i++)
            hist[i + hist_start] = input[(lfes * 64 + i) * c->channels + lfech];

        hist_start = (hist_start + 64) & 511;
    }
}

typedef struct {
    int32_t re;
    int32_t im;
} cplx32;

static void fft(const int32_t in[2 * 256], cplx32 out[256])
{
    cplx32 buf[256], rin[256], rout[256];
    int i, j, k, l;

    /* do two transforms in parallel */
    for (i = 0; i < 256; i++) {
        /* Apply the Hann window */
        rin[i].re = mul32(in[2 * i], 0x3fffffff - (cos_t(8 * i + 2) >> 1));
        rin[i].im = mul32(in[2 * i + 1], 0x3fffffff - (cos_t(8 * i + 6) >> 1));
    }
    /* pre-rotation */
    for (i = 0; i < 256; i++) {
        buf[i].re = mul32(cos_t(4 * i + 2), rin[i].re)
                  - mul32(sin_t(4 * i + 2), rin[i].im);
        buf[i].im = mul32(cos_t(4 * i + 2), rin[i].im)
                  + mul32(sin_t(4 * i + 2), rin[i].re);
    }

    for (j = 256, l = 1; j != 1; j >>= 1, l <<= 1) {
        for (k = 0; k < 256; k += j) {
            for (i = k; i < k + j / 2; i++) {
                cplx32 sum, diff;
                int t = 8 * l * i;

                sum.re = buf[i].re + buf[i + j / 2].re;
                sum.im = buf[i].im + buf[i + j / 2].im;

                diff.re = buf[i].re - buf[i + j / 2].re;
                diff.im = buf[i].im - buf[i + j / 2].im;

                buf[i].re = half32(sum.re);
                buf[i].im = half32(sum.im);

                buf[i + j / 2].re = mul32(diff.re, cos_t(t))
                                  - mul32(diff.im, sin_t(t));
                buf[i + j / 2].im = mul32(diff.im, cos_t(t))
                                  + mul32(diff.re, sin_t(t));
            }
        }
    }
    /* post-rotation */
    for (i = 0; i < 256; i++) {
        int b = ff_reverse[i];
        rout[i].re = mul32(buf[b].re, cos_t(4 * i))
                   - mul32(buf[b].im, sin_t(4 * i));
        rout[i].im = mul32(buf[b].im, cos_t(4 * i))
                   + mul32(buf[b].re, sin_t(4 * i));
    }
    for (i = 0; i < 256; i++) {
        /* separate the results of the two transforms */
        cplx32 o1, o2;

        o1.re =  rout[i].re - rout[255 - i].re;
        o1.im =  rout[i].im + rout[255 - i].im;

        o2.re =  rout[i].im - rout[255 - i].im;
        o2.im = -rout[i].re - rout[255 - i].re;

        /* combine them into one long transform */
        out[i].re = mul32( o1.re + o2.re, cos_t(2 * i + 1))
                  + mul32( o1.im - o2.im, sin_t(2 * i + 1));
        out[i].im = mul32( o1.im + o2.im, cos_t(2 * i + 1))
                  + mul32(-o1.re + o2.re, sin_t(2 * i + 1));
    }
}

static int32_t get_cb(int32_t in)
{
    int i, res;

    res = 0;
    if (in < 0)
        in = -in;
    for (i = 1024; i > 0; i >>= 1) {
        if (cb_to_level[i + res] >= in)
            res += i;
    }
    return -res;
}

static int32_t add_cb(int32_t a, int32_t b)
{
    if (a < b)
        FFSWAP(int32_t, a, b);

    if (a - b >= 256)
        return a;
    return a + cb_to_add[a - b];
}

static void adjust_jnd(int samplerate_index,
                       const int32_t in[512], int32_t out_cb[256])
{
    int32_t power[256];
    cplx32 out[256];
    int32_t out_cb_unnorm[256];
    int32_t denom;
    const int32_t ca_cb = -1114;
    const int32_t cs_cb = 928;
    int i, j;

    fft(in, out);

    for (j = 0; j < 256; j++) {
        power[j] = add_cb(get_cb(out[j].re), get_cb(out[j].im));
        out_cb_unnorm[j] = -2047; /* and can only grow */
    }

    for (i = 0; i < AUBANDS; i++) {
        denom = ca_cb; /* and can only grow */
        for (j = 0; j < 256; j++)
            denom = add_cb(denom, power[j] + auf[samplerate_index][i][j]);
        for (j = 0; j < 256; j++)
            out_cb_unnorm[j] = add_cb(out_cb_unnorm[j],
                    -denom + auf[samplerate_index][i][j]);
    }

    for (j = 0; j < 256; j++)
        out_cb[j] = add_cb(out_cb[j], -out_cb_unnorm[j] - ca_cb - cs_cb);
}

typedef void (*walk_band_t)(DCAEncContext *c, int band1, int band2, int f,
                            int32_t spectrum1, int32_t spectrum2, int channel,
                            int32_t * arg);

static void walk_band_low(DCAEncContext *c, int band, int channel,
                          walk_band_t walk, int32_t *arg)
{
    int f;

    if (band == 0) {
        for (f = 0; f < 4; f++)
            walk(c, 0, 0, f, 0, -2047, channel, arg);
    } else {
        for (f = 0; f < 8; f++)
            walk(c, band, band - 1, 8 * band - 4 + f,
                    c->band_spectrum[7 - f], c->band_spectrum[f], channel, arg);
    }
}

static void walk_band_high(DCAEncContext *c, int band, int channel,
                           walk_band_t walk, int32_t *arg)
{
    int f;

    if (band == 31) {
        for (f = 0; f < 4; f++)
            walk(c, 31, 31, 256 - 4 + f, 0, -2047, channel, arg);
    } else {
        for (f = 0; f < 8; f++)
            walk(c, band, band + 1, 8 * band + 4 + f,
                    c->band_spectrum[f], c->band_spectrum[7 - f], channel, arg);
    }
}

static void update_band_masking(DCAEncContext *c, int band1, int band2,
                                int f, int32_t spectrum1, int32_t spectrum2,
                                int channel, int32_t * arg)
{
    int32_t value = c->eff_masking_curve_cb[f] - spectrum1;

    if (value < c->band_masking_cb[band1])
        c->band_masking_cb[band1] = value;
}

static void calc_masking(DCAEncContext *c, const int32_t *input)
{
    int i, k, band, ch, ssf;
    int32_t data[512];

    for (i = 0; i < 256; i++)
        for (ssf = 0; ssf < SUBSUBFRAMES; ssf++)
            c->masking_curve_cb[ssf][i] = -2047;

    for (ssf = 0; ssf < SUBSUBFRAMES; ssf++)
        for (ch = 0; ch < c->fullband_channels; ch++) {
            const int chi = c->channel_order_tab[ch];

            for (i = 0, k = 128 + 256 * ssf; k < 512; i++, k++)
                data[i] = c->history[ch][k];
            for (k -= 512; i < 512; i++, k++)
                data[i] = input[k * c->channels + chi];
            adjust_jnd(c->samplerate_index, data, c->masking_curve_cb[ssf]);
        }
    for (i = 0; i < 256; i++) {
        int32_t m = 2048;

        for (ssf = 0; ssf < SUBSUBFRAMES; ssf++)
            if (c->masking_curve_cb[ssf][i] < m)
                m = c->masking_curve_cb[ssf][i];
        c->eff_masking_curve_cb[i] = m;
    }

    for (band = 0; band < 32; band++) {
        c->band_masking_cb[band] = 2048;
        walk_band_low(c, band, 0, update_band_masking, NULL);
        walk_band_high(c, band, 0, update_band_masking, NULL);
    }
}

static void find_peaks(DCAEncContext *c)
{
    int band, ch;

    for (ch = 0; ch < c->fullband_channels; ch++)
        for (band = 0; band < 32; band++) {
            int sample;
            int32_t m = 0;

            for (sample = 0; sample < SUBBAND_SAMPLES; sample++) {
                int32_t s = abs(c->subband[ch][band][sample]);
                if (m < s)
                    m = s;
            }
            c->peak_cb[ch][band] = get_cb(m);
        }

    if (c->lfe_channel) {
        int sample;
        int32_t m = 0;

        for (sample = 0; sample < DCA_LFE_SAMPLES; sample++)
            if (m < abs(c->downsampled_lfe[sample]))
                m = abs(c->downsampled_lfe[sample]);
        c->lfe_peak_cb = get_cb(m);
    }
}

static const int snr_fudge = 128;
#define USED_1ABITS 1
#define USED_NABITS 2
#define USED_26ABITS 4

static int32_t quantize_value(int32_t value, softfloat quant)
{
    int32_t offset = 1 << (quant.e - 1);

    value = mul32(value, quant.m) + offset;
    value = value >> quant.e;
    return value;
}

static int calc_one_scale(int32_t peak_cb, int abits, softfloat *quant)
{
    int32_t peak;
    int our_nscale, try_remove;
    softfloat our_quant;

    av_assert0(peak_cb <= 0);
    av_assert0(peak_cb >= -2047);

    our_nscale = 127;
    peak = cb_to_level[-peak_cb];

    for (try_remove = 64; try_remove > 0; try_remove >>= 1) {
        if (scalefactor_inv[our_nscale - try_remove].e + stepsize_inv[abits].e <= 17)
            continue;
        our_quant.m = mul32(scalefactor_inv[our_nscale - try_remove].m, stepsize_inv[abits].m);
        our_quant.e = scalefactor_inv[our_nscale - try_remove].e + stepsize_inv[abits].e - 17;
        if ((ff_dca_quant_levels[abits] - 1) / 2 < quantize_value(peak, our_quant))
            continue;
        our_nscale -= try_remove;
    }

    if (our_nscale >= 125)
        our_nscale = 124;

    quant->m = mul32(scalefactor_inv[our_nscale].m, stepsize_inv[abits].m);
    quant->e = scalefactor_inv[our_nscale].e + stepsize_inv[abits].e - 17;
    av_assert0((ff_dca_quant_levels[abits] - 1) / 2 >= quantize_value(peak, *quant));

    return our_nscale;
}

static void quantize_all(DCAEncContext *c)
{
    int sample, band, ch;

    for (ch = 0; ch < c->fullband_channels; ch++)
        for (band = 0; band < 32; band++)
            for (sample = 0; sample < SUBBAND_SAMPLES; sample++)
                c->quantized[ch][band][sample] = quantize_value(c->subband[ch][band][sample], c->quant[ch][band]);
}

static void accumulate_huff_bit_consumption(int abits, int32_t *quantized, uint32_t *result)
{
    uint8_t sel, id = abits - 1;
    for (sel = 0; sel < ff_dca_quant_index_group_size[id]; sel++)
        result[sel] += ff_dca_vlc_calc_quant_bits(quantized, SUBBAND_SAMPLES, sel, id);
}

static uint32_t set_best_code(uint32_t vlc_bits[DCA_CODE_BOOKS][7], uint32_t clc_bits[DCA_CODE_BOOKS], int32_t res[DCA_CODE_BOOKS])
{
    uint8_t i, sel;
    uint32_t best_sel_bits[DCA_CODE_BOOKS];
    int32_t best_sel_id[DCA_CODE_BOOKS];
    uint32_t t, bits = 0;

    for (i = 0; i < DCA_CODE_BOOKS; i++) {

        av_assert0(!((!!vlc_bits[i][0]) ^ (!!clc_bits[i])));
        if (vlc_bits[i][0] == 0) {
            /* do not transmit adjustment index for empty codebooks */
            res[i] = ff_dca_quant_index_group_size[i];
            /* and skip it */
            continue;
        }

        best_sel_bits[i] = vlc_bits[i][0];
        best_sel_id[i] = 0;
        for (sel = 0; sel < ff_dca_quant_index_group_size[i]; sel++) {
            if (best_sel_bits[i] > vlc_bits[i][sel] && vlc_bits[i][sel]) {
                best_sel_bits[i] = vlc_bits[i][sel];
                best_sel_id[i] = sel;
            }
        }

        /* 2 bits to transmit scale factor adjustment index */
        t = best_sel_bits[i] + 2;
        if (t < clc_bits[i]) {
            res[i] = best_sel_id[i];
            bits += t;
        } else {
            res[i] = ff_dca_quant_index_group_size[i];
            bits += clc_bits[i];
        }
    }
    return bits;
}

static int init_quantization_noise(DCAEncContext *c, int noise)
{
    int ch, band, ret = 0;
    uint32_t huff_bit_count_accum[MAX_CHANNELS][DCA_CODE_BOOKS][7];
    uint32_t clc_bit_count_accum[MAX_CHANNELS][DCA_CODE_BOOKS];
    uint32_t bits_counter = 0;

    c->consumed_bits = 132 + 493 * c->fullband_channels;
    if (c->lfe_channel)
        c->consumed_bits += 72;

    /* attempt to guess the bit distribution based on the prevoius frame */
    for (ch = 0; ch < c->fullband_channels; ch++) {
        for (band = 0; band < 32; band++) {
            int snr_cb = c->peak_cb[ch][band] - c->band_masking_cb[band] - noise;

            if (snr_cb >= 1312) {
                c->abits[ch][band] = 26;
                ret |= USED_26ABITS;
            } else if (snr_cb >= 222) {
                c->abits[ch][band] = 8 + mul32(snr_cb - 222, 69000000);
                ret |= USED_NABITS;
            } else if (snr_cb >= 0) {
                c->abits[ch][band] = 2 + mul32(snr_cb, 106000000);
                ret |= USED_NABITS;
            } else {
                c->abits[ch][band] = 1;
                ret |= USED_1ABITS;
            }
        }
    }

    /* Recalc scale_factor each time to get bits consumption in case of Huffman coding.
       It is suboptimal solution */
    /* TODO: May be cache scaled values */
    for (ch = 0; ch < c->fullband_channels; ch++) {
        for (band = 0; band < 32; band++) {
            c->scale_factor[ch][band] = calc_one_scale(c->peak_cb[ch][band],
                                                       c->abits[ch][band],
                                                       &c->quant[ch][band]);
        }
    }
    quantize_all(c);

    memset(huff_bit_count_accum, 0, MAX_CHANNELS * DCA_CODE_BOOKS * 7 * sizeof(uint32_t));
    memset(clc_bit_count_accum, 0, MAX_CHANNELS * DCA_CODE_BOOKS * sizeof(uint32_t));
    for (ch = 0; ch < c->fullband_channels; ch++) {
        for (band = 0; band < 32; band++) {
            if (c->abits[ch][band] && c->abits[ch][band] <= DCA_CODE_BOOKS) {
                accumulate_huff_bit_consumption(c->abits[ch][band], c->quantized[ch][band], huff_bit_count_accum[ch][c->abits[ch][band] - 1]);
                clc_bit_count_accum[ch][c->abits[ch][band] - 1] += bit_consumption[c->abits[ch][band]];
            } else {
                bits_counter += bit_consumption[c->abits[ch][band]];
            }
        }
    }

    for (ch = 0; ch < c->fullband_channels; ch++) {
        bits_counter += set_best_code(huff_bit_count_accum[ch], clc_bit_count_accum[ch], c->quant_index_sel[ch]);
    }

    c->consumed_bits += bits_counter;

    return ret;
}

static void assign_bits(DCAEncContext *c)
{
    /* Find the bounds where the binary search should work */
    int low, high, down;
    int used_abits = 0;

    init_quantization_noise(c, c->worst_quantization_noise);
    low = high = c->worst_quantization_noise;
    if (c->consumed_bits > c->frame_bits) {
        while (c->consumed_bits > c->frame_bits) {
            av_assert0(used_abits != USED_1ABITS);
            low = high;
            high += snr_fudge;
            used_abits = init_quantization_noise(c, high);
        }
    } else {
        while (c->consumed_bits <= c->frame_bits) {
            high = low;
            if (used_abits == USED_26ABITS)
                goto out; /* The requested bitrate is too high, pad with zeros */
            low -= snr_fudge;
            used_abits = init_quantization_noise(c, low);
        }
    }

    /* Now do a binary search between low and high to see what fits */
    for (down = snr_fudge >> 1; down; down >>= 1) {
        init_quantization_noise(c, high - down);
        if (c->consumed_bits <= c->frame_bits)
            high -= down;
    }
    init_quantization_noise(c, high);
out:
    c->worst_quantization_noise = high;
    if (high > c->worst_noise_ever)
        c->worst_noise_ever = high;
}

static void shift_history(DCAEncContext *c, const int32_t *input)
{
    int k, ch;

    for (k = 0; k < 512; k++)
        for (ch = 0; ch < c->channels; ch++) {
            const int chi = c->channel_order_tab[ch];

            c->history[ch][k] = input[k * c->channels + chi];
        }
}

static void calc_lfe_scales(DCAEncContext *c)
{
    if (c->lfe_channel)
        c->lfe_scale_factor = calc_one_scale(c->lfe_peak_cb, 11, &c->lfe_quant);
}

static void put_frame_header(DCAEncContext *c)
{
    /* SYNC */
    put_bits(&c->pb, 16, 0x7ffe);
    put_bits(&c->pb, 16, 0x8001);

    /* Frame type: normal */
    put_bits(&c->pb, 1, 1);

    /* Deficit sample count: none */
    put_bits(&c->pb, 5, 31);

    /* CRC is not present */
    put_bits(&c->pb, 1, 0);

    /* Number of PCM sample blocks */
    put_bits(&c->pb, 7, SUBBAND_SAMPLES - 1);

    /* Primary frame byte size */
    put_bits(&c->pb, 14, c->frame_size - 1);

    /* Audio channel arrangement */
    put_bits(&c->pb, 6, c->channel_config);

    /* Core audio sampling frequency */
    put_bits(&c->pb, 4, bitstream_sfreq[c->samplerate_index]);

    /* Transmission bit rate */
    put_bits(&c->pb, 5, c->bitrate_index);

    /* Embedded down mix: disabled */
    put_bits(&c->pb, 1, 0);

    /* Embedded dynamic range flag: not present */
    put_bits(&c->pb, 1, 0);

    /* Embedded time stamp flag: not present */
    put_bits(&c->pb, 1, 0);

    /* Auxiliary data flag: not present */
    put_bits(&c->pb, 1, 0);

    /* HDCD source: no */
    put_bits(&c->pb, 1, 0);

    /* Extension audio ID: N/A */
    put_bits(&c->pb, 3, 0);

    /* Extended audio data: not present */
    put_bits(&c->pb, 1, 0);

    /* Audio sync word insertion flag: after each sub-frame */
    put_bits(&c->pb, 1, 0);

    /* Low frequency effects flag: not present or 64x subsampling */
    put_bits(&c->pb, 2, c->lfe_channel ? 2 : 0);

    /* Predictor history switch flag: on */
    put_bits(&c->pb, 1, 1);

    /* No CRC */
    /* Multirate interpolator switch: non-perfect reconstruction */
    put_bits(&c->pb, 1, 0);

    /* Encoder software revision: 7 */
    put_bits(&c->pb, 4, 7);

    /* Copy history: 0 */
    put_bits(&c->pb, 2, 0);

    /* Source PCM resolution: 16 bits, not DTS ES */
    put_bits(&c->pb, 3, 0);

    /* Front sum/difference coding: no */
    put_bits(&c->pb, 1, 0);

    /* Surrounds sum/difference coding: no */
    put_bits(&c->pb, 1, 0);

    /* Dialog normalization: 0 dB */
    put_bits(&c->pb, 4, 0);
}

static void put_primary_audio_header(DCAEncContext *c)
{
    int ch, i;
    /* Number of subframes */
    put_bits(&c->pb, 4, SUBFRAMES - 1);

    /* Number of primary audio channels */
    put_bits(&c->pb, 3, c->fullband_channels - 1);

    /* Subband activity count */
    for (ch = 0; ch < c->fullband_channels; ch++)
        put_bits(&c->pb, 5, DCAENC_SUBBANDS - 2);

    /* High frequency VQ start subband */
    for (ch = 0; ch < c->fullband_channels; ch++)
        put_bits(&c->pb, 5, DCAENC_SUBBANDS - 1);

    /* Joint intensity coding index: 0, 0 */
    for (ch = 0; ch < c->fullband_channels; ch++)
        put_bits(&c->pb, 3, 0);

    /* Transient mode codebook: A4, A4 (arbitrary) */
    for (ch = 0; ch < c->fullband_channels; ch++)
        put_bits(&c->pb, 2, 0);

    /* Scale factor code book: 7 bit linear, 7-bit sqrt table (for each channel) */
    for (ch = 0; ch < c->fullband_channels; ch++)
        put_bits(&c->pb, 3, 6);

    /* Bit allocation quantizer select: linear 5-bit */
    for (ch = 0; ch < c->fullband_channels; ch++)
        put_bits(&c->pb, 3, 6);

    /* Quantization index codebook select */
    for (i = 0; i < DCA_CODE_BOOKS; i++)
        for (ch = 0; ch < c->fullband_channels; ch++)
            put_bits(&c->pb, ff_dca_quant_index_sel_nbits[i], c->quant_index_sel[ch][i]);

    /* Scale factor adjustment index: transmitted in case of Huffman coding */
    for (i = 0; i < DCA_CODE_BOOKS; i++)
        for (ch = 0; ch < c->fullband_channels; ch++)
            if (c->quant_index_sel[ch][i] < ff_dca_quant_index_group_size[i])
                put_bits(&c->pb, 2, 0);

    /* Audio header CRC check word: not transmitted */
}

static void put_subframe_samples(DCAEncContext *c, int ss, int band, int ch)
{
    int i, j, sum, bits, sel;
    if (c->abits[ch][band] <= DCA_CODE_BOOKS) {
        av_assert0(c->abits[ch][band] > 0);
        sel = c->quant_index_sel[ch][c->abits[ch][band] - 1];
        // Huffman codes
        if (sel < ff_dca_quant_index_group_size[c->abits[ch][band] - 1]) {
            ff_dca_vlc_enc_quant(&c->pb, &c->quantized[ch][band][ss * 8], 8, sel, c->abits[ch][band] - 1);
            return;
        }

        // Block codes
        if (c->abits[ch][band] <= 7) {
            for (i = 0; i < 8; i += 4) {
                sum = 0;
                for (j = 3; j >= 0; j--) {
                    sum *= ff_dca_quant_levels[c->abits[ch][band]];
                    sum += c->quantized[ch][band][ss * 8 + i + j];
                    sum += (ff_dca_quant_levels[c->abits[ch][band]] - 1) / 2;
                }
                put_bits(&c->pb, bit_consumption[c->abits[ch][band]] / 4, sum);
            }
            return;
        }
    }

    for (i = 0; i < 8; i++) {
        bits = bit_consumption[c->abits[ch][band]] / 16;
        put_sbits(&c->pb, bits, c->quantized[ch][band][ss * 8 + i]);
    }
}

static void put_subframe(DCAEncContext *c, int subframe)
{
    int i, band, ss, ch;

    /* Subsubframes count */
    put_bits(&c->pb, 2, SUBSUBFRAMES -1);

    /* Partial subsubframe sample count: dummy */
    put_bits(&c->pb, 3, 0);

    /* Prediction mode: no ADPCM, in each channel and subband */
    for (ch = 0; ch < c->fullband_channels; ch++)
        for (band = 0; band < DCAENC_SUBBANDS; band++)
            put_bits(&c->pb, 1, 0);

    /* Prediction VQ address: not transmitted */
    /* Bit allocation index */
    for (ch = 0; ch < c->fullband_channels; ch++)
        for (band = 0; band < DCAENC_SUBBANDS; band++)
            put_bits(&c->pb, 5, c->abits[ch][band]);

    if (SUBSUBFRAMES > 1) {
        /* Transition mode: none for each channel and subband */
        for (ch = 0; ch < c->fullband_channels; ch++)
            for (band = 0; band < DCAENC_SUBBANDS; band++)
                put_bits(&c->pb, 1, 0); /* codebook A4 */
    }

    /* Scale factors */
    for (ch = 0; ch < c->fullband_channels; ch++)
        for (band = 0; band < DCAENC_SUBBANDS; band++)
            put_bits(&c->pb, 7, c->scale_factor[ch][band]);

    /* Joint subband scale factor codebook select: not transmitted */
    /* Scale factors for joint subband coding: not transmitted */
    /* Stereo down-mix coefficients: not transmitted */
    /* Dynamic range coefficient: not transmitted */
    /* Stde information CRC check word: not transmitted */
    /* VQ encoded high frequency subbands: not transmitted */

    /* LFE data: 8 samples and scalefactor */
    if (c->lfe_channel) {
        for (i = 0; i < DCA_LFE_SAMPLES; i++)
            put_bits(&c->pb, 8, quantize_value(c->downsampled_lfe[i], c->lfe_quant) & 0xff);
        put_bits(&c->pb, 8, c->lfe_scale_factor);
    }

    /* Audio data (subsubframes) */
    for (ss = 0; ss < SUBSUBFRAMES ; ss++)
        for (ch = 0; ch < c->fullband_channels; ch++)
            for (band = 0; band < DCAENC_SUBBANDS; band++)
                    put_subframe_samples(c, ss, band, ch);

    /* DSYNC */
    put_bits(&c->pb, 16, 0xffff);
}

static int encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
                        const AVFrame *frame, int *got_packet_ptr)
{
    DCAEncContext *c = avctx->priv_data;
    const int32_t *samples;
    int ret, i;

    if ((ret = ff_alloc_packet2(avctx, avpkt, c->frame_size, 0)) < 0)
        return ret;

    samples = (const int32_t *)frame->data[0];

    subband_transform(c, samples);
    if (c->lfe_channel)
        lfe_downsample(c, samples);

    calc_masking(c, samples);
    find_peaks(c);
    assign_bits(c);
    calc_lfe_scales(c);
    shift_history(c, samples);

    init_put_bits(&c->pb, avpkt->data, avpkt->size);
    put_frame_header(c);
    put_primary_audio_header(c);
    for (i = 0; i < SUBFRAMES; i++)
        put_subframe(c, i);


    for (i = put_bits_count(&c->pb); i < 8*c->frame_size; i++)
        put_bits(&c->pb, 1, 0);

    flush_put_bits(&c->pb);

    avpkt->pts      = frame->pts;
    avpkt->duration = ff_samples_to_time_base(avctx, frame->nb_samples);
    avpkt->size     = put_bits_count(&c->pb) >> 3;
    *got_packet_ptr = 1;
    return 0;
}

static const AVCodecDefault defaults[] = {
    { "b",          "1411200" },
    { NULL },
};

AVCodec ff_dca_encoder = {
    .name                  = "dca",
    .long_name             = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
    .type                  = AVMEDIA_TYPE_AUDIO,
    .id                    = AV_CODEC_ID_DTS,
    .priv_data_size        = sizeof(DCAEncContext),
    .init                  = encode_init,
    .encode2               = encode_frame,
    .capabilities          = AV_CODEC_CAP_EXPERIMENTAL,
    .sample_fmts           = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_S32,
                                                            AV_SAMPLE_FMT_NONE },
    .supported_samplerates = sample_rates,
    .channel_layouts       = (const uint64_t[]) { AV_CH_LAYOUT_MONO,
                                                  AV_CH_LAYOUT_STEREO,
                                                  AV_CH_LAYOUT_2_2,
                                                  AV_CH_LAYOUT_5POINT0,
                                                  AV_CH_LAYOUT_5POINT1,
                                                  0 },
    .defaults              = defaults,
};