aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/apedec.c
blob: f3635836d0ce3f4679dfbe934e2594158106686b (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
/*
 * Monkey's Audio lossless audio decoder
 * Copyright (c) 2007 Benjamin Zores <ben@geexbox.org>
 *  based upon libdemac from Dave Chapman.
 *
 * This file is part of Libav.
 *
 * Libav 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.
 *
 * Libav 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 Libav; 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/opt.h"
#include "avcodec.h"
#include "dsputil.h"
#include "bytestream.h"
#include "internal.h"

/**
 * @file
 * Monkey's Audio lossless audio decoder
 */

#define MAX_CHANNELS        2
#define MAX_BYTESPERSAMPLE  3

#define APE_FRAMECODE_MONO_SILENCE    1
#define APE_FRAMECODE_STEREO_SILENCE  3
#define APE_FRAMECODE_PSEUDO_STEREO   4

#define HISTORY_SIZE 512
#define PREDICTOR_ORDER 8
/** Total size of all predictor histories */
#define PREDICTOR_SIZE 50

#define YDELAYA (18 + PREDICTOR_ORDER*4)
#define YDELAYB (18 + PREDICTOR_ORDER*3)
#define XDELAYA (18 + PREDICTOR_ORDER*2)
#define XDELAYB (18 + PREDICTOR_ORDER)

#define YADAPTCOEFFSA 18
#define XADAPTCOEFFSA 14
#define YADAPTCOEFFSB 10
#define XADAPTCOEFFSB 5

/**
 * Possible compression levels
 * @{
 */
enum APECompressionLevel {
    COMPRESSION_LEVEL_FAST       = 1000,
    COMPRESSION_LEVEL_NORMAL     = 2000,
    COMPRESSION_LEVEL_HIGH       = 3000,
    COMPRESSION_LEVEL_EXTRA_HIGH = 4000,
    COMPRESSION_LEVEL_INSANE     = 5000
};
/** @} */

#define APE_FILTER_LEVELS 3

/** Filter orders depending on compression level */
static const uint16_t ape_filter_orders[5][APE_FILTER_LEVELS] = {
    {  0,   0,    0 },
    { 16,   0,    0 },
    { 64,   0,    0 },
    { 32, 256,    0 },
    { 16, 256, 1280 }
};

/** Filter fraction bits depending on compression level */
static const uint8_t ape_filter_fracbits[5][APE_FILTER_LEVELS] = {
    {  0,  0,  0 },
    { 11,  0,  0 },
    { 11,  0,  0 },
    { 10, 13,  0 },
    { 11, 13, 15 }
};


/** Filters applied to the decoded data */
typedef struct APEFilter {
    int16_t *coeffs;        ///< actual coefficients used in filtering
    int16_t *adaptcoeffs;   ///< adaptive filter coefficients used for correcting of actual filter coefficients
    int16_t *historybuffer; ///< filter memory
    int16_t *delay;         ///< filtered values

    int avg;
} APEFilter;

typedef struct APERice {
    uint32_t k;
    uint32_t ksum;
} APERice;

typedef struct APERangecoder {
    uint32_t low;           ///< low end of interval
    uint32_t range;         ///< length of interval
    uint32_t help;          ///< bytes_to_follow resp. intermediate value
    unsigned int buffer;    ///< buffer for input/output
} APERangecoder;

/** Filter histories */
typedef struct APEPredictor {
    int32_t *buf;

    int32_t lastA[2];

    int32_t filterA[2];
    int32_t filterB[2];

    int32_t coeffsA[2][4];  ///< adaption coefficients
    int32_t coeffsB[2][5];  ///< adaption coefficients
    int32_t historybuffer[HISTORY_SIZE + PREDICTOR_SIZE];
} APEPredictor;

/** Decoder context */
typedef struct APEContext {
    AVClass *class;                          ///< class for AVOptions
    AVCodecContext *avctx;
    DSPContext dsp;
    int channels;
    int samples;                             ///< samples left to decode in current frame
    int bps;

    int fileversion;                         ///< codec version, very important in decoding process
    int compression_level;                   ///< compression levels
    int fset;                                ///< which filter set to use (calculated from compression level)
    int flags;                               ///< global decoder flags

    uint32_t CRC;                            ///< frame CRC
    int frameflags;                          ///< frame flags
    APEPredictor predictor;                  ///< predictor used for final reconstruction

    int32_t *decoded_buffer;
    int decoded_size;
    int32_t *decoded[MAX_CHANNELS];          ///< decoded data for each channel
    int blocks_per_loop;                     ///< maximum number of samples to decode for each call

    int16_t* filterbuf[APE_FILTER_LEVELS];   ///< filter memory

    APERangecoder rc;                        ///< rangecoder used to decode actual values
    APERice riceX;                           ///< rice code parameters for the second channel
    APERice riceY;                           ///< rice code parameters for the first channel
    APEFilter filters[APE_FILTER_LEVELS][2]; ///< filters used for reconstruction

    uint8_t *data;                           ///< current frame data
    uint8_t *data_end;                       ///< frame data end
    int data_size;                           ///< frame data allocated size
    const uint8_t *ptr;                      ///< current position in frame data

    int error;
} APEContext;

// TODO: dsputilize

static av_cold int ape_decode_close(AVCodecContext *avctx)
{
    APEContext *s = avctx->priv_data;
    int i;

    for (i = 0; i < APE_FILTER_LEVELS; i++)
        av_freep(&s->filterbuf[i]);

    av_freep(&s->decoded_buffer);
    av_freep(&s->data);
    s->decoded_size = s->data_size = 0;

    return 0;
}

static av_cold int ape_decode_init(AVCodecContext *avctx)
{
    APEContext *s = avctx->priv_data;
    int i;

    if (avctx->extradata_size != 6) {
        av_log(avctx, AV_LOG_ERROR, "Incorrect extradata\n");
        return AVERROR(EINVAL);
    }
    if (avctx->channels > 2) {
        av_log(avctx, AV_LOG_ERROR, "Only mono and stereo is supported\n");
        return AVERROR(EINVAL);
    }
    s->bps = avctx->bits_per_coded_sample;
    switch (s->bps) {
    case 8:
        avctx->sample_fmt = AV_SAMPLE_FMT_U8P;
        break;
    case 16:
        avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
        break;
    case 24:
        avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
        break;
    default:
        av_log_ask_for_sample(avctx, "Unsupported bits per coded sample %d\n",
                              s->bps);
        return AVERROR_PATCHWELCOME;
    }
    s->avctx             = avctx;
    s->channels          = avctx->channels;
    s->fileversion       = AV_RL16(avctx->extradata);
    s->compression_level = AV_RL16(avctx->extradata + 2);
    s->flags             = AV_RL16(avctx->extradata + 4);

    av_log(avctx, AV_LOG_DEBUG, "Compression Level: %d - Flags: %d\n",
           s->compression_level, s->flags);
    if (s->compression_level % 1000 || s->compression_level > COMPRESSION_LEVEL_INSANE) {
        av_log(avctx, AV_LOG_ERROR, "Incorrect compression level %d\n",
               s->compression_level);
        return AVERROR_INVALIDDATA;
    }
    s->fset = s->compression_level / 1000 - 1;
    for (i = 0; i < APE_FILTER_LEVELS; i++) {
        if (!ape_filter_orders[s->fset][i])
            break;
        FF_ALLOC_OR_GOTO(avctx, s->filterbuf[i],
                         (ape_filter_orders[s->fset][i] * 3 + HISTORY_SIZE) * 4,
                         filter_alloc_fail);
    }

    ff_dsputil_init(&s->dsp, avctx);
    avctx->channel_layout = (avctx->channels==2) ? AV_CH_LAYOUT_STEREO : AV_CH_LAYOUT_MONO;

    return 0;
filter_alloc_fail:
    ape_decode_close(avctx);
    return AVERROR(ENOMEM);
}

/**
 * @name APE range decoding functions
 * @{
 */

#define CODE_BITS    32
#define TOP_VALUE    ((unsigned int)1 << (CODE_BITS-1))
#define SHIFT_BITS   (CODE_BITS - 9)
#define EXTRA_BITS   ((CODE_BITS-2) % 8 + 1)
#define BOTTOM_VALUE (TOP_VALUE >> 8)

/** Start the decoder */
static inline void range_start_decoding(APEContext *ctx)
{
    ctx->rc.buffer = bytestream_get_byte(&ctx->ptr);
    ctx->rc.low    = ctx->rc.buffer >> (8 - EXTRA_BITS);
    ctx->rc.range  = (uint32_t) 1 << EXTRA_BITS;
}

/** Perform normalization */
static inline void range_dec_normalize(APEContext *ctx)
{
    while (ctx->rc.range <= BOTTOM_VALUE) {
        ctx->rc.buffer <<= 8;
        if(ctx->ptr < ctx->data_end) {
            ctx->rc.buffer += *ctx->ptr;
            ctx->ptr++;
        } else {
            ctx->error = 1;
        }
        ctx->rc.low    = (ctx->rc.low << 8)    | ((ctx->rc.buffer >> 1) & 0xFF);
        ctx->rc.range  <<= 8;
    }
}

/**
 * Calculate culmulative frequency for next symbol. Does NO update!
 * @param ctx decoder context
 * @param tot_f is the total frequency or (code_value)1<<shift
 * @return the culmulative frequency
 */
static inline int range_decode_culfreq(APEContext *ctx, int tot_f)
{
    range_dec_normalize(ctx);
    ctx->rc.help = ctx->rc.range / tot_f;
    return ctx->rc.low / ctx->rc.help;
}

/**
 * Decode value with given size in bits
 * @param ctx decoder context
 * @param shift number of bits to decode
 */
static inline int range_decode_culshift(APEContext *ctx, int shift)
{
    range_dec_normalize(ctx);
    ctx->rc.help = ctx->rc.range >> shift;
    return ctx->rc.low / ctx->rc.help;
}


/**
 * Update decoding state
 * @param ctx decoder context
 * @param sy_f the interval length (frequency of the symbol)
 * @param lt_f the lower end (frequency sum of < symbols)
 */
static inline void range_decode_update(APEContext *ctx, int sy_f, int lt_f)
{
    ctx->rc.low  -= ctx->rc.help * lt_f;
    ctx->rc.range = ctx->rc.help * sy_f;
}

/** Decode n bits (n <= 16) without modelling */
static inline int range_decode_bits(APEContext *ctx, int n)
{
    int sym = range_decode_culshift(ctx, n);
    range_decode_update(ctx, 1, sym);
    return sym;
}


#define MODEL_ELEMENTS 64

/**
 * Fixed probabilities for symbols in Monkey Audio version 3.97
 */
static const uint16_t counts_3970[22] = {
        0, 14824, 28224, 39348, 47855, 53994, 58171, 60926,
    62682, 63786, 64463, 64878, 65126, 65276, 65365, 65419,
    65450, 65469, 65480, 65487, 65491, 65493,
};

/**
 * Probability ranges for symbols in Monkey Audio version 3.97
 */
static const uint16_t counts_diff_3970[21] = {
    14824, 13400, 11124, 8507, 6139, 4177, 2755, 1756,
    1104, 677, 415, 248, 150, 89, 54, 31,
    19, 11, 7, 4, 2,
};

/**
 * Fixed probabilities for symbols in Monkey Audio version 3.98
 */
static const uint16_t counts_3980[22] = {
        0, 19578, 36160, 48417, 56323, 60899, 63265, 64435,
    64971, 65232, 65351, 65416, 65447, 65466, 65476, 65482,
    65485, 65488, 65490, 65491, 65492, 65493,
};

/**
 * Probability ranges for symbols in Monkey Audio version 3.98
 */
static const uint16_t counts_diff_3980[21] = {
    19578, 16582, 12257, 7906, 4576, 2366, 1170, 536,
    261, 119, 65, 31, 19, 10, 6, 3,
    3, 2, 1, 1, 1,
};

/**
 * Decode symbol
 * @param ctx decoder context
 * @param counts probability range start position
 * @param counts_diff probability range widths
 */
static inline int range_get_symbol(APEContext *ctx,
                                   const uint16_t counts[],
                                   const uint16_t counts_diff[])
{
    int symbol, cf;

    cf = range_decode_culshift(ctx, 16);

    if(cf > 65492){
        symbol= cf - 65535 + 63;
        range_decode_update(ctx, 1, cf);
        if(cf > 65535)
            ctx->error=1;
        return symbol;
    }
    /* figure out the symbol inefficiently; a binary search would be much better */
    for (symbol = 0; counts[symbol + 1] <= cf; symbol++);

    range_decode_update(ctx, counts_diff[symbol], counts[symbol]);

    return symbol;
}
/** @} */ // group rangecoder

static inline void update_rice(APERice *rice, unsigned int x)
{
    int lim = rice->k ? (1 << (rice->k + 4)) : 0;
    rice->ksum += ((x + 1) / 2) - ((rice->ksum + 16) >> 5);

    if (rice->ksum < lim)
        rice->k--;
    else if (rice->ksum >= (1 << (rice->k + 5)))
        rice->k++;
}

static inline int ape_decode_value(APEContext *ctx, APERice *rice)
{
    unsigned int x, overflow;

    if (ctx->fileversion < 3990) {
        int tmpk;

        overflow = range_get_symbol(ctx, counts_3970, counts_diff_3970);

        if (overflow == (MODEL_ELEMENTS - 1)) {
            tmpk = range_decode_bits(ctx, 5);
            overflow = 0;
        } else
            tmpk = (rice->k < 1) ? 0 : rice->k - 1;

        if (tmpk <= 16)
            x = range_decode_bits(ctx, tmpk);
        else if (tmpk <= 32) {
            x = range_decode_bits(ctx, 16);
            x |= (range_decode_bits(ctx, tmpk - 16) << 16);
        } else {
            av_log(ctx->avctx, AV_LOG_ERROR, "Too many bits: %d\n", tmpk);
            return AVERROR_INVALIDDATA;
        }
        x += overflow << tmpk;
    } else {
        int base, pivot;

        pivot = rice->ksum >> 5;
        if (pivot == 0)
            pivot = 1;

        overflow = range_get_symbol(ctx, counts_3980, counts_diff_3980);

        if (overflow == (MODEL_ELEMENTS - 1)) {
            overflow  = range_decode_bits(ctx, 16) << 16;
            overflow |= range_decode_bits(ctx, 16);
        }

        if (pivot < 0x10000) {
            base = range_decode_culfreq(ctx, pivot);
            range_decode_update(ctx, 1, base);
        } else {
            int base_hi = pivot, base_lo;
            int bbits = 0;

            while (base_hi & ~0xFFFF) {
                base_hi >>= 1;
                bbits++;
            }
            base_hi = range_decode_culfreq(ctx, base_hi + 1);
            range_decode_update(ctx, 1, base_hi);
            base_lo = range_decode_culfreq(ctx, 1 << bbits);
            range_decode_update(ctx, 1, base_lo);

            base = (base_hi << bbits) + base_lo;
        }

        x = base + overflow * pivot;
    }

    update_rice(rice, x);

    /* Convert to signed */
    if (x & 1)
        return (x >> 1) + 1;
    else
        return -(x >> 1);
}

static void entropy_decode(APEContext *ctx, int blockstodecode, int stereo)
{
    int32_t *decoded0 = ctx->decoded[0];
    int32_t *decoded1 = ctx->decoded[1];

    while (blockstodecode--) {
        *decoded0++ = ape_decode_value(ctx, &ctx->riceY);
        if (stereo)
            *decoded1++ = ape_decode_value(ctx, &ctx->riceX);
    }
}

static int init_entropy_decoder(APEContext *ctx)
{
    /* Read the CRC */
    if (ctx->data_end - ctx->ptr < 6)
        return AVERROR_INVALIDDATA;
    ctx->CRC = bytestream_get_be32(&ctx->ptr);

    /* Read the frame flags if they exist */
    ctx->frameflags = 0;
    if ((ctx->fileversion > 3820) && (ctx->CRC & 0x80000000)) {
        ctx->CRC &= ~0x80000000;

        if (ctx->data_end - ctx->ptr < 6)
            return AVERROR_INVALIDDATA;
        ctx->frameflags = bytestream_get_be32(&ctx->ptr);
    }

    /* Initialize the rice structs */
    ctx->riceX.k = 10;
    ctx->riceX.ksum = (1 << ctx->riceX.k) * 16;
    ctx->riceY.k = 10;
    ctx->riceY.ksum = (1 << ctx->riceY.k) * 16;

    /* The first 8 bits of input are ignored. */
    ctx->ptr++;

    range_start_decoding(ctx);

    return 0;
}

static const int32_t initial_coeffs[4] = {
    360, 317, -109, 98
};

static void init_predictor_decoder(APEContext *ctx)
{
    APEPredictor *p = &ctx->predictor;

    /* Zero the history buffers */
    memset(p->historybuffer, 0, PREDICTOR_SIZE * sizeof(*p->historybuffer));
    p->buf = p->historybuffer;

    /* Initialize and zero the coefficients */
    memcpy(p->coeffsA[0], initial_coeffs, sizeof(initial_coeffs));
    memcpy(p->coeffsA[1], initial_coeffs, sizeof(initial_coeffs));
    memset(p->coeffsB, 0, sizeof(p->coeffsB));

    p->filterA[0] = p->filterA[1] = 0;
    p->filterB[0] = p->filterB[1] = 0;
    p->lastA[0]   = p->lastA[1]   = 0;
}

/** Get inverse sign of integer (-1 for positive, 1 for negative and 0 for zero) */
static inline int APESIGN(int32_t x) {
    return (x < 0) - (x > 0);
}

static av_always_inline int predictor_update_filter(APEPredictor *p,
                                                    const int decoded, const int filter,
                                                    const int delayA,  const int delayB,
                                                    const int adaptA,  const int adaptB)
{
    int32_t predictionA, predictionB, sign;

    p->buf[delayA]     = p->lastA[filter];
    p->buf[adaptA]     = APESIGN(p->buf[delayA]);
    p->buf[delayA - 1] = p->buf[delayA] - p->buf[delayA - 1];
    p->buf[adaptA - 1] = APESIGN(p->buf[delayA - 1]);

    predictionA = p->buf[delayA    ] * p->coeffsA[filter][0] +
                  p->buf[delayA - 1] * p->coeffsA[filter][1] +
                  p->buf[delayA - 2] * p->coeffsA[filter][2] +
                  p->buf[delayA - 3] * p->coeffsA[filter][3];

    /*  Apply a scaled first-order filter compression */
    p->buf[delayB]     = p->filterA[filter ^ 1] - ((p->filterB[filter] * 31) >> 5);
    p->buf[adaptB]     = APESIGN(p->buf[delayB]);
    p->buf[delayB - 1] = p->buf[delayB] - p->buf[delayB - 1];
    p->buf[adaptB - 1] = APESIGN(p->buf[delayB - 1]);
    p->filterB[filter] = p->filterA[filter ^ 1];

    predictionB = p->buf[delayB    ] * p->coeffsB[filter][0] +
                  p->buf[delayB - 1] * p->coeffsB[filter][1] +
                  p->buf[delayB - 2] * p->coeffsB[filter][2] +
                  p->buf[delayB - 3] * p->coeffsB[filter][3] +
                  p->buf[delayB - 4] * p->coeffsB[filter][4];

    p->lastA[filter] = decoded + ((predictionA + (predictionB >> 1)) >> 10);
    p->filterA[filter] = p->lastA[filter] + ((p->filterA[filter] * 31) >> 5);

    sign = APESIGN(decoded);
    p->coeffsA[filter][0] += p->buf[adaptA    ] * sign;
    p->coeffsA[filter][1] += p->buf[adaptA - 1] * sign;
    p->coeffsA[filter][2] += p->buf[adaptA - 2] * sign;
    p->coeffsA[filter][3] += p->buf[adaptA - 3] * sign;
    p->coeffsB[filter][0] += p->buf[adaptB    ] * sign;
    p->coeffsB[filter][1] += p->buf[adaptB - 1] * sign;
    p->coeffsB[filter][2] += p->buf[adaptB - 2] * sign;
    p->coeffsB[filter][3] += p->buf[adaptB - 3] * sign;
    p->coeffsB[filter][4] += p->buf[adaptB - 4] * sign;

    return p->filterA[filter];
}

static void predictor_decode_stereo(APEContext *ctx, int count)
{
    APEPredictor *p = &ctx->predictor;
    int32_t *decoded0 = ctx->decoded[0];
    int32_t *decoded1 = ctx->decoded[1];

    while (count--) {
        /* Predictor Y */
        *decoded0 = predictor_update_filter(p, *decoded0, 0, YDELAYA, YDELAYB,
                                            YADAPTCOEFFSA, YADAPTCOEFFSB);
        decoded0++;
        *decoded1 = predictor_update_filter(p, *decoded1, 1, XDELAYA, XDELAYB,
                                            XADAPTCOEFFSA, XADAPTCOEFFSB);
        decoded1++;

        /* Combined */
        p->buf++;

        /* Have we filled the history buffer? */
        if (p->buf == p->historybuffer + HISTORY_SIZE) {
            memmove(p->historybuffer, p->buf,
                    PREDICTOR_SIZE * sizeof(*p->historybuffer));
            p->buf = p->historybuffer;
        }
    }
}

static void predictor_decode_mono(APEContext *ctx, int count)
{
    APEPredictor *p = &ctx->predictor;
    int32_t *decoded0 = ctx->decoded[0];
    int32_t predictionA, currentA, A, sign;

    currentA = p->lastA[0];

    while (count--) {
        A = *decoded0;

        p->buf[YDELAYA] = currentA;
        p->buf[YDELAYA - 1] = p->buf[YDELAYA] - p->buf[YDELAYA - 1];

        predictionA = p->buf[YDELAYA    ] * p->coeffsA[0][0] +
                      p->buf[YDELAYA - 1] * p->coeffsA[0][1] +
                      p->buf[YDELAYA - 2] * p->coeffsA[0][2] +
                      p->buf[YDELAYA - 3] * p->coeffsA[0][3];

        currentA = A + (predictionA >> 10);

        p->buf[YADAPTCOEFFSA]     = APESIGN(p->buf[YDELAYA    ]);
        p->buf[YADAPTCOEFFSA - 1] = APESIGN(p->buf[YDELAYA - 1]);

        sign = APESIGN(A);
        p->coeffsA[0][0] += p->buf[YADAPTCOEFFSA    ] * sign;
        p->coeffsA[0][1] += p->buf[YADAPTCOEFFSA - 1] * sign;
        p->coeffsA[0][2] += p->buf[YADAPTCOEFFSA - 2] * sign;
        p->coeffsA[0][3] += p->buf[YADAPTCOEFFSA - 3] * sign;

        p->buf++;

        /* Have we filled the history buffer? */
        if (p->buf == p->historybuffer + HISTORY_SIZE) {
            memmove(p->historybuffer, p->buf,
                    PREDICTOR_SIZE * sizeof(*p->historybuffer));
            p->buf = p->historybuffer;
        }

        p->filterA[0] = currentA + ((p->filterA[0] * 31) >> 5);
        *(decoded0++) = p->filterA[0];
    }

    p->lastA[0] = currentA;
}

static void do_init_filter(APEFilter *f, int16_t *buf, int order)
{
    f->coeffs = buf;
    f->historybuffer = buf + order;
    f->delay       = f->historybuffer + order * 2;
    f->adaptcoeffs = f->historybuffer + order;

    memset(f->historybuffer, 0, (order * 2) * sizeof(*f->historybuffer));
    memset(f->coeffs, 0, order * sizeof(*f->coeffs));
    f->avg = 0;
}

static void init_filter(APEContext *ctx, APEFilter *f, int16_t *buf, int order)
{
    do_init_filter(&f[0], buf, order);
    do_init_filter(&f[1], buf + order * 3 + HISTORY_SIZE, order);
}

static void do_apply_filter(APEContext *ctx, int version, APEFilter *f,
                            int32_t *data, int count, int order, int fracbits)
{
    int res;
    int absres;

    while (count--) {
        /* round fixedpoint scalar product */
        res = ctx->dsp.scalarproduct_and_madd_int16(f->coeffs, f->delay - order,
                                                    f->adaptcoeffs - order,
                                                    order, APESIGN(*data));
        res = (res + (1 << (fracbits - 1))) >> fracbits;
        res += *data;
        *data++ = res;

        /* Update the output history */
        *f->delay++ = av_clip_int16(res);

        if (version < 3980) {
            /* Version ??? to < 3.98 files (untested) */
            f->adaptcoeffs[0]  = (res == 0) ? 0 : ((res >> 28) & 8) - 4;
            f->adaptcoeffs[-4] >>= 1;
            f->adaptcoeffs[-8] >>= 1;
        } else {
            /* Version 3.98 and later files */

            /* Update the adaption coefficients */
            absres = FFABS(res);
            if (absres)
                *f->adaptcoeffs = ((res & (-1<<31)) ^ (-1<<30)) >>
                                  (25 + (absres <= f->avg*3) + (absres <= f->avg*4/3));
            else
                *f->adaptcoeffs = 0;

            f->avg += (absres - f->avg) / 16;

            f->adaptcoeffs[-1] >>= 1;
            f->adaptcoeffs[-2] >>= 1;
            f->adaptcoeffs[-8] >>= 1;
        }

        f->adaptcoeffs++;

        /* Have we filled the history buffer? */
        if (f->delay == f->historybuffer + HISTORY_SIZE + (order * 2)) {
            memmove(f->historybuffer, f->delay - (order * 2),
                    (order * 2) * sizeof(*f->historybuffer));
            f->delay = f->historybuffer + order * 2;
            f->adaptcoeffs = f->historybuffer + order;
        }
    }
}

static void apply_filter(APEContext *ctx, APEFilter *f,
                         int32_t *data0, int32_t *data1,
                         int count, int order, int fracbits)
{
    do_apply_filter(ctx, ctx->fileversion, &f[0], data0, count, order, fracbits);
    if (data1)
        do_apply_filter(ctx, ctx->fileversion, &f[1], data1, count, order, fracbits);
}

static void ape_apply_filters(APEContext *ctx, int32_t *decoded0,
                              int32_t *decoded1, int count)
{
    int i;

    for (i = 0; i < APE_FILTER_LEVELS; i++) {
        if (!ape_filter_orders[ctx->fset][i])
            break;
        apply_filter(ctx, ctx->filters[i], decoded0, decoded1, count,
                     ape_filter_orders[ctx->fset][i],
                     ape_filter_fracbits[ctx->fset][i]);
    }
}

static int init_frame_decoder(APEContext *ctx)
{
    int i, ret;
    if ((ret = init_entropy_decoder(ctx)) < 0)
        return ret;
    init_predictor_decoder(ctx);

    for (i = 0; i < APE_FILTER_LEVELS; i++) {
        if (!ape_filter_orders[ctx->fset][i])
            break;
        init_filter(ctx, ctx->filters[i], ctx->filterbuf[i],
                    ape_filter_orders[ctx->fset][i]);
    }
    return 0;
}

static void ape_unpack_mono(APEContext *ctx, int count)
{
    if (ctx->frameflags & APE_FRAMECODE_STEREO_SILENCE) {
        /* We are pure silence, so we're done. */
        av_log(ctx->avctx, AV_LOG_DEBUG, "pure silence mono\n");
        return;
    }

    entropy_decode(ctx, count, 0);
    ape_apply_filters(ctx, ctx->decoded[0], NULL, count);

    /* Now apply the predictor decoding */
    predictor_decode_mono(ctx, count);

    /* Pseudo-stereo - just copy left channel to right channel */
    if (ctx->channels == 2) {
        memcpy(ctx->decoded[1], ctx->decoded[0], count * sizeof(*ctx->decoded[1]));
    }
}

static void ape_unpack_stereo(APEContext *ctx, int count)
{
    int32_t left, right;
    int32_t *decoded0 = ctx->decoded[0];
    int32_t *decoded1 = ctx->decoded[1];

    if (ctx->frameflags & APE_FRAMECODE_STEREO_SILENCE) {
        /* We are pure silence, so we're done. */
        av_log(ctx->avctx, AV_LOG_DEBUG, "pure silence stereo\n");
        return;
    }

    entropy_decode(ctx, count, 1);
    ape_apply_filters(ctx, decoded0, decoded1, count);

    /* Now apply the predictor decoding */
    predictor_decode_stereo(ctx, count);

    /* Decorrelate and scale to output depth */
    while (count--) {
        left = *decoded1 - (*decoded0 / 2);
        right = left + *decoded0;

        *(decoded0++) = left;
        *(decoded1++) = right;
    }
}

static int ape_decode_frame(AVCodecContext *avctx, void *data,
                            int *got_frame_ptr, AVPacket *avpkt)
{
    AVFrame *frame     = data;
    const uint8_t *buf = avpkt->data;
    APEContext *s = avctx->priv_data;
    uint8_t *sample8;
    int16_t *sample16;
    int32_t *sample24;
    int i, ch, ret;
    int blockstodecode;
    int bytes_used = 0;

    /* this should never be negative, but bad things will happen if it is, so
       check it just to make sure. */
    av_assert0(s->samples >= 0);

    if(!s->samples){
        uint32_t nblocks, offset;
        int buf_size;

        if (!avpkt->size) {
            *got_frame_ptr = 0;
            return 0;
        }
        if (avpkt->size < 8) {
            av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
            return AVERROR_INVALIDDATA;
        }
        buf_size = avpkt->size & ~3;
        if (buf_size != avpkt->size) {
            av_log(avctx, AV_LOG_WARNING, "packet size is not a multiple of 4. "
                   "extra bytes at the end will be skipped.\n");
        }

        av_fast_malloc(&s->data, &s->data_size, buf_size);
        if (!s->data)
            return AVERROR(ENOMEM);
        s->dsp.bswap_buf((uint32_t*)s->data, (const uint32_t*)buf, buf_size >> 2);
        s->ptr = s->data;
        s->data_end = s->data + buf_size;

        nblocks = bytestream_get_be32(&s->ptr);
        offset  = bytestream_get_be32(&s->ptr);
        if (offset > 3) {
            av_log(avctx, AV_LOG_ERROR, "Incorrect offset passed\n");
            s->data = NULL;
            return AVERROR_INVALIDDATA;
        }
        if (s->data_end - s->ptr < offset) {
            av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
            return AVERROR_INVALIDDATA;
        }
        s->ptr += offset;

        if (!nblocks || nblocks > INT_MAX) {
            av_log(avctx, AV_LOG_ERROR, "Invalid sample count: %u.\n", nblocks);
            return AVERROR_INVALIDDATA;
        }
        s->samples = nblocks;

        /* Initialize the frame decoder */
        if (init_frame_decoder(s) < 0) {
            av_log(avctx, AV_LOG_ERROR, "Error reading frame header\n");
            return AVERROR_INVALIDDATA;
        }

        bytes_used = avpkt->size;
    }

    if (!s->data) {
        *got_frame_ptr = 0;
        return avpkt->size;
    }

    blockstodecode = FFMIN(s->blocks_per_loop, s->samples);

    /* reallocate decoded sample buffer if needed */
    av_fast_malloc(&s->decoded_buffer, &s->decoded_size,
                   2 * FFALIGN(blockstodecode, 8) * sizeof(*s->decoded_buffer));
    if (!s->decoded_buffer)
        return AVERROR(ENOMEM);
    memset(s->decoded_buffer, 0, s->decoded_size);
    s->decoded[0] = s->decoded_buffer;
    s->decoded[1] = s->decoded_buffer + FFALIGN(blockstodecode, 8);

    /* get output buffer */
    frame->nb_samples = blockstodecode;
    if ((ret = ff_get_buffer(avctx, frame)) < 0) {
        av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
        return ret;
    }

    s->error=0;

    if ((s->channels == 1) || (s->frameflags & APE_FRAMECODE_PSEUDO_STEREO))
        ape_unpack_mono(s, blockstodecode);
    else
        ape_unpack_stereo(s, blockstodecode);
    emms_c();

    if (s->error) {
        s->samples=0;
        av_log(avctx, AV_LOG_ERROR, "Error decoding frame\n");
        return AVERROR_INVALIDDATA;
    }

    switch (s->bps) {
    case 8:
        for (ch = 0; ch < s->channels; ch++) {
            sample8 = (uint8_t *)frame->data[ch];
            for (i = 0; i < blockstodecode; i++)
                *sample8++ = (s->decoded[ch][i] + 0x80) & 0xff;
        }
        break;
    case 16:
        for (ch = 0; ch < s->channels; ch++) {
            sample16 = (int16_t *)frame->data[ch];
            for (i = 0; i < blockstodecode; i++)
                *sample16++ = s->decoded[ch][i];
        }
        break;
    case 24:
        for (ch = 0; ch < s->channels; ch++) {
            sample24 = (int32_t *)frame->data[ch];
            for (i = 0; i < blockstodecode; i++)
                *sample24++ = s->decoded[ch][i] << 8;
        }
        break;
    }

    s->samples -= blockstodecode;

    *got_frame_ptr = 1;

    return bytes_used;
}

static void ape_flush(AVCodecContext *avctx)
{
    APEContext *s = avctx->priv_data;
    s->samples= 0;
}

#define OFFSET(x) offsetof(APEContext, x)
#define PAR (AV_OPT_FLAG_DECODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM)
static const AVOption options[] = {
    { "max_samples", "maximum number of samples decoded per call",             OFFSET(blocks_per_loop), AV_OPT_TYPE_INT,   { .i64 = 4608 },    1,       INT_MAX, PAR, "max_samples" },
    { "all",         "no maximum. decode all samples for each packet at once", 0,                       AV_OPT_TYPE_CONST, { .i64 = INT_MAX }, INT_MIN, INT_MAX, PAR, "max_samples" },
    { NULL},
};

static const AVClass ape_decoder_class = {
    .class_name = "APE decoder",
    .item_name  = av_default_item_name,
    .option     = options,
    .version    = LIBAVUTIL_VERSION_INT,
};

AVCodec ff_ape_decoder = {
    .name           = "ape",
    .type           = AVMEDIA_TYPE_AUDIO,
    .id             = AV_CODEC_ID_APE,
    .priv_data_size = sizeof(APEContext),
    .init           = ape_decode_init,
    .close          = ape_decode_close,
    .decode         = ape_decode_frame,
    .capabilities   = CODEC_CAP_SUBFRAMES | CODEC_CAP_DELAY | CODEC_CAP_DR1,
    .flush          = ape_flush,
    .long_name      = NULL_IF_CONFIG_SMALL("Monkey's Audio"),
    .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_U8P,
                                                      AV_SAMPLE_FMT_S16P,
                                                      AV_SAMPLE_FMT_S32P,
                                                      AV_SAMPLE_FMT_NONE },
    .priv_class     = &ape_decoder_class,
};