xref: /third_party/ffmpeg/libavcodec/adpcmenc.c (revision cabdff1a)
1/*
2 * Copyright (c) 2001-2003 The FFmpeg project
3 *
4 * first version by Francois Revol (revol@free.fr)
5 * fringe ADPCM codecs (e.g., DK3, DK4, Westwood)
6 *   by Mike Melanson (melanson@pcisys.net)
7 *
8 * This file is part of FFmpeg.
9 *
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
14 *
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18 * Lesser General Public License for more details.
19 *
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 */
24
25#include "config_components.h"
26
27#include "libavutil/opt.h"
28
29#include "avcodec.h"
30#include "put_bits.h"
31#include "bytestream.h"
32#include "adpcm.h"
33#include "adpcm_data.h"
34#include "codec_internal.h"
35#include "encode.h"
36
37/**
38 * @file
39 * ADPCM encoders
40 * See ADPCM decoder reference documents for codec information.
41 */
42
43#define CASE_0(codec_id, ...)
44#define CASE_1(codec_id, ...) \
45    case codec_id:            \
46    { __VA_ARGS__ }           \
47    break;
48#define CASE_2(enabled, codec_id, ...) \
49        CASE_ ## enabled(codec_id, __VA_ARGS__)
50#define CASE_3(config, codec_id, ...) \
51        CASE_2(config, codec_id, __VA_ARGS__)
52#define CASE(codec, ...) \
53        CASE_3(CONFIG_ ## codec ## _ENCODER, AV_CODEC_ID_ ## codec, __VA_ARGS__)
54
55typedef struct TrellisPath {
56    int nibble;
57    int prev;
58} TrellisPath;
59
60typedef struct TrellisNode {
61    uint32_t ssd;
62    int path;
63    int sample1;
64    int sample2;
65    int step;
66} TrellisNode;
67
68typedef struct ADPCMEncodeContext {
69    AVClass *class;
70    int block_size;
71
72    ADPCMChannelStatus status[6];
73    TrellisPath *paths;
74    TrellisNode *node_buf;
75    TrellisNode **nodep_buf;
76    uint8_t *trellis_hash;
77} ADPCMEncodeContext;
78
79#define FREEZE_INTERVAL 128
80
81static av_cold int adpcm_encode_init(AVCodecContext *avctx)
82{
83    ADPCMEncodeContext *s = avctx->priv_data;
84    int channels = avctx->ch_layout.nb_channels;
85
86    /*
87     * AMV's block size has to match that of the corresponding video
88     * stream. Relax the POT requirement.
89     */
90    if (avctx->codec->id != AV_CODEC_ID_ADPCM_IMA_AMV &&
91        (s->block_size & (s->block_size - 1))) {
92        av_log(avctx, AV_LOG_ERROR, "block size must be power of 2\n");
93        return AVERROR(EINVAL);
94    }
95
96    if (avctx->trellis) {
97        int frontier, max_paths;
98
99        if ((unsigned)avctx->trellis > 16U) {
100            av_log(avctx, AV_LOG_ERROR, "invalid trellis size\n");
101            return AVERROR(EINVAL);
102        }
103
104        if (avctx->codec->id == AV_CODEC_ID_ADPCM_IMA_SSI ||
105            avctx->codec->id == AV_CODEC_ID_ADPCM_IMA_APM ||
106            avctx->codec->id == AV_CODEC_ID_ADPCM_ARGO    ||
107            avctx->codec->id == AV_CODEC_ID_ADPCM_IMA_WS) {
108            /*
109             * The current trellis implementation doesn't work for extended
110             * runs of samples without periodic resets. Disallow it.
111             */
112            av_log(avctx, AV_LOG_ERROR, "trellis not supported\n");
113            return AVERROR_PATCHWELCOME;
114        }
115
116        frontier  = 1 << avctx->trellis;
117        max_paths =  frontier * FREEZE_INTERVAL;
118        if (!FF_ALLOC_TYPED_ARRAY(s->paths,        max_paths)    ||
119            !FF_ALLOC_TYPED_ARRAY(s->node_buf,     2 * frontier) ||
120            !FF_ALLOC_TYPED_ARRAY(s->nodep_buf,    2 * frontier) ||
121            !FF_ALLOC_TYPED_ARRAY(s->trellis_hash, 65536))
122            return AVERROR(ENOMEM);
123    }
124
125    avctx->bits_per_coded_sample = av_get_bits_per_sample(avctx->codec->id);
126
127    switch (avctx->codec->id) {
128    CASE(ADPCM_IMA_WAV,
129        /* each 16 bits sample gives one nibble
130           and we have 4 bytes per channel overhead */
131        avctx->frame_size = (s->block_size - 4 * channels) * 8 /
132                            (4 * channels) + 1;
133        /* seems frame_size isn't taken into account...
134           have to buffer the samples :-( */
135        avctx->block_align = s->block_size;
136        avctx->bits_per_coded_sample = 4;
137        ) /* End of CASE */
138    CASE(ADPCM_IMA_QT,
139        avctx->frame_size  = 64;
140        avctx->block_align = 34 * channels;
141        ) /* End of CASE */
142    CASE(ADPCM_MS,
143        uint8_t *extradata;
144        /* each 16 bits sample gives one nibble
145           and we have 7 bytes per channel overhead */
146        avctx->frame_size = (s->block_size - 7 * channels) * 2 / channels + 2;
147        avctx->bits_per_coded_sample = 4;
148        avctx->block_align     = s->block_size;
149        if (!(avctx->extradata = av_malloc(32 + AV_INPUT_BUFFER_PADDING_SIZE)))
150            return AVERROR(ENOMEM);
151        avctx->extradata_size = 32;
152        extradata = avctx->extradata;
153        bytestream_put_le16(&extradata, avctx->frame_size);
154        bytestream_put_le16(&extradata, 7); /* wNumCoef */
155        for (int i = 0; i < 7; i++) {
156            bytestream_put_le16(&extradata, ff_adpcm_AdaptCoeff1[i] * 4);
157            bytestream_put_le16(&extradata, ff_adpcm_AdaptCoeff2[i] * 4);
158        }
159        ) /* End of CASE */
160    CASE(ADPCM_YAMAHA,
161        avctx->frame_size  = s->block_size * 2 / channels;
162        avctx->block_align = s->block_size;
163        ) /* End of CASE */
164    CASE(ADPCM_SWF,
165        if (avctx->sample_rate != 11025 &&
166            avctx->sample_rate != 22050 &&
167            avctx->sample_rate != 44100) {
168            av_log(avctx, AV_LOG_ERROR, "Sample rate must be 11025, "
169                   "22050 or 44100\n");
170            return AVERROR(EINVAL);
171        }
172        avctx->frame_size  = 4096; /* Hardcoded according to the SWF spec. */
173        avctx->block_align = (2 + channels * (22 + 4 * (avctx->frame_size - 1)) + 7) / 8;
174        ) /* End of CASE */
175    case AV_CODEC_ID_ADPCM_IMA_SSI:
176    case AV_CODEC_ID_ADPCM_IMA_ALP:
177        avctx->frame_size  = s->block_size * 2 / channels;
178        avctx->block_align = s->block_size;
179        break;
180    CASE(ADPCM_IMA_AMV,
181        if (avctx->sample_rate != 22050) {
182            av_log(avctx, AV_LOG_ERROR, "Sample rate must be 22050\n");
183            return AVERROR(EINVAL);
184        }
185
186        if (channels != 1) {
187            av_log(avctx, AV_LOG_ERROR, "Only mono is supported\n");
188            return AVERROR(EINVAL);
189        }
190
191        avctx->frame_size  = s->block_size;
192        avctx->block_align = 8 + (FFALIGN(avctx->frame_size, 2) / 2);
193        ) /* End of CASE */
194    CASE(ADPCM_IMA_APM,
195        avctx->frame_size  = s->block_size * 2 / channels;
196        avctx->block_align = s->block_size;
197
198        if (!(avctx->extradata = av_mallocz(28 + AV_INPUT_BUFFER_PADDING_SIZE)))
199            return AVERROR(ENOMEM);
200        avctx->extradata_size = 28;
201        ) /* End of CASE */
202    CASE(ADPCM_ARGO,
203        avctx->frame_size = 32;
204        avctx->block_align = 17 * channels;
205        ) /* End of CASE */
206    CASE(ADPCM_IMA_WS,
207        /* each 16 bits sample gives one nibble */
208        avctx->frame_size = s->block_size * 2 / channels;
209        avctx->block_align = s->block_size;
210        ) /* End of CASE */
211    default:
212        return AVERROR(EINVAL);
213    }
214
215    return 0;
216}
217
218static av_cold int adpcm_encode_close(AVCodecContext *avctx)
219{
220    ADPCMEncodeContext *s = avctx->priv_data;
221    av_freep(&s->paths);
222    av_freep(&s->node_buf);
223    av_freep(&s->nodep_buf);
224    av_freep(&s->trellis_hash);
225
226    return 0;
227}
228
229
230static inline uint8_t adpcm_ima_compress_sample(ADPCMChannelStatus *c,
231                                                int16_t sample)
232{
233    int delta  = sample - c->prev_sample;
234    int nibble = FFMIN(7, abs(delta) * 4 /
235                       ff_adpcm_step_table[c->step_index]) + (delta < 0) * 8;
236    c->prev_sample += ((ff_adpcm_step_table[c->step_index] *
237                        ff_adpcm_yamaha_difflookup[nibble]) / 8);
238    c->prev_sample = av_clip_int16(c->prev_sample);
239    c->step_index  = av_clip(c->step_index + ff_adpcm_index_table[nibble], 0, 88);
240    return nibble;
241}
242
243static inline uint8_t adpcm_ima_alp_compress_sample(ADPCMChannelStatus *c, int16_t sample)
244{
245    const int delta  = sample - c->prev_sample;
246    const int step   = ff_adpcm_step_table[c->step_index];
247    const int sign   = (delta < 0) * 8;
248
249    int nibble = FFMIN(abs(delta) * 4 / step, 7);
250    int diff   = (step * nibble) >> 2;
251    if (sign)
252        diff = -diff;
253
254    nibble = sign | nibble;
255
256    c->prev_sample += diff;
257    c->prev_sample  = av_clip_int16(c->prev_sample);
258    c->step_index   = av_clip(c->step_index + ff_adpcm_index_table[nibble], 0, 88);
259    return nibble;
260}
261
262static inline uint8_t adpcm_ima_qt_compress_sample(ADPCMChannelStatus *c,
263                                                   int16_t sample)
264{
265    int delta  = sample - c->prev_sample;
266    int diff, step = ff_adpcm_step_table[c->step_index];
267    int nibble = 8*(delta < 0);
268
269    delta= abs(delta);
270    diff = delta + (step >> 3);
271
272    if (delta >= step) {
273        nibble |= 4;
274        delta  -= step;
275    }
276    step >>= 1;
277    if (delta >= step) {
278        nibble |= 2;
279        delta  -= step;
280    }
281    step >>= 1;
282    if (delta >= step) {
283        nibble |= 1;
284        delta  -= step;
285    }
286    diff -= delta;
287
288    if (nibble & 8)
289        c->prev_sample -= diff;
290    else
291        c->prev_sample += diff;
292
293    c->prev_sample = av_clip_int16(c->prev_sample);
294    c->step_index  = av_clip(c->step_index + ff_adpcm_index_table[nibble], 0, 88);
295
296    return nibble;
297}
298
299static inline uint8_t adpcm_ms_compress_sample(ADPCMChannelStatus *c,
300                                               int16_t sample)
301{
302    int predictor, nibble, bias;
303
304    predictor = (((c->sample1) * (c->coeff1)) +
305                (( c->sample2) * (c->coeff2))) / 64;
306
307    nibble = sample - predictor;
308    if (nibble >= 0)
309        bias =  c->idelta / 2;
310    else
311        bias = -c->idelta / 2;
312
313    nibble = (nibble + bias) / c->idelta;
314    nibble = av_clip_intp2(nibble, 3) & 0x0F;
315
316    predictor += ((nibble & 0x08) ? (nibble - 0x10) : nibble) * c->idelta;
317
318    c->sample2 = c->sample1;
319    c->sample1 = av_clip_int16(predictor);
320
321    c->idelta = (ff_adpcm_AdaptationTable[nibble] * c->idelta) >> 8;
322    if (c->idelta < 16)
323        c->idelta = 16;
324
325    return nibble;
326}
327
328static inline uint8_t adpcm_yamaha_compress_sample(ADPCMChannelStatus *c,
329                                                   int16_t sample)
330{
331    int nibble, delta;
332
333    if (!c->step) {
334        c->predictor = 0;
335        c->step      = 127;
336    }
337
338    delta = sample - c->predictor;
339
340    nibble = FFMIN(7, abs(delta) * 4 / c->step) + (delta < 0) * 8;
341
342    c->predictor += ((c->step * ff_adpcm_yamaha_difflookup[nibble]) / 8);
343    c->predictor = av_clip_int16(c->predictor);
344    c->step = (c->step * ff_adpcm_yamaha_indexscale[nibble]) >> 8;
345    c->step = av_clip(c->step, 127, 24576);
346
347    return nibble;
348}
349
350static void adpcm_compress_trellis(AVCodecContext *avctx,
351                                   const int16_t *samples, uint8_t *dst,
352                                   ADPCMChannelStatus *c, int n, int stride)
353{
354    //FIXME 6% faster if frontier is a compile-time constant
355    ADPCMEncodeContext *s = avctx->priv_data;
356    const int frontier = 1 << avctx->trellis;
357    const int version  = avctx->codec->id;
358    TrellisPath *paths       = s->paths, *p;
359    TrellisNode *node_buf    = s->node_buf;
360    TrellisNode **nodep_buf  = s->nodep_buf;
361    TrellisNode **nodes      = nodep_buf; // nodes[] is always sorted by .ssd
362    TrellisNode **nodes_next = nodep_buf + frontier;
363    int pathn = 0, froze = -1, i, j, k, generation = 0;
364    uint8_t *hash = s->trellis_hash;
365    memset(hash, 0xff, 65536 * sizeof(*hash));
366
367    memset(nodep_buf, 0, 2 * frontier * sizeof(*nodep_buf));
368    nodes[0]          = node_buf + frontier;
369    nodes[0]->ssd     = 0;
370    nodes[0]->path    = 0;
371    nodes[0]->step    = c->step_index;
372    nodes[0]->sample1 = c->sample1;
373    nodes[0]->sample2 = c->sample2;
374    if (version == AV_CODEC_ID_ADPCM_IMA_WAV ||
375        version == AV_CODEC_ID_ADPCM_IMA_QT  ||
376        version == AV_CODEC_ID_ADPCM_IMA_AMV ||
377        version == AV_CODEC_ID_ADPCM_SWF)
378        nodes[0]->sample1 = c->prev_sample;
379    if (version == AV_CODEC_ID_ADPCM_MS)
380        nodes[0]->step = c->idelta;
381    if (version == AV_CODEC_ID_ADPCM_YAMAHA) {
382        if (c->step == 0) {
383            nodes[0]->step    = 127;
384            nodes[0]->sample1 = 0;
385        } else {
386            nodes[0]->step    = c->step;
387            nodes[0]->sample1 = c->predictor;
388        }
389    }
390
391    for (i = 0; i < n; i++) {
392        TrellisNode *t = node_buf + frontier*(i&1);
393        TrellisNode **u;
394        int sample   = samples[i * stride];
395        int heap_pos = 0;
396        memset(nodes_next, 0, frontier * sizeof(TrellisNode*));
397        for (j = 0; j < frontier && nodes[j]; j++) {
398            // higher j have higher ssd already, so they're likely
399            // to yield a suboptimal next sample too
400            const int range = (j < frontier / 2) ? 1 : 0;
401            const int step  = nodes[j]->step;
402            int nidx;
403            if (version == AV_CODEC_ID_ADPCM_MS) {
404                const int predictor = ((nodes[j]->sample1 * c->coeff1) +
405                                       (nodes[j]->sample2 * c->coeff2)) / 64;
406                const int div  = (sample - predictor) / step;
407                const int nmin = av_clip(div-range, -8, 6);
408                const int nmax = av_clip(div+range, -7, 7);
409                for (nidx = nmin; nidx <= nmax; nidx++) {
410                    const int nibble = nidx & 0xf;
411                    int dec_sample   = predictor + nidx * step;
412#define STORE_NODE(NAME, STEP_INDEX)\
413                    int d;\
414                    uint32_t ssd;\
415                    int pos;\
416                    TrellisNode *u;\
417                    uint8_t *h;\
418                    dec_sample = av_clip_int16(dec_sample);\
419                    d = sample - dec_sample;\
420                    ssd = nodes[j]->ssd + d*(unsigned)d;\
421                    /* Check for wraparound, skip such samples completely. \
422                     * Note, changing ssd to a 64 bit variable would be \
423                     * simpler, avoiding this check, but it's slower on \
424                     * x86 32 bit at the moment. */\
425                    if (ssd < nodes[j]->ssd)\
426                        goto next_##NAME;\
427                    /* Collapse any two states with the same previous sample value. \
428                     * One could also distinguish states by step and by 2nd to last
429                     * sample, but the effects of that are negligible.
430                     * Since nodes in the previous generation are iterated
431                     * through a heap, they're roughly ordered from better to
432                     * worse, but not strictly ordered. Therefore, an earlier
433                     * node with the same sample value is better in most cases
434                     * (and thus the current is skipped), but not strictly
435                     * in all cases. Only skipping samples where ssd >=
436                     * ssd of the earlier node with the same sample gives
437                     * slightly worse quality, though, for some reason. */ \
438                    h = &hash[(uint16_t) dec_sample];\
439                    if (*h == generation)\
440                        goto next_##NAME;\
441                    if (heap_pos < frontier) {\
442                        pos = heap_pos++;\
443                    } else {\
444                        /* Try to replace one of the leaf nodes with the new \
445                         * one, but try a different slot each time. */\
446                        pos = (frontier >> 1) +\
447                              (heap_pos & ((frontier >> 1) - 1));\
448                        if (ssd > nodes_next[pos]->ssd)\
449                            goto next_##NAME;\
450                        heap_pos++;\
451                    }\
452                    *h = generation;\
453                    u  = nodes_next[pos];\
454                    if (!u) {\
455                        av_assert1(pathn < FREEZE_INTERVAL << avctx->trellis);\
456                        u = t++;\
457                        nodes_next[pos] = u;\
458                        u->path = pathn++;\
459                    }\
460                    u->ssd  = ssd;\
461                    u->step = STEP_INDEX;\
462                    u->sample2 = nodes[j]->sample1;\
463                    u->sample1 = dec_sample;\
464                    paths[u->path].nibble = nibble;\
465                    paths[u->path].prev   = nodes[j]->path;\
466                    /* Sift the newly inserted node up in the heap to \
467                     * restore the heap property. */\
468                    while (pos > 0) {\
469                        int parent = (pos - 1) >> 1;\
470                        if (nodes_next[parent]->ssd <= ssd)\
471                            break;\
472                        FFSWAP(TrellisNode*, nodes_next[parent], nodes_next[pos]);\
473                        pos = parent;\
474                    }\
475                    next_##NAME:;
476                    STORE_NODE(ms, FFMAX(16,
477                               (ff_adpcm_AdaptationTable[nibble] * step) >> 8));
478                }
479            } else if (version == AV_CODEC_ID_ADPCM_IMA_WAV ||
480                       version == AV_CODEC_ID_ADPCM_IMA_QT  ||
481                       version == AV_CODEC_ID_ADPCM_IMA_AMV ||
482                       version == AV_CODEC_ID_ADPCM_SWF) {
483#define LOOP_NODES(NAME, STEP_TABLE, STEP_INDEX)\
484                const int predictor = nodes[j]->sample1;\
485                const int div = (sample - predictor) * 4 / STEP_TABLE;\
486                int nmin = av_clip(div - range, -7, 6);\
487                int nmax = av_clip(div + range, -6, 7);\
488                if (nmin <= 0)\
489                    nmin--; /* distinguish -0 from +0 */\
490                if (nmax < 0)\
491                    nmax--;\
492                for (nidx = nmin; nidx <= nmax; nidx++) {\
493                    const int nibble = nidx < 0 ? 7 - nidx : nidx;\
494                    int dec_sample = predictor +\
495                                    (STEP_TABLE *\
496                                     ff_adpcm_yamaha_difflookup[nibble]) / 8;\
497                    STORE_NODE(NAME, STEP_INDEX);\
498                }
499                LOOP_NODES(ima, ff_adpcm_step_table[step],
500                           av_clip(step + ff_adpcm_index_table[nibble], 0, 88));
501            } else { //AV_CODEC_ID_ADPCM_YAMAHA
502                LOOP_NODES(yamaha, step,
503                           av_clip((step * ff_adpcm_yamaha_indexscale[nibble]) >> 8,
504                                   127, 24576));
505#undef LOOP_NODES
506#undef STORE_NODE
507            }
508        }
509
510        u = nodes;
511        nodes = nodes_next;
512        nodes_next = u;
513
514        generation++;
515        if (generation == 255) {
516            memset(hash, 0xff, 65536 * sizeof(*hash));
517            generation = 0;
518        }
519
520        // prevent overflow
521        if (nodes[0]->ssd > (1 << 28)) {
522            for (j = 1; j < frontier && nodes[j]; j++)
523                nodes[j]->ssd -= nodes[0]->ssd;
524            nodes[0]->ssd = 0;
525        }
526
527        // merge old paths to save memory
528        if (i == froze + FREEZE_INTERVAL) {
529            p = &paths[nodes[0]->path];
530            for (k = i; k > froze; k--) {
531                dst[k] = p->nibble;
532                p = &paths[p->prev];
533            }
534            froze = i;
535            pathn = 0;
536            // other nodes might use paths that don't coincide with the frozen one.
537            // checking which nodes do so is too slow, so just kill them all.
538            // this also slightly improves quality, but I don't know why.
539            memset(nodes + 1, 0, (frontier - 1) * sizeof(TrellisNode*));
540        }
541    }
542
543    p = &paths[nodes[0]->path];
544    for (i = n - 1; i > froze; i--) {
545        dst[i] = p->nibble;
546        p = &paths[p->prev];
547    }
548
549    c->predictor  = nodes[0]->sample1;
550    c->sample1    = nodes[0]->sample1;
551    c->sample2    = nodes[0]->sample2;
552    c->step_index = nodes[0]->step;
553    c->step       = nodes[0]->step;
554    c->idelta     = nodes[0]->step;
555}
556
557#if CONFIG_ADPCM_ARGO_ENCODER
558static inline int adpcm_argo_compress_nibble(const ADPCMChannelStatus *cs, int16_t s,
559                                             int shift, int flag)
560{
561    int nibble;
562
563    if (flag)
564        nibble = 4 * s - 8 * cs->sample1 + 4 * cs->sample2;
565    else
566        nibble = 4 * s - 4 * cs->sample1;
567
568    return (nibble >> shift) & 0x0F;
569}
570
571static int64_t adpcm_argo_compress_block(ADPCMChannelStatus *cs, PutBitContext *pb,
572                                         const int16_t *samples, int nsamples,
573                                         int shift, int flag)
574{
575    int64_t error = 0;
576
577    if (pb) {
578        put_bits(pb, 4, shift - 2);
579        put_bits(pb, 1, 0);
580        put_bits(pb, 1, !!flag);
581        put_bits(pb, 2, 0);
582    }
583
584    for (int n = 0; n < nsamples; n++) {
585        /* Compress the nibble, then expand it to see how much precision we've lost. */
586        int nibble = adpcm_argo_compress_nibble(cs, samples[n], shift, flag);
587        int16_t sample = ff_adpcm_argo_expand_nibble(cs, nibble, shift, flag);
588
589        error += abs(samples[n] - sample);
590
591        if (pb)
592            put_bits(pb, 4, nibble);
593    }
594
595    return error;
596}
597#endif
598
599static int adpcm_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
600                              const AVFrame *frame, int *got_packet_ptr)
601{
602    int st, pkt_size, ret;
603    const int16_t *samples;
604    int16_t **samples_p;
605    uint8_t *dst;
606    ADPCMEncodeContext *c = avctx->priv_data;
607    int channels = avctx->ch_layout.nb_channels;
608
609    samples = (const int16_t *)frame->data[0];
610    samples_p = (int16_t **)frame->extended_data;
611    st = channels == 2;
612
613    if (avctx->codec_id == AV_CODEC_ID_ADPCM_IMA_SSI ||
614        avctx->codec_id == AV_CODEC_ID_ADPCM_IMA_ALP ||
615        avctx->codec_id == AV_CODEC_ID_ADPCM_IMA_APM ||
616        avctx->codec_id == AV_CODEC_ID_ADPCM_IMA_WS)
617        pkt_size = (frame->nb_samples * channels) / 2;
618    else
619        pkt_size = avctx->block_align;
620    if ((ret = ff_get_encode_buffer(avctx, avpkt, pkt_size, 0)) < 0)
621        return ret;
622    dst = avpkt->data;
623
624    switch(avctx->codec->id) {
625    CASE(ADPCM_IMA_WAV,
626        int blocks = (frame->nb_samples - 1) / 8;
627
628        for (int ch = 0; ch < channels; ch++) {
629            ADPCMChannelStatus *status = &c->status[ch];
630            status->prev_sample = samples_p[ch][0];
631            /* status->step_index = 0;
632               XXX: not sure how to init the state machine */
633            bytestream_put_le16(&dst, status->prev_sample);
634            *dst++ = status->step_index;
635            *dst++ = 0; /* unknown */
636        }
637
638        /* stereo: 4 bytes (8 samples) for left, 4 bytes for right */
639        if (avctx->trellis > 0) {
640            uint8_t *buf;
641            if (!FF_ALLOC_TYPED_ARRAY(buf, channels * blocks * 8))
642                return AVERROR(ENOMEM);
643            for (int ch = 0; ch < channels; ch++) {
644                adpcm_compress_trellis(avctx, &samples_p[ch][1],
645                                       buf + ch * blocks * 8, &c->status[ch],
646                                       blocks * 8, 1);
647            }
648            for (int i = 0; i < blocks; i++) {
649                for (int ch = 0; ch < channels; ch++) {
650                    uint8_t *buf1 = buf + ch * blocks * 8 + i * 8;
651                    for (int j = 0; j < 8; j += 2)
652                        *dst++ = buf1[j] | (buf1[j + 1] << 4);
653                }
654            }
655            av_free(buf);
656        } else {
657            for (int i = 0; i < blocks; i++) {
658                for (int ch = 0; ch < channels; ch++) {
659                    ADPCMChannelStatus *status = &c->status[ch];
660                    const int16_t *smp = &samples_p[ch][1 + i * 8];
661                    for (int j = 0; j < 8; j += 2) {
662                        uint8_t v = adpcm_ima_compress_sample(status, smp[j    ]);
663                        v        |= adpcm_ima_compress_sample(status, smp[j + 1]) << 4;
664                        *dst++ = v;
665                    }
666                }
667            }
668        }
669        ) /* End of CASE */
670    CASE(ADPCM_IMA_QT,
671        PutBitContext pb;
672        init_put_bits(&pb, dst, pkt_size);
673
674        for (int ch = 0; ch < channels; ch++) {
675            ADPCMChannelStatus *status = &c->status[ch];
676            put_bits(&pb, 9, (status->prev_sample & 0xFFFF) >> 7);
677            put_bits(&pb, 7,  status->step_index);
678            if (avctx->trellis > 0) {
679                uint8_t buf[64];
680                adpcm_compress_trellis(avctx, &samples_p[ch][0], buf, status,
681                                       64, 1);
682                for (int i = 0; i < 64; i++)
683                    put_bits(&pb, 4, buf[i ^ 1]);
684                status->prev_sample = status->predictor;
685            } else {
686                for (int i = 0; i < 64; i += 2) {
687                    int t1, t2;
688                    t1 = adpcm_ima_qt_compress_sample(status, samples_p[ch][i    ]);
689                    t2 = adpcm_ima_qt_compress_sample(status, samples_p[ch][i + 1]);
690                    put_bits(&pb, 4, t2);
691                    put_bits(&pb, 4, t1);
692                }
693            }
694        }
695
696        flush_put_bits(&pb);
697        ) /* End of CASE */
698    CASE(ADPCM_IMA_SSI,
699        PutBitContext pb;
700        init_put_bits(&pb, dst, pkt_size);
701
702        av_assert0(avctx->trellis == 0);
703
704        for (int i = 0; i < frame->nb_samples; i++) {
705            for (int ch = 0; ch < channels; ch++) {
706                put_bits(&pb, 4, adpcm_ima_qt_compress_sample(c->status + ch, *samples++));
707            }
708        }
709
710        flush_put_bits(&pb);
711        ) /* End of CASE */
712    CASE(ADPCM_IMA_ALP,
713        PutBitContext pb;
714        init_put_bits(&pb, dst, pkt_size);
715
716        av_assert0(avctx->trellis == 0);
717
718        for (int n = frame->nb_samples / 2; n > 0; n--) {
719            for (int ch = 0; ch < channels; ch++) {
720                put_bits(&pb, 4, adpcm_ima_alp_compress_sample(c->status + ch, *samples++));
721                put_bits(&pb, 4, adpcm_ima_alp_compress_sample(c->status + ch, samples[st]));
722            }
723            samples += channels;
724        }
725
726        flush_put_bits(&pb);
727        ) /* End of CASE */
728    CASE(ADPCM_SWF,
729        const int n = frame->nb_samples - 1;
730        PutBitContext pb;
731        init_put_bits(&pb, dst, pkt_size);
732
733        /* NB: This is safe as we don't have AV_CODEC_CAP_SMALL_LAST_FRAME. */
734        av_assert0(n == 4095);
735
736        // store AdpcmCodeSize
737        put_bits(&pb, 2, 2);    // set 4-bit flash adpcm format
738
739        // init the encoder state
740        for (int i = 0; i < channels; i++) {
741            // clip step so it fits 6 bits
742            c->status[i].step_index = av_clip_uintp2(c->status[i].step_index, 6);
743            put_sbits(&pb, 16, samples[i]);
744            put_bits(&pb, 6, c->status[i].step_index);
745            c->status[i].prev_sample = samples[i];
746        }
747
748        if (avctx->trellis > 0) {
749            uint8_t buf[8190 /* = 2 * n */];
750            adpcm_compress_trellis(avctx, samples + channels, buf,
751                                   &c->status[0], n, channels);
752            if (channels == 2)
753                adpcm_compress_trellis(avctx, samples + channels + 1,
754                                       buf + n, &c->status[1], n,
755                                       channels);
756            for (int i = 0; i < n; i++) {
757                put_bits(&pb, 4, buf[i]);
758                if (channels == 2)
759                    put_bits(&pb, 4, buf[n + i]);
760            }
761        } else {
762            for (int i = 1; i < frame->nb_samples; i++) {
763                put_bits(&pb, 4, adpcm_ima_compress_sample(&c->status[0],
764                         samples[channels * i]));
765                if (channels == 2)
766                    put_bits(&pb, 4, adpcm_ima_compress_sample(&c->status[1],
767                             samples[2 * i + 1]));
768            }
769        }
770        flush_put_bits(&pb);
771        ) /* End of CASE */
772    CASE(ADPCM_MS,
773        for (int i = 0; i < channels; i++) {
774            int predictor = 0;
775            *dst++ = predictor;
776            c->status[i].coeff1 = ff_adpcm_AdaptCoeff1[predictor];
777            c->status[i].coeff2 = ff_adpcm_AdaptCoeff2[predictor];
778        }
779        for (int i = 0; i < channels; i++) {
780            if (c->status[i].idelta < 16)
781                c->status[i].idelta = 16;
782            bytestream_put_le16(&dst, c->status[i].idelta);
783        }
784        for (int i = 0; i < channels; i++)
785            c->status[i].sample2= *samples++;
786        for (int i = 0; i < channels; i++) {
787            c->status[i].sample1 = *samples++;
788            bytestream_put_le16(&dst, c->status[i].sample1);
789        }
790        for (int i = 0; i < channels; i++)
791            bytestream_put_le16(&dst, c->status[i].sample2);
792
793        if (avctx->trellis > 0) {
794            const int n  = avctx->block_align - 7 * channels;
795            uint8_t *buf = av_malloc(2 * n);
796            if (!buf)
797                return AVERROR(ENOMEM);
798            if (channels == 1) {
799                adpcm_compress_trellis(avctx, samples, buf, &c->status[0], n,
800                                       channels);
801                for (int i = 0; i < n; i += 2)
802                    *dst++ = (buf[i] << 4) | buf[i + 1];
803            } else {
804                adpcm_compress_trellis(avctx, samples,     buf,
805                                       &c->status[0], n, channels);
806                adpcm_compress_trellis(avctx, samples + 1, buf + n,
807                                       &c->status[1], n, channels);
808                for (int i = 0; i < n; i++)
809                    *dst++ = (buf[i] << 4) | buf[n + i];
810            }
811            av_free(buf);
812        } else {
813            for (int i = 7 * channels; i < avctx->block_align; i++) {
814                int nibble;
815                nibble  = adpcm_ms_compress_sample(&c->status[ 0], *samples++) << 4;
816                nibble |= adpcm_ms_compress_sample(&c->status[st], *samples++);
817                *dst++  = nibble;
818            }
819        }
820        ) /* End of CASE */
821    CASE(ADPCM_YAMAHA,
822        int n = frame->nb_samples / 2;
823        if (avctx->trellis > 0) {
824            uint8_t *buf = av_malloc(2 * n * 2);
825            if (!buf)
826                return AVERROR(ENOMEM);
827            n *= 2;
828            if (channels == 1) {
829                adpcm_compress_trellis(avctx, samples, buf, &c->status[0], n,
830                                       channels);
831                for (int i = 0; i < n; i += 2)
832                    *dst++ = buf[i] | (buf[i + 1] << 4);
833            } else {
834                adpcm_compress_trellis(avctx, samples,     buf,
835                                       &c->status[0], n, channels);
836                adpcm_compress_trellis(avctx, samples + 1, buf + n,
837                                       &c->status[1], n, channels);
838                for (int i = 0; i < n; i++)
839                    *dst++ = buf[i] | (buf[n + i] << 4);
840            }
841            av_free(buf);
842        } else
843            for (n *= channels; n > 0; n--) {
844                int nibble;
845                nibble  = adpcm_yamaha_compress_sample(&c->status[ 0], *samples++);
846                nibble |= adpcm_yamaha_compress_sample(&c->status[st], *samples++) << 4;
847                *dst++  = nibble;
848            }
849        ) /* End of CASE */
850    CASE(ADPCM_IMA_APM,
851        PutBitContext pb;
852        init_put_bits(&pb, dst, pkt_size);
853
854        av_assert0(avctx->trellis == 0);
855
856        for (int n = frame->nb_samples / 2; n > 0; n--) {
857            for (int ch = 0; ch < channels; ch++) {
858                put_bits(&pb, 4, adpcm_ima_qt_compress_sample(c->status + ch, *samples++));
859                put_bits(&pb, 4, adpcm_ima_qt_compress_sample(c->status + ch, samples[st]));
860            }
861            samples += channels;
862        }
863
864        flush_put_bits(&pb);
865        ) /* End of CASE */
866    CASE(ADPCM_IMA_AMV,
867        av_assert0(channels == 1);
868
869        c->status[0].prev_sample = *samples;
870        bytestream_put_le16(&dst, c->status[0].prev_sample);
871        bytestream_put_byte(&dst, c->status[0].step_index);
872        bytestream_put_byte(&dst, 0);
873        bytestream_put_le32(&dst, avctx->frame_size);
874
875        if (avctx->trellis > 0) {
876            const int n  = frame->nb_samples >> 1;
877            uint8_t *buf = av_malloc(2 * n);
878
879            if (!buf)
880                return AVERROR(ENOMEM);
881
882            adpcm_compress_trellis(avctx, samples, buf, &c->status[0], 2 * n, channels);
883            for (int i = 0; i < n; i++)
884                bytestream_put_byte(&dst, (buf[2 * i] << 4) | buf[2 * i + 1]);
885
886            samples += 2 * n;
887            av_free(buf);
888        } else for (int n = frame->nb_samples >> 1; n > 0; n--) {
889            int nibble;
890            nibble  = adpcm_ima_compress_sample(&c->status[0], *samples++) << 4;
891            nibble |= adpcm_ima_compress_sample(&c->status[0], *samples++) & 0x0F;
892            bytestream_put_byte(&dst, nibble);
893        }
894
895        if (avctx->frame_size & 1) {
896            int nibble = adpcm_ima_compress_sample(&c->status[0], *samples++) << 4;
897            bytestream_put_byte(&dst, nibble);
898        }
899        ) /* End of CASE */
900    CASE(ADPCM_ARGO,
901        PutBitContext pb;
902        init_put_bits(&pb, dst, pkt_size);
903
904        av_assert0(frame->nb_samples == 32);
905
906        for (int ch = 0; ch < channels; ch++) {
907            int64_t error  = INT64_MAX, tmperr = INT64_MAX;
908            int     shift  = 2, flag = 0;
909            int     saved1 = c->status[ch].sample1;
910            int     saved2 = c->status[ch].sample2;
911
912            /* Find the optimal coefficients, bail early if we find a perfect result. */
913            for (int s = 2; s < 18 && tmperr != 0; s++) {
914                for (int f = 0; f < 2 && tmperr != 0; f++) {
915                    c->status[ch].sample1 = saved1;
916                    c->status[ch].sample2 = saved2;
917                    tmperr = adpcm_argo_compress_block(c->status + ch, NULL, samples_p[ch],
918                                                       frame->nb_samples, s, f);
919                    if (tmperr < error) {
920                        shift = s;
921                        flag  = f;
922                        error = tmperr;
923                    }
924                }
925            }
926
927            /* Now actually do the encode. */
928            c->status[ch].sample1 = saved1;
929            c->status[ch].sample2 = saved2;
930            adpcm_argo_compress_block(c->status + ch, &pb, samples_p[ch],
931                                      frame->nb_samples, shift, flag);
932        }
933
934        flush_put_bits(&pb);
935        ) /* End of CASE */
936    CASE(ADPCM_IMA_WS,
937        PutBitContext pb;
938        init_put_bits(&pb, dst, pkt_size);
939
940        av_assert0(avctx->trellis == 0);
941        for (int n = frame->nb_samples / 2; n > 0; n--) {
942            /* stereo: 1 byte (2 samples) for left, 1 byte for right */
943            for (int ch = 0; ch < channels; ch++) {
944                int t1, t2;
945                t1 = adpcm_ima_compress_sample(&c->status[ch], *samples++);
946                t2 = adpcm_ima_compress_sample(&c->status[ch], samples[st]);
947                put_bits(&pb, 4, t2);
948                put_bits(&pb, 4, t1);
949            }
950            samples += channels;
951        }
952        flush_put_bits(&pb);
953        ) /* End of CASE */
954    default:
955        return AVERROR(EINVAL);
956    }
957
958    *got_packet_ptr = 1;
959    return 0;
960}
961
962static const enum AVSampleFormat sample_fmts[] = {
963    AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
964};
965
966static const enum AVSampleFormat sample_fmts_p[] = {
967    AV_SAMPLE_FMT_S16P, AV_SAMPLE_FMT_NONE
968};
969
970static const AVChannelLayout ch_layouts[] = {
971    AV_CHANNEL_LAYOUT_MONO,
972    AV_CHANNEL_LAYOUT_STEREO,
973    { 0 },
974};
975
976static const AVOption options[] = {
977    {
978        .name        = "block_size",
979        .help        = "set the block size",
980        .offset      = offsetof(ADPCMEncodeContext, block_size),
981        .type        = AV_OPT_TYPE_INT,
982        .default_val = {.i64 = 1024},
983        .min         = 32,
984        .max         = 8192, /* Is this a reasonable upper limit? */
985        .flags       = AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_AUDIO_PARAM
986    },
987    { NULL }
988};
989
990static const AVClass adpcm_encoder_class = {
991    .class_name = "ADPCM encoder",
992    .item_name  = av_default_item_name,
993    .option     = options,
994    .version    = LIBAVUTIL_VERSION_INT,
995};
996
997#define ADPCM_ENCODER_0(id_, name_, sample_fmts_, capabilities_, long_name_)
998#define ADPCM_ENCODER_1(id_, name_, sample_fmts_, capabilities_, long_name_) \
999const FFCodec ff_ ## name_ ## _encoder = {                                 \
1000    .p.name         = #name_,                                              \
1001    .p.long_name    = NULL_IF_CONFIG_SMALL(long_name_),                    \
1002    .p.type         = AVMEDIA_TYPE_AUDIO,                                  \
1003    .p.id           = id_,                                                 \
1004    .p.sample_fmts  = sample_fmts_,                                        \
1005    .p.ch_layouts   = ch_layouts,                                          \
1006    .p.capabilities = capabilities_ | AV_CODEC_CAP_DR1,                    \
1007    .p.priv_class   = &adpcm_encoder_class,                                \
1008    .priv_data_size = sizeof(ADPCMEncodeContext),                          \
1009    .init           = adpcm_encode_init,                                   \
1010    FF_CODEC_ENCODE_CB(adpcm_encode_frame),                                \
1011    .close          = adpcm_encode_close,                                  \
1012    .caps_internal  = FF_CODEC_CAP_INIT_CLEANUP | FF_CODEC_CAP_INIT_THREADSAFE, \
1013};
1014#define ADPCM_ENCODER_2(enabled, codec_id, name, sample_fmts, capabilities, long_name) \
1015    ADPCM_ENCODER_ ## enabled(codec_id, name, sample_fmts, capabilities, long_name)
1016#define ADPCM_ENCODER_3(config, codec_id, name, sample_fmts, capabilities, long_name) \
1017    ADPCM_ENCODER_2(config, codec_id, name, sample_fmts, capabilities, long_name)
1018#define ADPCM_ENCODER(codec, name, sample_fmts, capabilities, long_name) \
1019    ADPCM_ENCODER_3(CONFIG_ ## codec ## _ENCODER, AV_CODEC_ID_ ## codec, \
1020                    name, sample_fmts, capabilities, long_name)
1021
1022ADPCM_ENCODER(ADPCM_ARGO,    adpcm_argo,    sample_fmts_p, 0,                             "ADPCM Argonaut Games")
1023ADPCM_ENCODER(ADPCM_IMA_AMV, adpcm_ima_amv, sample_fmts,   0,                             "ADPCM IMA AMV")
1024ADPCM_ENCODER(ADPCM_IMA_APM, adpcm_ima_apm, sample_fmts,   AV_CODEC_CAP_SMALL_LAST_FRAME, "ADPCM IMA Ubisoft APM")
1025ADPCM_ENCODER(ADPCM_IMA_ALP, adpcm_ima_alp, sample_fmts,   AV_CODEC_CAP_SMALL_LAST_FRAME, "ADPCM IMA High Voltage Software ALP")
1026ADPCM_ENCODER(ADPCM_IMA_QT,  adpcm_ima_qt,  sample_fmts_p, 0,                             "ADPCM IMA QuickTime")
1027ADPCM_ENCODER(ADPCM_IMA_SSI, adpcm_ima_ssi, sample_fmts,   AV_CODEC_CAP_SMALL_LAST_FRAME, "ADPCM IMA Simon & Schuster Interactive")
1028ADPCM_ENCODER(ADPCM_IMA_WAV, adpcm_ima_wav, sample_fmts_p, 0,                             "ADPCM IMA WAV")
1029ADPCM_ENCODER(ADPCM_IMA_WS,  adpcm_ima_ws,  sample_fmts,   AV_CODEC_CAP_SMALL_LAST_FRAME, "ADPCM IMA Westwood")
1030ADPCM_ENCODER(ADPCM_MS,      adpcm_ms,      sample_fmts,   0,                             "ADPCM Microsoft")
1031ADPCM_ENCODER(ADPCM_SWF,     adpcm_swf,     sample_fmts,   0,                             "ADPCM Shockwave Flash")
1032ADPCM_ENCODER(ADPCM_YAMAHA,  adpcm_yamaha,  sample_fmts,   0,                             "ADPCM Yamaha")
1033