flacdec: split off channel decorrelation as flacdsp
[libav.git] / libavcodec / flacdec.c
1 /*
2 * FLAC (Free Lossless Audio Codec) decoder
3 * Copyright (c) 2003 Alex Beregszaszi
4 *
5 * This file is part of Libav.
6 *
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * FLAC (Free Lossless Audio Codec) decoder
25 * @author Alex Beregszaszi
26 * @see http://flac.sourceforge.net/
27 *
28 * This decoder can be used in 1 of 2 ways: Either raw FLAC data can be fed
29 * through, starting from the initial 'fLaC' signature; or by passing the
30 * 34-byte streaminfo structure through avctx->extradata[_size] followed
31 * by data starting with the 0xFFF8 marker.
32 */
33
34 #include <limits.h>
35
36 #include "libavutil/audioconvert.h"
37 #include "libavutil/crc.h"
38 #include "avcodec.h"
39 #include "internal.h"
40 #include "get_bits.h"
41 #include "bytestream.h"
42 #include "golomb.h"
43 #include "flac.h"
44 #include "flacdata.h"
45 #include "flacdsp.h"
46
47 #undef NDEBUG
48 #include <assert.h>
49
50 typedef struct FLACContext {
51 FLACSTREAMINFO
52
53 AVCodecContext *avctx; ///< parent AVCodecContext
54 AVFrame frame;
55 GetBitContext gb; ///< GetBitContext initialized to start at the current frame
56
57 int blocksize; ///< number of samples in the current frame
58 int sample_shift; ///< shift required to make output samples 16-bit or 32-bit
59 int ch_mode; ///< channel decorrelation type in the current frame
60 int got_streaminfo; ///< indicates if the STREAMINFO has been read
61
62 int32_t *decoded[FLAC_MAX_CHANNELS]; ///< decoded samples
63
64 FLACDSPContext dsp;
65 } FLACContext;
66
67 static const int64_t flac_channel_layouts[6] = {
68 AV_CH_LAYOUT_MONO,
69 AV_CH_LAYOUT_STEREO,
70 AV_CH_LAYOUT_SURROUND,
71 AV_CH_LAYOUT_QUAD,
72 AV_CH_LAYOUT_5POINT0,
73 AV_CH_LAYOUT_5POINT1
74 };
75
76 static void allocate_buffers(FLACContext *s);
77
78 int avpriv_flac_is_extradata_valid(AVCodecContext *avctx,
79 enum FLACExtradataFormat *format,
80 uint8_t **streaminfo_start)
81 {
82 if (!avctx->extradata || avctx->extradata_size < FLAC_STREAMINFO_SIZE) {
83 av_log(avctx, AV_LOG_ERROR, "extradata NULL or too small.\n");
84 return 0;
85 }
86 if (AV_RL32(avctx->extradata) != MKTAG('f','L','a','C')) {
87 /* extradata contains STREAMINFO only */
88 if (avctx->extradata_size != FLAC_STREAMINFO_SIZE) {
89 av_log(avctx, AV_LOG_WARNING, "extradata contains %d bytes too many.\n",
90 FLAC_STREAMINFO_SIZE-avctx->extradata_size);
91 }
92 *format = FLAC_EXTRADATA_FORMAT_STREAMINFO;
93 *streaminfo_start = avctx->extradata;
94 } else {
95 if (avctx->extradata_size < 8+FLAC_STREAMINFO_SIZE) {
96 av_log(avctx, AV_LOG_ERROR, "extradata too small.\n");
97 return 0;
98 }
99 *format = FLAC_EXTRADATA_FORMAT_FULL_HEADER;
100 *streaminfo_start = &avctx->extradata[8];
101 }
102 return 1;
103 }
104
105 static void flac_set_bps(FLACContext *s)
106 {
107 if (s->bps > 16) {
108 s->avctx->sample_fmt = AV_SAMPLE_FMT_S32;
109 s->sample_shift = 32 - s->bps;
110 } else {
111 s->avctx->sample_fmt = AV_SAMPLE_FMT_S16;
112 s->sample_shift = 16 - s->bps;
113 }
114 }
115
116 static av_cold int flac_decode_init(AVCodecContext *avctx)
117 {
118 enum FLACExtradataFormat format;
119 uint8_t *streaminfo;
120 FLACContext *s = avctx->priv_data;
121 s->avctx = avctx;
122
123 /* for now, the raw FLAC header is allowed to be passed to the decoder as
124 frame data instead of extradata. */
125 if (!avctx->extradata)
126 return 0;
127
128 if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
129 return -1;
130
131 /* initialize based on the demuxer-supplied streamdata header */
132 avpriv_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s, streaminfo);
133 allocate_buffers(s);
134 flac_set_bps(s);
135 ff_flacdsp_init(&s->dsp, avctx->sample_fmt);
136 s->got_streaminfo = 1;
137
138 avcodec_get_frame_defaults(&s->frame);
139 avctx->coded_frame = &s->frame;
140
141 if (avctx->channels <= FF_ARRAY_ELEMS(flac_channel_layouts))
142 avctx->channel_layout = flac_channel_layouts[avctx->channels - 1];
143
144 return 0;
145 }
146
147 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
148 {
149 av_log(avctx, AV_LOG_DEBUG, " Max Blocksize: %d\n", s->max_blocksize);
150 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
151 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
152 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
153 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
154 }
155
156 static void allocate_buffers(FLACContext *s)
157 {
158 int i;
159
160 assert(s->max_blocksize);
161
162 for (i = 0; i < s->channels; i++) {
163 s->decoded[i] = av_malloc(sizeof(int32_t)*s->max_blocksize);
164 }
165 }
166
167 void avpriv_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
168 const uint8_t *buffer)
169 {
170 GetBitContext gb;
171 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
172
173 skip_bits(&gb, 16); /* skip min blocksize */
174 s->max_blocksize = get_bits(&gb, 16);
175 if (s->max_blocksize < FLAC_MIN_BLOCKSIZE) {
176 av_log(avctx, AV_LOG_WARNING, "invalid max blocksize: %d\n",
177 s->max_blocksize);
178 s->max_blocksize = 16;
179 }
180
181 skip_bits(&gb, 24); /* skip min frame size */
182 s->max_framesize = get_bits_long(&gb, 24);
183
184 s->samplerate = get_bits_long(&gb, 20);
185 s->channels = get_bits(&gb, 3) + 1;
186 s->bps = get_bits(&gb, 5) + 1;
187
188 avctx->channels = s->channels;
189 avctx->sample_rate = s->samplerate;
190 avctx->bits_per_raw_sample = s->bps;
191
192 s->samples = get_bits_long(&gb, 32) << 4;
193 s->samples |= get_bits(&gb, 4);
194
195 skip_bits_long(&gb, 64); /* md5 sum */
196 skip_bits_long(&gb, 64); /* md5 sum */
197
198 dump_headers(avctx, s);
199 }
200
201 void avpriv_flac_parse_block_header(const uint8_t *block_header,
202 int *last, int *type, int *size)
203 {
204 int tmp = bytestream_get_byte(&block_header);
205 if (last)
206 *last = tmp & 0x80;
207 if (type)
208 *type = tmp & 0x7F;
209 if (size)
210 *size = bytestream_get_be24(&block_header);
211 }
212
213 /**
214 * Parse the STREAMINFO from an inline header.
215 * @param s the flac decoding context
216 * @param buf input buffer, starting with the "fLaC" marker
217 * @param buf_size buffer size
218 * @return non-zero if metadata is invalid
219 */
220 static int parse_streaminfo(FLACContext *s, const uint8_t *buf, int buf_size)
221 {
222 int metadata_type, metadata_size;
223
224 if (buf_size < FLAC_STREAMINFO_SIZE+8) {
225 /* need more data */
226 return 0;
227 }
228 avpriv_flac_parse_block_header(&buf[4], NULL, &metadata_type, &metadata_size);
229 if (metadata_type != FLAC_METADATA_TYPE_STREAMINFO ||
230 metadata_size != FLAC_STREAMINFO_SIZE) {
231 return AVERROR_INVALIDDATA;
232 }
233 avpriv_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, &buf[8]);
234 allocate_buffers(s);
235 flac_set_bps(s);
236 ff_flacdsp_init(&s->dsp, s->avctx->sample_fmt);
237 s->got_streaminfo = 1;
238
239 return 0;
240 }
241
242 /**
243 * Determine the size of an inline header.
244 * @param buf input buffer, starting with the "fLaC" marker
245 * @param buf_size buffer size
246 * @return number of bytes in the header, or 0 if more data is needed
247 */
248 static int get_metadata_size(const uint8_t *buf, int buf_size)
249 {
250 int metadata_last, metadata_size;
251 const uint8_t *buf_end = buf + buf_size;
252
253 buf += 4;
254 do {
255 if (buf_end - buf < 4)
256 return 0;
257 avpriv_flac_parse_block_header(buf, &metadata_last, NULL, &metadata_size);
258 buf += 4;
259 if (buf_end - buf < metadata_size) {
260 /* need more data in order to read the complete header */
261 return 0;
262 }
263 buf += metadata_size;
264 } while (!metadata_last);
265
266 return buf_size - (buf_end - buf);
267 }
268
269 static int decode_residuals(FLACContext *s, int channel, int pred_order)
270 {
271 int i, tmp, partition, method_type, rice_order;
272 int sample = 0, samples;
273
274 method_type = get_bits(&s->gb, 2);
275 if (method_type > 1) {
276 av_log(s->avctx, AV_LOG_ERROR, "illegal residual coding method %d\n",
277 method_type);
278 return -1;
279 }
280
281 rice_order = get_bits(&s->gb, 4);
282
283 samples= s->blocksize >> rice_order;
284 if (pred_order > samples) {
285 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n",
286 pred_order, samples);
287 return -1;
288 }
289
290 sample=
291 i= pred_order;
292 for (partition = 0; partition < (1 << rice_order); partition++) {
293 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
294 if (tmp == (method_type == 0 ? 15 : 31)) {
295 tmp = get_bits(&s->gb, 5);
296 for (; i < samples; i++, sample++)
297 s->decoded[channel][sample] = get_sbits_long(&s->gb, tmp);
298 } else {
299 for (; i < samples; i++, sample++) {
300 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
301 }
302 }
303 i= 0;
304 }
305
306 return 0;
307 }
308
309 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order,
310 int bps)
311 {
312 const int blocksize = s->blocksize;
313 int32_t *decoded = s->decoded[channel];
314 int a, b, c, d, i;
315
316 /* warm up samples */
317 for (i = 0; i < pred_order; i++) {
318 decoded[i] = get_sbits_long(&s->gb, bps);
319 }
320
321 if (decode_residuals(s, channel, pred_order) < 0)
322 return -1;
323
324 if (pred_order > 0)
325 a = decoded[pred_order-1];
326 if (pred_order > 1)
327 b = a - decoded[pred_order-2];
328 if (pred_order > 2)
329 c = b - decoded[pred_order-2] + decoded[pred_order-3];
330 if (pred_order > 3)
331 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
332
333 switch (pred_order) {
334 case 0:
335 break;
336 case 1:
337 for (i = pred_order; i < blocksize; i++)
338 decoded[i] = a += decoded[i];
339 break;
340 case 2:
341 for (i = pred_order; i < blocksize; i++)
342 decoded[i] = a += b += decoded[i];
343 break;
344 case 3:
345 for (i = pred_order; i < blocksize; i++)
346 decoded[i] = a += b += c += decoded[i];
347 break;
348 case 4:
349 for (i = pred_order; i < blocksize; i++)
350 decoded[i] = a += b += c += d += decoded[i];
351 break;
352 default:
353 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
354 return -1;
355 }
356
357 return 0;
358 }
359
360 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order,
361 int bps)
362 {
363 int i, j;
364 int coeff_prec, qlevel;
365 int coeffs[32];
366 int32_t *decoded = s->decoded[channel];
367
368 /* warm up samples */
369 for (i = 0; i < pred_order; i++) {
370 decoded[i] = get_sbits_long(&s->gb, bps);
371 }
372
373 coeff_prec = get_bits(&s->gb, 4) + 1;
374 if (coeff_prec == 16) {
375 av_log(s->avctx, AV_LOG_ERROR, "invalid coeff precision\n");
376 return -1;
377 }
378 qlevel = get_sbits(&s->gb, 5);
379 if (qlevel < 0) {
380 av_log(s->avctx, AV_LOG_ERROR, "qlevel %d not supported, maybe buggy stream\n",
381 qlevel);
382 return -1;
383 }
384
385 for (i = 0; i < pred_order; i++) {
386 coeffs[i] = get_sbits(&s->gb, coeff_prec);
387 }
388
389 if (decode_residuals(s, channel, pred_order) < 0)
390 return -1;
391
392 if (s->bps > 16) {
393 int64_t sum;
394 for (i = pred_order; i < s->blocksize; i++) {
395 sum = 0;
396 for (j = 0; j < pred_order; j++)
397 sum += (int64_t)coeffs[j] * decoded[i-j-1];
398 decoded[i] += sum >> qlevel;
399 }
400 } else {
401 for (i = pred_order; i < s->blocksize-1; i += 2) {
402 int c;
403 int d = decoded[i-pred_order];
404 int s0 = 0, s1 = 0;
405 for (j = pred_order-1; j > 0; j--) {
406 c = coeffs[j];
407 s0 += c*d;
408 d = decoded[i-j];
409 s1 += c*d;
410 }
411 c = coeffs[0];
412 s0 += c*d;
413 d = decoded[i] += s0 >> qlevel;
414 s1 += c*d;
415 decoded[i+1] += s1 >> qlevel;
416 }
417 if (i < s->blocksize) {
418 int sum = 0;
419 for (j = 0; j < pred_order; j++)
420 sum += coeffs[j] * decoded[i-j-1];
421 decoded[i] += sum >> qlevel;
422 }
423 }
424
425 return 0;
426 }
427
428 static inline int decode_subframe(FLACContext *s, int channel)
429 {
430 int type, wasted = 0;
431 int bps = s->bps;
432 int i, tmp;
433
434 if (channel == 0) {
435 if (s->ch_mode == FLAC_CHMODE_RIGHT_SIDE)
436 bps++;
437 } else {
438 if (s->ch_mode == FLAC_CHMODE_LEFT_SIDE || s->ch_mode == FLAC_CHMODE_MID_SIDE)
439 bps++;
440 }
441
442 if (get_bits1(&s->gb)) {
443 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
444 return -1;
445 }
446 type = get_bits(&s->gb, 6);
447
448 if (get_bits1(&s->gb)) {
449 int left = get_bits_left(&s->gb);
450 wasted = 1;
451 if ( left < 0 ||
452 (left < bps && !show_bits_long(&s->gb, left)) ||
453 !show_bits_long(&s->gb, bps)) {
454 av_log(s->avctx, AV_LOG_ERROR,
455 "Invalid number of wasted bits > available bits (%d) - left=%d\n",
456 bps, left);
457 return AVERROR_INVALIDDATA;
458 }
459 while (!get_bits1(&s->gb))
460 wasted++;
461 bps -= wasted;
462 }
463 if (bps > 32) {
464 av_log_missing_feature(s->avctx, "decorrelated bit depth > 32", 0);
465 return -1;
466 }
467
468 //FIXME use av_log2 for types
469 if (type == 0) {
470 tmp = get_sbits_long(&s->gb, bps);
471 for (i = 0; i < s->blocksize; i++)
472 s->decoded[channel][i] = tmp;
473 } else if (type == 1) {
474 for (i = 0; i < s->blocksize; i++)
475 s->decoded[channel][i] = get_sbits_long(&s->gb, bps);
476 } else if ((type >= 8) && (type <= 12)) {
477 if (decode_subframe_fixed(s, channel, type & ~0x8, bps) < 0)
478 return -1;
479 } else if (type >= 32) {
480 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1, bps) < 0)
481 return -1;
482 } else {
483 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
484 return -1;
485 }
486
487 if (wasted) {
488 int i;
489 for (i = 0; i < s->blocksize; i++)
490 s->decoded[channel][i] <<= wasted;
491 }
492
493 return 0;
494 }
495
496 static int decode_frame(FLACContext *s)
497 {
498 int i;
499 GetBitContext *gb = &s->gb;
500 FLACFrameInfo fi;
501
502 if (ff_flac_decode_frame_header(s->avctx, gb, &fi, 0)) {
503 av_log(s->avctx, AV_LOG_ERROR, "invalid frame header\n");
504 return -1;
505 }
506
507 if (s->channels && fi.channels != s->channels) {
508 av_log(s->avctx, AV_LOG_ERROR, "switching channel layout mid-stream "
509 "is not supported\n");
510 return -1;
511 }
512 s->channels = s->avctx->channels = fi.channels;
513 s->ch_mode = fi.ch_mode;
514
515 if (!s->bps && !fi.bps) {
516 av_log(s->avctx, AV_LOG_ERROR, "bps not found in STREAMINFO or frame header\n");
517 return -1;
518 }
519 if (!fi.bps) {
520 fi.bps = s->bps;
521 } else if (s->bps && fi.bps != s->bps) {
522 av_log(s->avctx, AV_LOG_ERROR, "switching bps mid-stream is not "
523 "supported\n");
524 return -1;
525 }
526 s->bps = s->avctx->bits_per_raw_sample = fi.bps;
527
528 flac_set_bps(s);
529
530 if (!s->max_blocksize)
531 s->max_blocksize = FLAC_MAX_BLOCKSIZE;
532 if (fi.blocksize > s->max_blocksize) {
533 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", fi.blocksize,
534 s->max_blocksize);
535 return -1;
536 }
537 s->blocksize = fi.blocksize;
538
539 if (!s->samplerate && !fi.samplerate) {
540 av_log(s->avctx, AV_LOG_ERROR, "sample rate not found in STREAMINFO"
541 " or frame header\n");
542 return -1;
543 }
544 if (fi.samplerate == 0) {
545 fi.samplerate = s->samplerate;
546 } else if (s->samplerate && fi.samplerate != s->samplerate) {
547 av_log(s->avctx, AV_LOG_WARNING, "sample rate changed from %d to %d\n",
548 s->samplerate, fi.samplerate);
549 }
550 s->samplerate = s->avctx->sample_rate = fi.samplerate;
551
552 if (!s->got_streaminfo) {
553 allocate_buffers(s);
554 ff_flacdsp_init(&s->dsp, s->avctx->sample_fmt);
555 s->got_streaminfo = 1;
556 dump_headers(s->avctx, (FLACStreaminfo *)s);
557 }
558
559 // dump_headers(s->avctx, (FLACStreaminfo *)s);
560
561 /* subframes */
562 for (i = 0; i < s->channels; i++) {
563 if (decode_subframe(s, i) < 0)
564 return -1;
565 }
566
567 align_get_bits(gb);
568
569 /* frame footer */
570 skip_bits(gb, 16); /* data crc */
571
572 return 0;
573 }
574
575 static int flac_decode_frame(AVCodecContext *avctx, void *data,
576 int *got_frame_ptr, AVPacket *avpkt)
577 {
578 const uint8_t *buf = avpkt->data;
579 int buf_size = avpkt->size;
580 FLACContext *s = avctx->priv_data;
581 int bytes_read = 0;
582 int ret;
583
584 *got_frame_ptr = 0;
585
586 if (s->max_framesize == 0) {
587 s->max_framesize =
588 ff_flac_get_max_frame_size(s->max_blocksize ? s->max_blocksize : FLAC_MAX_BLOCKSIZE,
589 FLAC_MAX_CHANNELS, 32);
590 }
591
592 /* check that there is at least the smallest decodable amount of data.
593 this amount corresponds to the smallest valid FLAC frame possible.
594 FF F8 69 02 00 00 9A 00 00 34 46 */
595 if (buf_size < FLAC_MIN_FRAME_SIZE)
596 return buf_size;
597
598 /* check for inline header */
599 if (AV_RB32(buf) == MKBETAG('f','L','a','C')) {
600 if (!s->got_streaminfo && parse_streaminfo(s, buf, buf_size)) {
601 av_log(s->avctx, AV_LOG_ERROR, "invalid header\n");
602 return -1;
603 }
604 return get_metadata_size(buf, buf_size);
605 }
606
607 /* decode frame */
608 init_get_bits(&s->gb, buf, buf_size*8);
609 if (decode_frame(s) < 0) {
610 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
611 return -1;
612 }
613 bytes_read = (get_bits_count(&s->gb)+7)/8;
614
615 /* get output buffer */
616 s->frame.nb_samples = s->blocksize;
617 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
618 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
619 return ret;
620 }
621
622 s->dsp.decorrelate[s->ch_mode](s->frame.data, s->decoded, s->channels,
623 s->blocksize, s->sample_shift);
624
625 if (bytes_read > buf_size) {
626 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", bytes_read - buf_size);
627 return -1;
628 }
629 if (bytes_read < buf_size) {
630 av_log(s->avctx, AV_LOG_DEBUG, "underread: %d orig size: %d\n",
631 buf_size - bytes_read, buf_size);
632 }
633
634 *got_frame_ptr = 1;
635 *(AVFrame *)data = s->frame;
636
637 return bytes_read;
638 }
639
640 static av_cold int flac_decode_close(AVCodecContext *avctx)
641 {
642 FLACContext *s = avctx->priv_data;
643 int i;
644
645 for (i = 0; i < s->channels; i++) {
646 av_freep(&s->decoded[i]);
647 }
648
649 return 0;
650 }
651
652 AVCodec ff_flac_decoder = {
653 .name = "flac",
654 .type = AVMEDIA_TYPE_AUDIO,
655 .id = CODEC_ID_FLAC,
656 .priv_data_size = sizeof(FLACContext),
657 .init = flac_decode_init,
658 .close = flac_decode_close,
659 .decode = flac_decode_frame,
660 .capabilities = CODEC_CAP_DR1,
661 .long_name = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),
662 };