2 * FLAC (Free Lossless Audio Codec) decoder
3 * Copyright (c) 2003 Alex Beregszaszi
5 * This file is part of FFmpeg.
7 * FFmpeg 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.
12 * FFmpeg 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * FLAC (Free Lossless Audio Codec) decoder
25 * @author Alex Beregszaszi
27 * For more information on the FLAC format, visit:
28 * http://flac.sourceforge.net/
30 * This decoder can be used in 1 of 2 ways: Either raw FLAC data can be fed
31 * through, starting from the initial 'fLaC' signature; or by passing the
32 * 34-byte streaminfo structure through avctx->extradata[_size] followed
33 * by data starting with the 0xFFF8 marker.
38 #define ALT_BITSTREAM_READER
39 #include "libavutil/crc.h"
41 #include "bitstream.h"
48 #define MAX_CHANNELS 8
49 #define MAX_BLOCKSIZE 65535
50 #define FLAC_STREAMINFO_SIZE 34
52 enum decorrelation_type {
59 typedef struct FLACContext {
62 AVCodecContext *avctx;
65 int blocksize/*, last_blocksize*/;
67 enum decorrelation_type decorrelation;
69 int32_t *decoded[MAX_CHANNELS];
71 unsigned int bitstream_size;
72 unsigned int bitstream_index;
73 unsigned int allocated_bitstream_size;
76 static const int sample_rate_table[] =
78 88200, 176400, 192000,
79 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000,
82 static const int sample_size_table[] =
83 { 0, 8, 12, 0, 16, 20, 24, 0 };
85 static const int blocksize_table[] = {
86 0, 192, 576<<0, 576<<1, 576<<2, 576<<3, 0, 0,
87 256<<0, 256<<1, 256<<2, 256<<3, 256<<4, 256<<5, 256<<6, 256<<7
90 static int64_t get_utf8(GetBitContext *gb)
93 GET_UTF8(val, get_bits(gb, 8), return -1;)
97 static void allocate_buffers(FLACContext *s);
98 static int metadata_parse(FLACContext *s);
100 static av_cold int flac_decode_init(AVCodecContext *avctx)
102 FLACContext *s = avctx->priv_data;
105 if (avctx->extradata_size > 4) {
106 /* initialize based on the demuxer-supplied streamdata header */
107 if (avctx->extradata_size == FLAC_STREAMINFO_SIZE) {
108 ff_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s,
112 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size*8);
117 avctx->sample_fmt = SAMPLE_FMT_S16;
121 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
123 av_log(avctx, AV_LOG_DEBUG, " Blocksize: %d .. %d\n", s->min_blocksize,
125 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
126 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
127 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
128 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
131 static void allocate_buffers(FLACContext *s)
135 assert(s->max_blocksize);
137 if (s->max_framesize == 0 && s->max_blocksize) {
138 // FIXME header overhead
139 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8;
142 for (i = 0; i < s->channels; i++) {
143 s->decoded[i] = av_realloc(s->decoded[i],
144 sizeof(int32_t)*s->max_blocksize);
147 if (s->allocated_bitstream_size < s->max_framesize)
148 s->bitstream= av_fast_realloc(s->bitstream,
149 &s->allocated_bitstream_size,
153 void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
154 const uint8_t *buffer)
157 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
159 /* mandatory streaminfo */
160 s->min_blocksize = get_bits(&gb, 16);
161 s->max_blocksize = get_bits(&gb, 16);
163 skip_bits(&gb, 24); /* skip min frame size */
164 s->max_framesize = get_bits_long(&gb, 24);
166 s->samplerate = get_bits_long(&gb, 20);
167 s->channels = get_bits(&gb, 3) + 1;
168 s->bps = get_bits(&gb, 5) + 1;
170 avctx->channels = s->channels;
171 avctx->sample_rate = s->samplerate;
173 skip_bits(&gb, 36); /* total num of samples */
175 skip_bits(&gb, 64); /* md5 sum */
176 skip_bits(&gb, 64); /* md5 sum */
178 dump_headers(avctx, s);
182 * Parse a list of metadata blocks. This list of blocks must begin with
184 * @param s the flac decoding context containing the gb bit reader used to
186 * @return 1 if some metadata was read, 0 if no fLaC marker was found
188 static int metadata_parse(FLACContext *s)
190 int i, metadata_last, metadata_type, metadata_size, streaminfo_updated=0;
191 int initial_pos= get_bits_count(&s->gb);
193 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
194 skip_bits(&s->gb, 32);
197 metadata_last = get_bits1(&s->gb);
198 metadata_type = get_bits(&s->gb, 7);
199 metadata_size = get_bits_long(&s->gb, 24);
201 if (get_bits_count(&s->gb) + 8*metadata_size > s->gb.size_in_bits) {
202 skip_bits_long(&s->gb, initial_pos - get_bits_count(&s->gb));
207 switch (metadata_type) {
208 case FLAC_METADATA_TYPE_STREAMINFO:
209 ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s,
210 s->gb.buffer+get_bits_count(&s->gb)/8);
211 streaminfo_updated = 1;
214 for (i = 0; i < metadata_size; i++)
215 skip_bits(&s->gb, 8);
218 } while (!metadata_last);
220 if (streaminfo_updated)
227 static int decode_residuals(FLACContext *s, int channel, int pred_order)
229 int i, tmp, partition, method_type, rice_order;
230 int sample = 0, samples;
232 method_type = get_bits(&s->gb, 2);
233 if (method_type > 1) {
234 av_log(s->avctx, AV_LOG_ERROR, "illegal residual coding method %d\n",
239 rice_order = get_bits(&s->gb, 4);
241 samples= s->blocksize >> rice_order;
242 if (pred_order > samples) {
243 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n",
244 pred_order, samples);
250 for (partition = 0; partition < (1 << rice_order); partition++) {
251 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
252 if (tmp == (method_type == 0 ? 15 : 31)) {
253 tmp = get_bits(&s->gb, 5);
254 for (; i < samples; i++, sample++)
255 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
257 for (; i < samples; i++, sample++) {
258 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
267 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
269 const int blocksize = s->blocksize;
270 int32_t *decoded = s->decoded[channel];
273 /* warm up samples */
274 for (i = 0; i < pred_order; i++) {
275 decoded[i] = get_sbits(&s->gb, s->curr_bps);
278 if (decode_residuals(s, channel, pred_order) < 0)
282 a = decoded[pred_order-1];
284 b = a - decoded[pred_order-2];
286 c = b - decoded[pred_order-2] + decoded[pred_order-3];
288 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
290 switch (pred_order) {
294 for (i = pred_order; i < blocksize; i++)
295 decoded[i] = a += decoded[i];
298 for (i = pred_order; i < blocksize; i++)
299 decoded[i] = a += b += decoded[i];
302 for (i = pred_order; i < blocksize; i++)
303 decoded[i] = a += b += c += decoded[i];
306 for (i = pred_order; i < blocksize; i++)
307 decoded[i] = a += b += c += d += decoded[i];
310 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
317 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
320 int coeff_prec, qlevel;
321 int coeffs[pred_order];
322 int32_t *decoded = s->decoded[channel];
324 /* warm up samples */
325 for (i = 0; i < pred_order; i++) {
326 decoded[i] = get_sbits(&s->gb, s->curr_bps);
329 coeff_prec = get_bits(&s->gb, 4) + 1;
330 if (coeff_prec == 16) {
331 av_log(s->avctx, AV_LOG_ERROR, "invalid coeff precision\n");
334 qlevel = get_sbits(&s->gb, 5);
336 av_log(s->avctx, AV_LOG_ERROR, "qlevel %d not supported, maybe buggy stream\n",
341 for (i = 0; i < pred_order; i++) {
342 coeffs[i] = get_sbits(&s->gb, coeff_prec);
345 if (decode_residuals(s, channel, pred_order) < 0)
350 for (i = pred_order; i < s->blocksize; i++) {
352 for (j = 0; j < pred_order; j++)
353 sum += (int64_t)coeffs[j] * decoded[i-j-1];
354 decoded[i] += sum >> qlevel;
357 for (i = pred_order; i < s->blocksize-1; i += 2) {
359 int d = decoded[i-pred_order];
361 for (j = pred_order-1; j > 0; j--) {
369 d = decoded[i] += s0 >> qlevel;
371 decoded[i+1] += s1 >> qlevel;
373 if (i < s->blocksize) {
375 for (j = 0; j < pred_order; j++)
376 sum += coeffs[j] * decoded[i-j-1];
377 decoded[i] += sum >> qlevel;
384 static inline int decode_subframe(FLACContext *s, int channel)
386 int type, wasted = 0;
389 s->curr_bps = s->bps;
391 if (s->decorrelation == RIGHT_SIDE)
394 if (s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
398 if (get_bits1(&s->gb)) {
399 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
402 type = get_bits(&s->gb, 6);
404 if (get_bits1(&s->gb)) {
406 while (!get_bits1(&s->gb))
408 s->curr_bps -= wasted;
411 //FIXME use av_log2 for types
413 tmp = get_sbits(&s->gb, s->curr_bps);
414 for (i = 0; i < s->blocksize; i++)
415 s->decoded[channel][i] = tmp;
416 } else if (type == 1) {
417 for (i = 0; i < s->blocksize; i++)
418 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
419 } else if ((type >= 8) && (type <= 12)) {
420 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
422 } else if (type >= 32) {
423 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
426 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
432 for (i = 0; i < s->blocksize; i++)
433 s->decoded[channel][i] <<= wasted;
439 static int decode_frame(FLACContext *s, int alloc_data_size)
441 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
442 int decorrelation, bps, blocksize, samplerate;
444 blocksize_code = get_bits(&s->gb, 4);
446 sample_rate_code = get_bits(&s->gb, 4);
448 assignment = get_bits(&s->gb, 4); /* channel assignment */
449 if (assignment < 8 && s->channels == assignment+1)
450 decorrelation = INDEPENDENT;
451 else if (assignment >=8 && assignment < 11 && s->channels == 2)
452 decorrelation = LEFT_SIDE + assignment - 8;
454 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n",
455 assignment, s->channels);
459 sample_size_code = get_bits(&s->gb, 3);
460 if (sample_size_code == 0)
462 else if ((sample_size_code != 3) && (sample_size_code != 7))
463 bps = sample_size_table[sample_size_code];
465 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n",
470 if (get_bits1(&s->gb)) {
471 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
475 if (get_utf8(&s->gb) < 0) {
476 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
480 if (blocksize_code == 0)
481 blocksize = s->min_blocksize;
482 else if (blocksize_code == 6)
483 blocksize = get_bits(&s->gb, 8)+1;
484 else if (blocksize_code == 7)
485 blocksize = get_bits(&s->gb, 16)+1;
487 blocksize = blocksize_table[blocksize_code];
489 if (blocksize > s->max_blocksize) {
490 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize,
495 if (blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
498 if (sample_rate_code == 0)
499 samplerate= s->samplerate;
500 else if (sample_rate_code < 12)
501 samplerate = sample_rate_table[sample_rate_code];
502 else if (sample_rate_code == 12)
503 samplerate = get_bits(&s->gb, 8) * 1000;
504 else if (sample_rate_code == 13)
505 samplerate = get_bits(&s->gb, 16);
506 else if (sample_rate_code == 14)
507 samplerate = get_bits(&s->gb, 16) * 10;
509 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n",
514 skip_bits(&s->gb, 8);
515 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
516 s->gb.buffer, get_bits_count(&s->gb)/8);
518 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
522 s->blocksize = blocksize;
523 s->samplerate = samplerate;
525 s->decorrelation= decorrelation;
527 // dump_headers(s->avctx, (FLACStreaminfo *)s);
530 for (i = 0; i < s->channels; i++) {
531 if (decode_subframe(s, i) < 0)
535 align_get_bits(&s->gb);
538 skip_bits(&s->gb, 16); /* data crc */
543 static int flac_decode_frame(AVCodecContext *avctx,
544 void *data, int *data_size,
545 const uint8_t *buf, int buf_size)
547 FLACContext *s = avctx->priv_data;
548 int tmp = 0, i, j = 0, input_buf_size = 0;
549 int16_t *samples = data;
550 int alloc_data_size= *data_size;
554 if (s->max_framesize == 0) {
555 s->max_framesize= FFMAX(4, buf_size); // should hopefully be enough for the first header
556 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
559 if (1 && s->max_framesize) { //FIXME truncated
560 if (s->bitstream_size < 4 || AV_RL32(s->bitstream) != MKTAG('f','L','a','C'))
561 buf_size= FFMIN(buf_size, s->max_framesize - FFMIN(s->bitstream_size, s->max_framesize));
562 input_buf_size= buf_size;
564 if (s->bitstream_size + buf_size < buf_size || s->bitstream_index + s->bitstream_size + buf_size < s->bitstream_index)
567 if (s->allocated_bitstream_size < s->bitstream_size + buf_size)
568 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->bitstream_size + buf_size);
570 if (s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size) {
571 memmove(s->bitstream, &s->bitstream[s->bitstream_index],
573 s->bitstream_index=0;
575 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size],
577 buf= &s->bitstream[s->bitstream_index];
578 buf_size += s->bitstream_size;
579 s->bitstream_size= buf_size;
581 if (buf_size < s->max_framesize && input_buf_size) {
582 return input_buf_size;
586 init_get_bits(&s->gb, buf, buf_size*8);
588 if (metadata_parse(s))
591 tmp = show_bits(&s->gb, 16);
592 if ((tmp & 0xFFFE) != 0xFFF8) {
593 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
594 while (get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
595 skip_bits(&s->gb, 8);
596 goto end; // we may not have enough bits left to decode a frame, so try next time
598 skip_bits(&s->gb, 16);
599 if (decode_frame(s, alloc_data_size) < 0) {
600 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
602 s->bitstream_index=0;
606 #define DECORRELATE(left, right)\
607 assert(s->channels == 2);\
608 for (i = 0; i < s->blocksize; i++) {\
609 int a= s->decoded[0][i];\
610 int b= s->decoded[1][i];\
611 *samples++ = ((left) << (24 - s->bps)) >> 8;\
612 *samples++ = ((right) << (24 - s->bps)) >> 8;\
616 switch (s->decorrelation) {
618 for (j = 0; j < s->blocksize; j++) {
619 for (i = 0; i < s->channels; i++)
620 *samples++ = (s->decoded[i][j] << (24 - s->bps)) >> 8;
628 DECORRELATE( (a-=b>>1) + b, a)
631 *data_size = (int8_t *)samples - (int8_t *)data;
634 i= (get_bits_count(&s->gb)+7)/8;
636 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
638 s->bitstream_index=0;
642 if (s->bitstream_size) {
643 s->bitstream_index += i;
644 s->bitstream_size -= i;
645 return input_buf_size;
650 static av_cold int flac_decode_close(AVCodecContext *avctx)
652 FLACContext *s = avctx->priv_data;
655 for (i = 0; i < s->channels; i++) {
656 av_freep(&s->decoded[i]);
658 av_freep(&s->bitstream);
663 static void flac_flush(AVCodecContext *avctx)
665 FLACContext *s = avctx->priv_data;
668 s->bitstream_index= 0;
671 AVCodec flac_decoder = {
682 .long_name= NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),