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
40 #include "bitstream.h"
47 #define MAX_CHANNELS 8
48 #define MAX_BLOCKSIZE 65535
49 #define FLAC_STREAMINFO_SIZE 34
51 enum decorrelation_type {
58 typedef struct FLACContext {
59 AVCodecContext *avctx;
62 int min_blocksize, max_blocksize;
63 int min_framesize, max_framesize;
64 int samplerate, channels;
65 int blocksize/*, last_blocksize*/;
67 enum decorrelation_type decorrelation;
69 int32_t *decoded[MAX_CHANNELS];
73 unsigned int allocated_bitstream_size;
76 #define METADATA_TYPE_STREAMINFO 0
78 static int sample_rate_table[] =
80 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000,
83 static int sample_size_table[] =
84 { 0, 8, 12, 0, 16, 20, 24, 0 };
86 static int blocksize_table[] = {
87 0, 192, 576<<0, 576<<1, 576<<2, 576<<3, 0, 0,
88 256<<0, 256<<1, 256<<2, 256<<3, 256<<4, 256<<5, 256<<6, 256<<7
91 static int64_t get_utf8(GetBitContext *gb){
93 GET_UTF8(val, get_bits(gb, 8), return -1;)
97 static void metadata_streaminfo(FLACContext *s);
98 static void allocate_buffers(FLACContext *s);
99 static int metadata_parse(FLACContext *s);
101 static int flac_decode_init(AVCodecContext * avctx)
103 FLACContext *s = avctx->priv_data;
106 if (avctx->extradata_size > 4) {
107 /* initialize based on the demuxer-supplied streamdata header */
108 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size*8);
109 if (avctx->extradata_size == FLAC_STREAMINFO_SIZE) {
110 metadata_streaminfo(s);
120 static void dump_headers(FLACContext *s)
122 av_log(s->avctx, AV_LOG_DEBUG, " Blocksize: %d .. %d (%d)\n", s->min_blocksize, s->max_blocksize, s->blocksize);
123 av_log(s->avctx, AV_LOG_DEBUG, " Framesize: %d .. %d\n", s->min_framesize, s->max_framesize);
124 av_log(s->avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
125 av_log(s->avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
126 av_log(s->avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
129 static void allocate_buffers(FLACContext *s){
132 assert(s->max_blocksize);
134 if(s->max_framesize == 0 && s->max_blocksize){
135 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8; //FIXME header overhead
138 for (i = 0; i < s->channels; i++)
140 s->decoded[i] = av_realloc(s->decoded[i], sizeof(int32_t)*s->max_blocksize);
143 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
146 static void metadata_streaminfo(FLACContext *s)
148 /* mandatory streaminfo */
149 s->min_blocksize = get_bits(&s->gb, 16);
150 s->max_blocksize = get_bits(&s->gb, 16);
152 s->min_framesize = get_bits_long(&s->gb, 24);
153 s->max_framesize = get_bits_long(&s->gb, 24);
155 s->samplerate = get_bits_long(&s->gb, 20);
156 s->channels = get_bits(&s->gb, 3) + 1;
157 s->bps = get_bits(&s->gb, 5) + 1;
159 s->avctx->channels = s->channels;
160 s->avctx->sample_rate = s->samplerate;
162 skip_bits(&s->gb, 36); /* total num of samples */
164 skip_bits(&s->gb, 64); /* md5 sum */
165 skip_bits(&s->gb, 64); /* md5 sum */
171 * Parse a list of metadata blocks. This list of blocks must begin with
173 * @param s the flac decoding context containing the gb bit reader used to
175 * @return 1 if some metadata was read, 0 if no fLaC marker was found
177 static int metadata_parse(FLACContext *s)
179 int i, metadata_last, metadata_type, metadata_size, streaminfo_updated=0;
181 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
182 skip_bits(&s->gb, 32);
184 av_log(s->avctx, AV_LOG_DEBUG, "STREAM HEADER\n");
186 metadata_last = get_bits(&s->gb, 1);
187 metadata_type = get_bits(&s->gb, 7);
188 metadata_size = get_bits_long(&s->gb, 24);
190 av_log(s->avctx, AV_LOG_DEBUG,
191 " metadata block: flag = %d, type = %d, size = %d\n",
192 metadata_last, metadata_type, metadata_size);
194 switch (metadata_type) {
195 case METADATA_TYPE_STREAMINFO:
196 metadata_streaminfo(s);
197 streaminfo_updated = 1;
201 for (i=0; i<metadata_size; i++)
202 skip_bits(&s->gb, 8);
205 } while (!metadata_last);
207 if (streaminfo_updated)
214 static int decode_residuals(FLACContext *s, int channel, int pred_order)
216 int i, tmp, partition, method_type, rice_order;
217 int sample = 0, samples;
219 method_type = get_bits(&s->gb, 2);
220 if (method_type != 0){
221 av_log(s->avctx, AV_LOG_DEBUG, "illegal residual coding method %d\n", method_type);
225 rice_order = get_bits(&s->gb, 4);
227 samples= s->blocksize >> rice_order;
231 for (partition = 0; partition < (1 << rice_order); partition++)
233 tmp = get_bits(&s->gb, 4);
236 av_log(s->avctx, AV_LOG_DEBUG, "fixed len partition\n");
237 tmp = get_bits(&s->gb, 5);
238 for (; i < samples; i++, sample++)
239 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
243 // av_log(s->avctx, AV_LOG_DEBUG, "rice coded partition k=%d\n", tmp);
244 for (; i < samples; i++, sample++){
245 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
251 // av_log(s->avctx, AV_LOG_DEBUG, "partitions: %d, samples: %d\n", 1 << rice_order, sample);
256 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
260 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME FIXED\n");
262 /* warm up samples */
263 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
265 for (i = 0; i < pred_order; i++)
267 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
268 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
271 if (decode_residuals(s, channel, pred_order) < 0)
279 for (i = pred_order; i < s->blocksize; i++)
280 s->decoded[channel][i] += s->decoded[channel][i-1];
283 for (i = pred_order; i < s->blocksize; i++)
284 s->decoded[channel][i] += 2*s->decoded[channel][i-1]
285 - s->decoded[channel][i-2];
288 for (i = pred_order; i < s->blocksize; i++)
289 s->decoded[channel][i] += 3*s->decoded[channel][i-1]
290 - 3*s->decoded[channel][i-2]
291 + s->decoded[channel][i-3];
294 for (i = pred_order; i < s->blocksize; i++)
295 s->decoded[channel][i] += 4*s->decoded[channel][i-1]
296 - 6*s->decoded[channel][i-2]
297 + 4*s->decoded[channel][i-3]
298 - s->decoded[channel][i-4];
301 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
308 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
311 int coeff_prec, qlevel;
312 int coeffs[pred_order];
314 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
316 /* warm up samples */
317 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
319 for (i = 0; i < pred_order; i++)
321 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
322 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
325 coeff_prec = get_bits(&s->gb, 4) + 1;
326 if (coeff_prec == 16)
328 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
331 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
332 qlevel = get_sbits(&s->gb, 5);
333 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
335 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
339 for (i = 0; i < pred_order; i++)
341 coeffs[i] = get_sbits(&s->gb, coeff_prec);
342 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
345 if (decode_residuals(s, channel, pred_order) < 0)
350 for (i = pred_order; i < s->blocksize; i++)
353 for (j = 0; j < pred_order; j++)
354 sum += (int64_t)coeffs[j] * s->decoded[channel][i-j-1];
355 s->decoded[channel][i] += sum >> qlevel;
359 for (i = pred_order; i < s->blocksize; i++)
362 for (j = 0; j < pred_order; j++)
363 sum += coeffs[j] * s->decoded[channel][i-j-1];
364 s->decoded[channel][i] += sum >> qlevel;
371 static inline int decode_subframe(FLACContext *s, int channel)
373 int type, wasted = 0;
376 s->curr_bps = s->bps;
378 if(s->decorrelation == RIGHT_SIDE)
381 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
385 if (get_bits1(&s->gb))
387 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
390 type = get_bits(&s->gb, 6);
391 // wasted = get_bits1(&s->gb);
395 // while (!get_bits1(&s->gb))
399 // s->curr_bps -= wasted;
402 wasted= 16 - av_log2(show_bits(&s->gb, 17));
403 skip_bits(&s->gb, wasted+1);
404 s->curr_bps -= wasted;
406 if (get_bits1(&s->gb))
409 while (!get_bits1(&s->gb))
411 s->curr_bps -= wasted;
412 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
415 //FIXME use av_log2 for types
418 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
419 tmp = get_sbits(&s->gb, s->curr_bps);
420 for (i = 0; i < s->blocksize; i++)
421 s->decoded[channel][i] = tmp;
425 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
426 for (i = 0; i < s->blocksize; i++)
427 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
429 else if ((type >= 8) && (type <= 12))
431 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
432 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
437 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
438 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
443 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
450 for (i = 0; i < s->blocksize; i++)
451 s->decoded[channel][i] <<= wasted;
457 static int decode_frame(FLACContext *s)
459 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
460 int decorrelation, bps, blocksize, samplerate;
462 blocksize_code = get_bits(&s->gb, 4);
464 sample_rate_code = get_bits(&s->gb, 4);
466 assignment = get_bits(&s->gb, 4); /* channel assignment */
467 if (assignment < 8 && s->channels == assignment+1)
468 decorrelation = INDEPENDENT;
469 else if (assignment >=8 && assignment < 11 && s->channels == 2)
470 decorrelation = LEFT_SIDE + assignment - 8;
473 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
477 sample_size_code = get_bits(&s->gb, 3);
478 if(sample_size_code == 0)
480 else if((sample_size_code != 3) && (sample_size_code != 7))
481 bps = sample_size_table[sample_size_code];
484 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
488 if (get_bits1(&s->gb))
490 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
494 if(get_utf8(&s->gb) < 0){
495 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
499 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
500 (s->min_blocksize != s->max_blocksize)){
505 if (blocksize_code == 0)
506 blocksize = s->min_blocksize;
507 else if (blocksize_code == 6)
508 blocksize = get_bits(&s->gb, 8)+1;
509 else if (blocksize_code == 7)
510 blocksize = get_bits(&s->gb, 16)+1;
512 blocksize = blocksize_table[blocksize_code];
514 if(blocksize > s->max_blocksize){
515 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
519 if (sample_rate_code == 0){
520 samplerate= s->samplerate;
521 }else if ((sample_rate_code > 3) && (sample_rate_code < 12))
522 samplerate = sample_rate_table[sample_rate_code];
523 else if (sample_rate_code == 12)
524 samplerate = get_bits(&s->gb, 8) * 1000;
525 else if (sample_rate_code == 13)
526 samplerate = get_bits(&s->gb, 16);
527 else if (sample_rate_code == 14)
528 samplerate = get_bits(&s->gb, 16) * 10;
530 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
534 skip_bits(&s->gb, 8);
535 crc8= av_crc(av_crc07, 0, s->gb.buffer, get_bits_count(&s->gb)/8);
537 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
541 s->blocksize = blocksize;
542 s->samplerate = samplerate;
544 s->decorrelation= decorrelation;
549 for (i = 0; i < s->channels; i++)
551 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
552 if (decode_subframe(s, i) < 0)
556 align_get_bits(&s->gb);
559 skip_bits(&s->gb, 16); /* data crc */
564 static inline int16_t shift_to_16_bits(int32_t data, int bps)
568 } else if (bps == 20) {
575 static int flac_decode_frame(AVCodecContext *avctx,
576 void *data, int *data_size,
577 uint8_t *buf, int buf_size)
579 FLACContext *s = avctx->priv_data;
580 int tmp = 0, i, j = 0, input_buf_size = 0;
581 int16_t *samples = data;
583 if(s->max_framesize == 0){
584 s->max_framesize= 65536; // should hopefully be enough for the first header
585 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
588 if(1 && s->max_framesize){//FIXME truncated
589 buf_size= FFMAX(FFMIN(buf_size, s->max_framesize - s->bitstream_size), 0);
590 input_buf_size= buf_size;
592 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
593 // printf("memmove\n");
594 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
595 s->bitstream_index=0;
597 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
598 buf= &s->bitstream[s->bitstream_index];
599 buf_size += s->bitstream_size;
600 s->bitstream_size= buf_size;
602 if(buf_size < s->max_framesize){
603 // printf("wanna more data ...\n");
604 return input_buf_size;
608 init_get_bits(&s->gb, buf, buf_size*8);
610 if (!metadata_parse(s))
612 tmp = show_bits(&s->gb, 16);
614 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
615 while(get_bits_count(&s->gb)/8+2 < buf_size && show_bits(&s->gb, 16) != 0xFFF8)
616 skip_bits(&s->gb, 8);
617 goto end; // we may not have enough bits left to decode a frame, so try next time
619 skip_bits(&s->gb, 16);
620 if (decode_frame(s) < 0){
621 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
623 s->bitstream_index=0;
630 /* fix the channel order here */
631 if (s->order == MID_SIDE)
633 short *left = samples;
634 short *right = samples + s->blocksize;
635 for (i = 0; i < s->blocksize; i += 2)
637 uint32_t x = s->decoded[0][i];
638 uint32_t y = s->decoded[0][i+1];
640 right[i] = x - (y / 2);
641 left[i] = right[i] + y;
643 *data_size = 2 * s->blocksize;
647 for (i = 0; i < s->channels; i++)
652 for (j = 0; j < s->blocksize; j++)
653 samples[(s->blocksize*i)+j] = s->decoded[i][j];
658 for (j = 0; j < s->blocksize; j++)
659 samples[(s->blocksize*i)+j] = s->decoded[0][j];
661 for (j = 0; j < s->blocksize; j++)
662 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
665 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
667 *data_size += s->blocksize;
671 #define DECORRELATE(left, right)\
672 assert(s->channels == 2);\
673 for (i = 0; i < s->blocksize; i++)\
675 int a= s->decoded[0][i];\
676 int b= s->decoded[1][i];\
677 *(samples++) = (left ) >> (16 - s->bps);\
678 *(samples++) = (right) >> (16 - s->bps);\
682 switch(s->decorrelation)
685 for (j = 0; j < s->blocksize; j++)
687 for (i = 0; i < s->channels; i++)
688 *(samples++) = shift_to_16_bits(s->decoded[i][j], s->bps);
696 DECORRELATE( (a-=b>>1) + b, a)
700 *data_size = (int8_t *)samples - (int8_t *)data;
701 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
703 // s->last_blocksize = s->blocksize;
705 i= (get_bits_count(&s->gb)+7)/8;;
707 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
709 s->bitstream_index=0;
713 if(s->bitstream_size){
714 s->bitstream_index += i;
715 s->bitstream_size -= i;
716 return input_buf_size;
721 static int flac_decode_close(AVCodecContext *avctx)
723 FLACContext *s = avctx->priv_data;
726 for (i = 0; i < s->channels; i++)
728 av_freep(&s->decoded[i]);
730 av_freep(&s->bitstream);
735 static void flac_flush(AVCodecContext *avctx){
736 FLACContext *s = avctx->priv_data;
739 s->bitstream_index= 0;
742 AVCodec flac_decoder = {