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 #define METADATA_TYPE_STREAMINFO 0
78 static const int sample_rate_table[] =
80 88200, 176400, 192000,
81 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000,
84 static const int sample_size_table[] =
85 { 0, 8, 12, 0, 16, 20, 24, 0 };
87 static const int blocksize_table[] = {
88 0, 192, 576<<0, 576<<1, 576<<2, 576<<3, 0, 0,
89 256<<0, 256<<1, 256<<2, 256<<3, 256<<4, 256<<5, 256<<6, 256<<7
92 static int64_t get_utf8(GetBitContext *gb){
94 GET_UTF8(val, get_bits(gb, 8), return -1;)
98 static void allocate_buffers(FLACContext *s);
99 static int metadata_parse(FLACContext *s);
101 static av_cold 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 if (avctx->extradata_size == FLAC_STREAMINFO_SIZE) {
109 ff_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s, avctx->extradata);
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, s->max_blocksize);
124 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
125 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
126 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
127 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
130 static void allocate_buffers(FLACContext *s){
133 assert(s->max_blocksize);
135 if(s->max_framesize == 0 && s->max_blocksize){
136 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8; //FIXME header overhead
139 for (i = 0; i < s->channels; i++)
141 s->decoded[i] = av_realloc(s->decoded[i], sizeof(int32_t)*s->max_blocksize);
144 if(s->allocated_bitstream_size < s->max_framesize)
145 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
148 void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
149 const uint8_t *buffer)
152 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
154 /* mandatory streaminfo */
155 s->min_blocksize = get_bits(&gb, 16);
156 s->max_blocksize = get_bits(&gb, 16);
158 skip_bits(&gb, 24); /* skip min frame size */
159 s->max_framesize = get_bits_long(&gb, 24);
161 s->samplerate = get_bits_long(&gb, 20);
162 s->channels = get_bits(&gb, 3) + 1;
163 s->bps = get_bits(&gb, 5) + 1;
165 avctx->channels = s->channels;
166 avctx->sample_rate = s->samplerate;
168 skip_bits(&gb, 36); /* total num of samples */
170 skip_bits(&gb, 64); /* md5 sum */
171 skip_bits(&gb, 64); /* md5 sum */
173 dump_headers(avctx, s);
177 * Parse a list of metadata blocks. This list of blocks must begin with
179 * @param s the flac decoding context containing the gb bit reader used to
181 * @return 1 if some metadata was read, 0 if no fLaC marker was found
183 static int metadata_parse(FLACContext *s)
185 int i, metadata_last, metadata_type, metadata_size, streaminfo_updated=0;
186 int initial_pos= get_bits_count(&s->gb);
188 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
189 skip_bits(&s->gb, 32);
191 av_log(s->avctx, AV_LOG_DEBUG, "STREAM HEADER\n");
193 metadata_last = get_bits1(&s->gb);
194 metadata_type = get_bits(&s->gb, 7);
195 metadata_size = get_bits_long(&s->gb, 24);
197 if(get_bits_count(&s->gb) + 8*metadata_size > s->gb.size_in_bits){
198 skip_bits_long(&s->gb, initial_pos - get_bits_count(&s->gb));
202 av_log(s->avctx, AV_LOG_DEBUG,
203 " metadata block: flag = %d, type = %d, size = %d\n",
204 metadata_last, metadata_type, metadata_size);
206 switch (metadata_type) {
207 case METADATA_TYPE_STREAMINFO:
208 ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, s->gb.buffer+get_bits_count(&s->gb)/8);
209 streaminfo_updated = 1;
212 for (i=0; i<metadata_size; i++)
213 skip_bits(&s->gb, 8);
216 } while (!metadata_last);
218 if (streaminfo_updated)
225 static int decode_residuals(FLACContext *s, int channel, int pred_order)
227 int i, tmp, partition, method_type, rice_order;
228 int sample = 0, samples;
230 method_type = get_bits(&s->gb, 2);
231 if (method_type > 1){
232 av_log(s->avctx, AV_LOG_DEBUG, "illegal residual coding method %d\n", method_type);
236 rice_order = get_bits(&s->gb, 4);
238 samples= s->blocksize >> rice_order;
239 if (pred_order > samples) {
240 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n", pred_order, samples);
246 for (partition = 0; partition < (1 << rice_order); partition++)
248 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
249 if (tmp == (method_type == 0 ? 15 : 31))
251 av_log(s->avctx, AV_LOG_DEBUG, "fixed len partition\n");
252 tmp = get_bits(&s->gb, 5);
253 for (; i < samples; i++, sample++)
254 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
258 // av_log(s->avctx, AV_LOG_DEBUG, "rice coded partition k=%d\n", tmp);
259 for (; i < samples; i++, sample++){
260 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
266 // av_log(s->avctx, AV_LOG_DEBUG, "partitions: %d, samples: %d\n", 1 << rice_order, sample);
271 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
273 const int blocksize = s->blocksize;
274 int32_t *decoded = s->decoded[channel];
277 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME FIXED\n");
279 /* warm up samples */
280 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
282 for (i = 0; i < pred_order; i++)
284 decoded[i] = get_sbits(&s->gb, s->curr_bps);
285 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
288 if (decode_residuals(s, channel, pred_order) < 0)
292 a = decoded[pred_order-1];
294 b = a - decoded[pred_order-2];
296 c = b - decoded[pred_order-2] + decoded[pred_order-3];
298 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
305 for (i = pred_order; i < blocksize; i++)
306 decoded[i] = a += decoded[i];
309 for (i = pred_order; i < blocksize; i++)
310 decoded[i] = a += b += decoded[i];
313 for (i = pred_order; i < blocksize; i++)
314 decoded[i] = a += b += c += decoded[i];
317 for (i = pred_order; i < blocksize; i++)
318 decoded[i] = a += b += c += d += decoded[i];
321 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
328 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
331 int coeff_prec, qlevel;
332 int coeffs[pred_order];
333 int32_t *decoded = s->decoded[channel];
335 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
337 /* warm up samples */
338 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
340 for (i = 0; i < pred_order; i++)
342 decoded[i] = get_sbits(&s->gb, s->curr_bps);
343 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, decoded[i]);
346 coeff_prec = get_bits(&s->gb, 4) + 1;
347 if (coeff_prec == 16)
349 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
352 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
353 qlevel = get_sbits(&s->gb, 5);
354 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
356 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
360 for (i = 0; i < pred_order; i++)
362 coeffs[i] = get_sbits(&s->gb, coeff_prec);
363 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
366 if (decode_residuals(s, channel, pred_order) < 0)
371 for (i = pred_order; i < s->blocksize; i++)
374 for (j = 0; j < pred_order; j++)
375 sum += (int64_t)coeffs[j] * decoded[i-j-1];
376 decoded[i] += sum >> qlevel;
379 for (i = pred_order; i < s->blocksize-1; i += 2)
382 int d = decoded[i-pred_order];
384 for (j = pred_order-1; j > 0; j--)
393 d = decoded[i] += s0 >> qlevel;
395 decoded[i+1] += s1 >> qlevel;
397 if (i < s->blocksize)
400 for (j = 0; j < pred_order; j++)
401 sum += coeffs[j] * decoded[i-j-1];
402 decoded[i] += sum >> qlevel;
409 static inline int decode_subframe(FLACContext *s, int channel)
411 int type, wasted = 0;
414 s->curr_bps = s->bps;
416 if(s->decorrelation == RIGHT_SIDE)
419 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
423 if (get_bits1(&s->gb))
425 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
428 type = get_bits(&s->gb, 6);
429 // wasted = get_bits1(&s->gb);
433 // while (!get_bits1(&s->gb))
437 // s->curr_bps -= wasted;
440 wasted= 16 - av_log2(show_bits(&s->gb, 17));
441 skip_bits(&s->gb, wasted+1);
442 s->curr_bps -= wasted;
444 if (get_bits1(&s->gb))
447 while (!get_bits1(&s->gb))
449 s->curr_bps -= wasted;
450 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
453 //FIXME use av_log2 for types
456 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
457 tmp = get_sbits(&s->gb, s->curr_bps);
458 for (i = 0; i < s->blocksize; i++)
459 s->decoded[channel][i] = tmp;
463 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
464 for (i = 0; i < s->blocksize; i++)
465 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
467 else if ((type >= 8) && (type <= 12))
469 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
470 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
475 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
476 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
481 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
488 for (i = 0; i < s->blocksize; i++)
489 s->decoded[channel][i] <<= wasted;
495 static int decode_frame(FLACContext *s, int alloc_data_size)
497 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
498 int decorrelation, bps, blocksize, samplerate;
500 blocksize_code = get_bits(&s->gb, 4);
502 sample_rate_code = get_bits(&s->gb, 4);
504 assignment = get_bits(&s->gb, 4); /* channel assignment */
505 if (assignment < 8 && s->channels == assignment+1)
506 decorrelation = INDEPENDENT;
507 else if (assignment >=8 && assignment < 11 && s->channels == 2)
508 decorrelation = LEFT_SIDE + assignment - 8;
511 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
515 sample_size_code = get_bits(&s->gb, 3);
516 if(sample_size_code == 0)
518 else if((sample_size_code != 3) && (sample_size_code != 7))
519 bps = sample_size_table[sample_size_code];
522 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
526 if (get_bits1(&s->gb))
528 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
532 if(get_utf8(&s->gb) < 0){
533 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
537 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
538 (s->min_blocksize != s->max_blocksize)){
543 if (blocksize_code == 0)
544 blocksize = s->min_blocksize;
545 else if (blocksize_code == 6)
546 blocksize = get_bits(&s->gb, 8)+1;
547 else if (blocksize_code == 7)
548 blocksize = get_bits(&s->gb, 16)+1;
550 blocksize = blocksize_table[blocksize_code];
552 if(blocksize > s->max_blocksize){
553 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
557 if(blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
560 if (sample_rate_code == 0){
561 samplerate= s->samplerate;
562 }else if (sample_rate_code < 12)
563 samplerate = sample_rate_table[sample_rate_code];
564 else if (sample_rate_code == 12)
565 samplerate = get_bits(&s->gb, 8) * 1000;
566 else if (sample_rate_code == 13)
567 samplerate = get_bits(&s->gb, 16);
568 else if (sample_rate_code == 14)
569 samplerate = get_bits(&s->gb, 16) * 10;
571 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
575 skip_bits(&s->gb, 8);
576 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
577 s->gb.buffer, get_bits_count(&s->gb)/8);
579 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
583 s->blocksize = blocksize;
584 s->samplerate = samplerate;
586 s->decorrelation= decorrelation;
588 // dump_headers(s->avctx, (FLACStreaminfo *)s);
591 for (i = 0; i < s->channels; i++)
593 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
594 if (decode_subframe(s, i) < 0)
598 align_get_bits(&s->gb);
601 skip_bits(&s->gb, 16); /* data crc */
606 static int flac_decode_frame(AVCodecContext *avctx,
607 void *data, int *data_size,
608 const uint8_t *buf, int buf_size)
610 FLACContext *s = avctx->priv_data;
611 int tmp = 0, i, j = 0, input_buf_size = 0;
612 int16_t *samples = data;
613 int alloc_data_size= *data_size;
617 if(s->max_framesize == 0){
618 s->max_framesize= FFMAX(4, buf_size); // should hopefully be enough for the first header
619 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
622 if(1 && s->max_framesize){//FIXME truncated
623 if(s->bitstream_size < 4 || AV_RL32(s->bitstream) != MKTAG('f','L','a','C'))
624 buf_size= FFMIN(buf_size, s->max_framesize - FFMIN(s->bitstream_size, s->max_framesize));
625 input_buf_size= buf_size;
627 if(s->bitstream_size + buf_size < buf_size || s->bitstream_index + s->bitstream_size + buf_size < s->bitstream_index)
630 if(s->allocated_bitstream_size < s->bitstream_size + buf_size)
631 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->bitstream_size + buf_size);
633 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
634 // printf("memmove\n");
635 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
636 s->bitstream_index=0;
638 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
639 buf= &s->bitstream[s->bitstream_index];
640 buf_size += s->bitstream_size;
641 s->bitstream_size= buf_size;
643 if(buf_size < s->max_framesize && input_buf_size){
644 // printf("wanna more data ...\n");
645 return input_buf_size;
649 init_get_bits(&s->gb, buf, buf_size*8);
651 if(metadata_parse(s))
654 tmp = show_bits(&s->gb, 16);
655 if((tmp & 0xFFFE) != 0xFFF8){
656 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
657 while(get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
658 skip_bits(&s->gb, 8);
659 goto end; // we may not have enough bits left to decode a frame, so try next time
661 skip_bits(&s->gb, 16);
662 if (decode_frame(s, alloc_data_size) < 0){
663 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
665 s->bitstream_index=0;
671 /* fix the channel order here */
672 if (s->order == MID_SIDE)
674 short *left = samples;
675 short *right = samples + s->blocksize;
676 for (i = 0; i < s->blocksize; i += 2)
678 uint32_t x = s->decoded[0][i];
679 uint32_t y = s->decoded[0][i+1];
681 right[i] = x - (y / 2);
682 left[i] = right[i] + y;
684 *data_size = 2 * s->blocksize;
688 for (i = 0; i < s->channels; i++)
693 for (j = 0; j < s->blocksize; j++)
694 samples[(s->blocksize*i)+j] = s->decoded[i][j];
699 for (j = 0; j < s->blocksize; j++)
700 samples[(s->blocksize*i)+j] = s->decoded[0][j];
702 for (j = 0; j < s->blocksize; j++)
703 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
706 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
708 *data_size += s->blocksize;
712 #define DECORRELATE(left, right)\
713 assert(s->channels == 2);\
714 for (i = 0; i < s->blocksize; i++)\
716 int a= s->decoded[0][i];\
717 int b= s->decoded[1][i];\
718 *samples++ = ((left) << (24 - s->bps)) >> 8;\
719 *samples++ = ((right) << (24 - s->bps)) >> 8;\
723 switch(s->decorrelation)
726 for (j = 0; j < s->blocksize; j++)
728 for (i = 0; i < s->channels; i++)
729 *samples++ = (s->decoded[i][j] << (24 - s->bps)) >> 8;
737 DECORRELATE( (a-=b>>1) + b, a)
741 *data_size = (int8_t *)samples - (int8_t *)data;
742 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
744 // s->last_blocksize = s->blocksize;
746 i= (get_bits_count(&s->gb)+7)/8;
748 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
750 s->bitstream_index=0;
754 if(s->bitstream_size){
755 s->bitstream_index += i;
756 s->bitstream_size -= i;
757 return input_buf_size;
762 static av_cold int flac_decode_close(AVCodecContext *avctx)
764 FLACContext *s = avctx->priv_data;
767 for (i = 0; i < s->channels; i++)
769 av_freep(&s->decoded[i]);
771 av_freep(&s->bitstream);
776 static void flac_flush(AVCodecContext *avctx){
777 FLACContext *s = avctx->priv_data;
780 s->bitstream_index= 0;
783 AVCodec flac_decoder = {
794 .long_name= NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),