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;
228 if (pred_order > samples) {
229 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n", pred_order, samples);
235 for (partition = 0; partition < (1 << rice_order); partition++)
237 tmp = get_bits(&s->gb, 4);
240 av_log(s->avctx, AV_LOG_DEBUG, "fixed len partition\n");
241 tmp = get_bits(&s->gb, 5);
242 for (; i < samples; i++, sample++)
243 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
247 // av_log(s->avctx, AV_LOG_DEBUG, "rice coded partition k=%d\n", tmp);
248 for (; i < samples; i++, sample++){
249 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
255 // av_log(s->avctx, AV_LOG_DEBUG, "partitions: %d, samples: %d\n", 1 << rice_order, sample);
260 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
264 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME FIXED\n");
266 /* warm up samples */
267 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
269 for (i = 0; i < pred_order; i++)
271 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
272 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
275 if (decode_residuals(s, channel, pred_order) < 0)
283 for (i = pred_order; i < s->blocksize; i++)
284 s->decoded[channel][i] += s->decoded[channel][i-1];
287 for (i = pred_order; i < s->blocksize; i++)
288 s->decoded[channel][i] += 2*s->decoded[channel][i-1]
289 - s->decoded[channel][i-2];
292 for (i = pred_order; i < s->blocksize; i++)
293 s->decoded[channel][i] += 3*s->decoded[channel][i-1]
294 - 3*s->decoded[channel][i-2]
295 + s->decoded[channel][i-3];
298 for (i = pred_order; i < s->blocksize; i++)
299 s->decoded[channel][i] += 4*s->decoded[channel][i-1]
300 - 6*s->decoded[channel][i-2]
301 + 4*s->decoded[channel][i-3]
302 - s->decoded[channel][i-4];
305 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
312 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
315 int coeff_prec, qlevel;
316 int coeffs[pred_order];
318 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
320 /* warm up samples */
321 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
323 for (i = 0; i < pred_order; i++)
325 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
326 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
329 coeff_prec = get_bits(&s->gb, 4) + 1;
330 if (coeff_prec == 16)
332 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
335 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
336 qlevel = get_sbits(&s->gb, 5);
337 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
339 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
343 for (i = 0; i < pred_order; i++)
345 coeffs[i] = get_sbits(&s->gb, coeff_prec);
346 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
349 if (decode_residuals(s, channel, pred_order) < 0)
354 for (i = pred_order; i < s->blocksize; i++)
357 for (j = 0; j < pred_order; j++)
358 sum += (int64_t)coeffs[j] * s->decoded[channel][i-j-1];
359 s->decoded[channel][i] += sum >> qlevel;
363 for (i = pred_order; i < s->blocksize; i++)
366 for (j = 0; j < pred_order; j++)
367 sum += coeffs[j] * s->decoded[channel][i-j-1];
368 s->decoded[channel][i] += sum >> qlevel;
375 static inline int decode_subframe(FLACContext *s, int channel)
377 int type, wasted = 0;
380 s->curr_bps = s->bps;
382 if(s->decorrelation == RIGHT_SIDE)
385 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
389 if (get_bits1(&s->gb))
391 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
394 type = get_bits(&s->gb, 6);
395 // wasted = get_bits1(&s->gb);
399 // while (!get_bits1(&s->gb))
403 // s->curr_bps -= wasted;
406 wasted= 16 - av_log2(show_bits(&s->gb, 17));
407 skip_bits(&s->gb, wasted+1);
408 s->curr_bps -= wasted;
410 if (get_bits1(&s->gb))
413 while (!get_bits1(&s->gb))
415 s->curr_bps -= wasted;
416 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
419 //FIXME use av_log2 for types
422 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
423 tmp = get_sbits(&s->gb, s->curr_bps);
424 for (i = 0; i < s->blocksize; i++)
425 s->decoded[channel][i] = tmp;
429 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
430 for (i = 0; i < s->blocksize; i++)
431 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
433 else if ((type >= 8) && (type <= 12))
435 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
436 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
441 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
442 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
447 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
454 for (i = 0; i < s->blocksize; i++)
455 s->decoded[channel][i] <<= wasted;
461 static int decode_frame(FLACContext *s, int alloc_data_size)
463 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
464 int decorrelation, bps, blocksize, samplerate;
466 blocksize_code = get_bits(&s->gb, 4);
468 sample_rate_code = get_bits(&s->gb, 4);
470 assignment = get_bits(&s->gb, 4); /* channel assignment */
471 if (assignment < 8 && s->channels == assignment+1)
472 decorrelation = INDEPENDENT;
473 else if (assignment >=8 && assignment < 11 && s->channels == 2)
474 decorrelation = LEFT_SIDE + assignment - 8;
477 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
481 sample_size_code = get_bits(&s->gb, 3);
482 if(sample_size_code == 0)
484 else if((sample_size_code != 3) && (sample_size_code != 7))
485 bps = sample_size_table[sample_size_code];
488 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
492 if (get_bits1(&s->gb))
494 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
498 if(get_utf8(&s->gb) < 0){
499 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
503 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
504 (s->min_blocksize != s->max_blocksize)){
509 if (blocksize_code == 0)
510 blocksize = s->min_blocksize;
511 else if (blocksize_code == 6)
512 blocksize = get_bits(&s->gb, 8)+1;
513 else if (blocksize_code == 7)
514 blocksize = get_bits(&s->gb, 16)+1;
516 blocksize = blocksize_table[blocksize_code];
518 if(blocksize > s->max_blocksize){
519 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
523 if(blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
526 if (sample_rate_code == 0){
527 samplerate= s->samplerate;
528 }else if ((sample_rate_code > 3) && (sample_rate_code < 12))
529 samplerate = sample_rate_table[sample_rate_code];
530 else if (sample_rate_code == 12)
531 samplerate = get_bits(&s->gb, 8) * 1000;
532 else if (sample_rate_code == 13)
533 samplerate = get_bits(&s->gb, 16);
534 else if (sample_rate_code == 14)
535 samplerate = get_bits(&s->gb, 16) * 10;
537 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
541 skip_bits(&s->gb, 8);
542 crc8= av_crc(av_crc07, 0, s->gb.buffer, get_bits_count(&s->gb)/8);
544 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
548 s->blocksize = blocksize;
549 s->samplerate = samplerate;
551 s->decorrelation= decorrelation;
556 for (i = 0; i < s->channels; i++)
558 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
559 if (decode_subframe(s, i) < 0)
563 align_get_bits(&s->gb);
566 skip_bits(&s->gb, 16); /* data crc */
571 static inline int16_t shift_to_16_bits(int32_t data, int bps)
575 } else if (bps == 20) {
582 static int flac_decode_frame(AVCodecContext *avctx,
583 void *data, int *data_size,
584 uint8_t *buf, int buf_size)
586 FLACContext *s = avctx->priv_data;
587 int tmp = 0, i, j = 0, input_buf_size = 0;
588 int16_t *samples = data;
589 int alloc_data_size= *data_size;
593 if(s->max_framesize == 0){
594 s->max_framesize= 65536; // should hopefully be enough for the first header
595 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
598 if(1 && s->max_framesize){//FIXME truncated
599 buf_size= FFMAX(FFMIN(buf_size, s->max_framesize - s->bitstream_size), 0);
600 input_buf_size= buf_size;
602 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
603 // printf("memmove\n");
604 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
605 s->bitstream_index=0;
607 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
608 buf= &s->bitstream[s->bitstream_index];
609 buf_size += s->bitstream_size;
610 s->bitstream_size= buf_size;
612 if(buf_size < s->max_framesize){
613 // printf("wanna more data ...\n");
614 return input_buf_size;
618 init_get_bits(&s->gb, buf, buf_size*8);
620 if (!metadata_parse(s))
622 tmp = show_bits(&s->gb, 16);
624 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
625 while(get_bits_count(&s->gb)/8+2 < buf_size && show_bits(&s->gb, 16) != 0xFFF8)
626 skip_bits(&s->gb, 8);
627 goto end; // we may not have enough bits left to decode a frame, so try next time
629 skip_bits(&s->gb, 16);
630 if (decode_frame(s, alloc_data_size) < 0){
631 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
633 s->bitstream_index=0;
640 /* fix the channel order here */
641 if (s->order == MID_SIDE)
643 short *left = samples;
644 short *right = samples + s->blocksize;
645 for (i = 0; i < s->blocksize; i += 2)
647 uint32_t x = s->decoded[0][i];
648 uint32_t y = s->decoded[0][i+1];
650 right[i] = x - (y / 2);
651 left[i] = right[i] + y;
653 *data_size = 2 * s->blocksize;
657 for (i = 0; i < s->channels; i++)
662 for (j = 0; j < s->blocksize; j++)
663 samples[(s->blocksize*i)+j] = s->decoded[i][j];
668 for (j = 0; j < s->blocksize; j++)
669 samples[(s->blocksize*i)+j] = s->decoded[0][j];
671 for (j = 0; j < s->blocksize; j++)
672 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
675 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
677 *data_size += s->blocksize;
681 #define DECORRELATE(left, right)\
682 assert(s->channels == 2);\
683 for (i = 0; i < s->blocksize; i++)\
685 int a= s->decoded[0][i];\
686 int b= s->decoded[1][i];\
687 *(samples++) = (left ) >> (16 - s->bps);\
688 *(samples++) = (right) >> (16 - s->bps);\
692 switch(s->decorrelation)
695 for (j = 0; j < s->blocksize; j++)
697 for (i = 0; i < s->channels; i++)
698 *(samples++) = shift_to_16_bits(s->decoded[i][j], s->bps);
706 DECORRELATE( (a-=b>>1) + b, a)
710 *data_size = (int8_t *)samples - (int8_t *)data;
711 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
713 // s->last_blocksize = s->blocksize;
715 i= (get_bits_count(&s->gb)+7)/8;;
717 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
719 s->bitstream_index=0;
723 if(s->bitstream_size){
724 s->bitstream_index += i;
725 s->bitstream_size -= i;
726 return input_buf_size;
731 static int flac_decode_close(AVCodecContext *avctx)
733 FLACContext *s = avctx->priv_data;
736 for (i = 0; i < s->channels; i++)
738 av_freep(&s->decoded[i]);
740 av_freep(&s->bitstream);
745 static void flac_flush(AVCodecContext *avctx){
746 FLACContext *s = avctx->priv_data;
749 s->bitstream_index= 0;
752 AVCodec flac_decoder = {