2 * FLAC (Free Lossless Audio Codec) decoder
3 * Copyright (c) 2003 Alex Beregszaszi
5 * This library is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2 of the License, or (at your option) any later version.
10 * This library is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with this library; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 * FLAC (Free Lossless Audio Codec) decoder
23 * @author Alex Beregszaszi
25 * For more information on the FLAC format, visit:
26 * http://flac.sourceforge.net/
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.
37 #include "bitstream.h"
44 #define MAX_CHANNELS 8
45 #define MAX_BLOCKSIZE 65535
46 #define FLAC_STREAMINFO_SIZE 34
48 enum decorrelation_type {
55 typedef struct FLACContext {
56 AVCodecContext *avctx;
59 int min_blocksize, max_blocksize;
60 int min_framesize, max_framesize;
61 int samplerate, channels;
62 int blocksize/*, last_blocksize*/;
64 enum decorrelation_type decorrelation;
66 int32_t *decoded[MAX_CHANNELS];
70 unsigned int allocated_bitstream_size;
73 #define METADATA_TYPE_STREAMINFO 0
75 static int sample_rate_table[] =
77 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000,
80 static int sample_size_table[] =
81 { 0, 8, 12, 0, 16, 20, 24, 0 };
83 static int blocksize_table[] = {
84 0, 192, 576<<0, 576<<1, 576<<2, 576<<3, 0, 0,
85 256<<0, 256<<1, 256<<2, 256<<3, 256<<4, 256<<5, 256<<6, 256<<7
88 static int64_t get_utf8(GetBitContext *gb)
97 else if(ones==1) return -1;
100 val= get_bits(gb, 7-ones);
102 const int tmp = get_bits(gb, 8);
113 static int skip_utf8(GetBitContext *gb)
120 if (ones==0) bytes=0;
121 else if(ones==1) return -1;
122 else bytes= ones - 1;
124 skip_bits(gb, 7-ones);
126 const int tmp = get_bits(gb, 8);
135 static void metadata_streaminfo(FLACContext *s);
136 static void dump_headers(FLACContext *s);
138 static int flac_decode_init(AVCodecContext * avctx)
140 FLACContext *s = avctx->priv_data;
143 /* initialize based on the demuxer-supplied streamdata header */
144 if (avctx->extradata_size == FLAC_STREAMINFO_SIZE) {
145 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size*8);
146 metadata_streaminfo(s);
153 static void dump_headers(FLACContext *s)
155 av_log(s->avctx, AV_LOG_DEBUG, " Blocksize: %d .. %d (%d)\n", s->min_blocksize, s->max_blocksize, s->blocksize);
156 av_log(s->avctx, AV_LOG_DEBUG, " Framesize: %d .. %d\n", s->min_framesize, s->max_framesize);
157 av_log(s->avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
158 av_log(s->avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
159 av_log(s->avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
162 static void allocate_buffers(FLACContext *s){
165 assert(s->max_blocksize);
167 if(s->max_framesize == 0 && s->max_blocksize){
168 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8; //FIXME header overhead
171 for (i = 0; i < s->channels; i++)
173 s->decoded[i] = av_realloc(s->decoded[i], sizeof(int32_t)*s->max_blocksize);
176 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
179 static void metadata_streaminfo(FLACContext *s)
181 /* mandatory streaminfo */
182 s->min_blocksize = get_bits(&s->gb, 16);
183 s->max_blocksize = get_bits(&s->gb, 16);
185 s->min_framesize = get_bits_long(&s->gb, 24);
186 s->max_framesize = get_bits_long(&s->gb, 24);
188 s->samplerate = get_bits_long(&s->gb, 20);
189 s->channels = get_bits(&s->gb, 3) + 1;
190 s->bps = get_bits(&s->gb, 5) + 1;
192 s->avctx->channels = s->channels;
193 s->avctx->sample_rate = s->samplerate;
195 skip_bits(&s->gb, 36); /* total num of samples */
197 skip_bits(&s->gb, 64); /* md5 sum */
198 skip_bits(&s->gb, 64); /* md5 sum */
203 static int decode_residuals(FLACContext *s, int channel, int pred_order)
205 int i, tmp, partition, method_type, rice_order;
206 int sample = 0, samples;
208 method_type = get_bits(&s->gb, 2);
209 if (method_type != 0){
210 av_log(s->avctx, AV_LOG_DEBUG, "illegal residual coding method %d\n", method_type);
214 rice_order = get_bits(&s->gb, 4);
216 samples= s->blocksize >> rice_order;
220 for (partition = 0; partition < (1 << rice_order); partition++)
222 tmp = get_bits(&s->gb, 4);
225 av_log(s->avctx, AV_LOG_DEBUG, "fixed len partition\n");
226 tmp = get_bits(&s->gb, 5);
227 for (; i < samples; i++, sample++)
228 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
232 // av_log(s->avctx, AV_LOG_DEBUG, "rice coded partition k=%d\n", tmp);
233 for (; i < samples; i++, sample++){
234 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
240 // av_log(s->avctx, AV_LOG_DEBUG, "partitions: %d, samples: %d\n", 1 << rice_order, sample);
245 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
249 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME FIXED\n");
251 /* warm up samples */
252 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
254 for (i = 0; i < pred_order; i++)
256 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
257 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
260 if (decode_residuals(s, channel, pred_order) < 0)
268 for (i = pred_order; i < s->blocksize; i++)
269 s->decoded[channel][i] += s->decoded[channel][i-1];
272 for (i = pred_order; i < s->blocksize; i++)
273 s->decoded[channel][i] += 2*s->decoded[channel][i-1]
274 - s->decoded[channel][i-2];
277 for (i = pred_order; i < s->blocksize; i++)
278 s->decoded[channel][i] += 3*s->decoded[channel][i-1]
279 - 3*s->decoded[channel][i-2]
280 + s->decoded[channel][i-3];
283 for (i = pred_order; i < s->blocksize; i++)
284 s->decoded[channel][i] += 4*s->decoded[channel][i-1]
285 - 6*s->decoded[channel][i-2]
286 + 4*s->decoded[channel][i-3]
287 - s->decoded[channel][i-4];
290 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
297 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
300 int coeff_prec, qlevel;
301 int coeffs[pred_order];
303 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
305 /* warm up samples */
306 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
308 for (i = 0; i < pred_order; i++)
310 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
311 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
314 coeff_prec = get_bits(&s->gb, 4) + 1;
315 if (coeff_prec == 16)
317 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
320 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
321 qlevel = get_sbits(&s->gb, 5);
322 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
324 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
328 for (i = 0; i < pred_order; i++)
330 coeffs[i] = get_sbits(&s->gb, coeff_prec);
331 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
334 if (decode_residuals(s, channel, pred_order) < 0)
339 for (i = pred_order; i < s->blocksize; i++)
342 for (j = 0; j < pred_order; j++)
343 sum += (int64_t)coeffs[j] * s->decoded[channel][i-j-1];
344 s->decoded[channel][i] += sum >> qlevel;
348 for (i = pred_order; i < s->blocksize; i++)
351 for (j = 0; j < pred_order; j++)
352 sum += coeffs[j] * s->decoded[channel][i-j-1];
353 s->decoded[channel][i] += sum >> qlevel;
360 static inline int decode_subframe(FLACContext *s, int channel)
362 int type, wasted = 0;
365 s->curr_bps = s->bps;
367 if(s->decorrelation == RIGHT_SIDE)
370 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
374 if (get_bits1(&s->gb))
376 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
379 type = get_bits(&s->gb, 6);
380 // wasted = get_bits1(&s->gb);
384 // while (!get_bits1(&s->gb))
388 // s->curr_bps -= wasted;
391 wasted= 16 - av_log2(show_bits(&s->gb, 17));
392 skip_bits(&s->gb, wasted+1);
393 s->curr_bps -= wasted;
395 if (get_bits1(&s->gb))
398 while (!get_bits1(&s->gb))
400 s->curr_bps -= wasted;
401 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
404 //FIXME use av_log2 for types
407 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
408 tmp = get_sbits(&s->gb, s->curr_bps);
409 for (i = 0; i < s->blocksize; i++)
410 s->decoded[channel][i] = tmp;
414 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
415 for (i = 0; i < s->blocksize; i++)
416 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
418 else if ((type >= 8) && (type <= 12))
420 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
421 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
426 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
427 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
432 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
439 for (i = 0; i < s->blocksize; i++)
440 s->decoded[channel][i] <<= wasted;
446 static int decode_frame(FLACContext *s)
448 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
449 int decorrelation, bps, blocksize, samplerate;
451 blocksize_code = get_bits(&s->gb, 4);
453 sample_rate_code = get_bits(&s->gb, 4);
455 assignment = get_bits(&s->gb, 4); /* channel assignment */
456 if (assignment < 8 && s->channels == assignment+1)
457 decorrelation = INDEPENDENT;
458 else if (assignment >=8 && assignment < 11 && s->channels == 2)
459 decorrelation = LEFT_SIDE + assignment - 8;
462 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
466 sample_size_code = get_bits(&s->gb, 3);
467 if(sample_size_code == 0)
469 else if((sample_size_code != 3) && (sample_size_code != 7))
470 bps = sample_size_table[sample_size_code];
473 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
477 if (get_bits1(&s->gb))
479 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
483 if(get_utf8(&s->gb) < 0){
484 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
488 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
489 (s->min_blocksize != s->max_blocksize)){
494 if (blocksize_code == 0)
495 blocksize = s->min_blocksize;
496 else if (blocksize_code == 6)
497 blocksize = get_bits(&s->gb, 8)+1;
498 else if (blocksize_code == 7)
499 blocksize = get_bits(&s->gb, 16)+1;
501 blocksize = blocksize_table[blocksize_code];
503 if(blocksize > s->max_blocksize){
504 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
508 if (sample_rate_code == 0){
509 samplerate= s->samplerate;
510 }else if ((sample_rate_code > 3) && (sample_rate_code < 12))
511 samplerate = sample_rate_table[sample_rate_code];
512 else if (sample_rate_code == 12)
513 samplerate = get_bits(&s->gb, 8) * 1000;
514 else if (sample_rate_code == 13)
515 samplerate = get_bits(&s->gb, 16);
516 else if (sample_rate_code == 14)
517 samplerate = get_bits(&s->gb, 16) * 10;
519 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
523 skip_bits(&s->gb, 8);
524 crc8= av_crc(av_crc07, 0, s->gb.buffer, get_bits_count(&s->gb)/8);
526 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
530 s->blocksize = blocksize;
531 s->samplerate = samplerate;
533 s->decorrelation= decorrelation;
538 for (i = 0; i < s->channels; i++)
540 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
541 if (decode_subframe(s, i) < 0)
545 align_get_bits(&s->gb);
548 skip_bits(&s->gb, 16); /* data crc */
553 static inline int16_t shift_to_16_bits(int32_t data, int bps)
557 } else if (bps == 20) {
564 static int flac_decode_frame(AVCodecContext *avctx,
565 void *data, int *data_size,
566 uint8_t *buf, int buf_size)
568 FLACContext *s = avctx->priv_data;
569 int metadata_last, metadata_type, metadata_size;
570 int tmp = 0, i, j = 0, input_buf_size = 0;
571 int16_t *samples = data;
573 if(s->max_framesize == 0){
574 s->max_framesize= 65536; // should hopefully be enough for the first header
575 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
578 if(1 && s->max_framesize){//FIXME truncated
579 buf_size= FFMAX(FFMIN(buf_size, s->max_framesize - s->bitstream_size), 0);
580 input_buf_size= buf_size;
582 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
583 // printf("memmove\n");
584 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
585 s->bitstream_index=0;
587 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
588 buf= &s->bitstream[s->bitstream_index];
589 buf_size += s->bitstream_size;
590 s->bitstream_size= buf_size;
592 if(buf_size < s->max_framesize){
593 // printf("wanna more data ...\n");
594 return input_buf_size;
598 init_get_bits(&s->gb, buf, buf_size*8);
600 /* fLaC signature (be) */
601 if (show_bits_long(&s->gb, 32) == bswap_32(ff_get_fourcc("fLaC")))
603 skip_bits(&s->gb, 32);
605 av_log(s->avctx, AV_LOG_DEBUG, "STREAM HEADER\n");
607 metadata_last = get_bits(&s->gb, 1);
608 metadata_type = get_bits(&s->gb, 7);
609 metadata_size = get_bits_long(&s->gb, 24);
611 av_log(s->avctx, AV_LOG_DEBUG, " metadata block: flag = %d, type = %d, size = %d\n",
612 metadata_last, metadata_type,
615 switch(metadata_type)
617 case METADATA_TYPE_STREAMINFO:{
618 metadata_streaminfo(s);
620 /* Buffer might have been reallocated, reinit bitreader */
621 if(buf != &s->bitstream[s->bitstream_index])
623 int bits_count = get_bits_count(&s->gb);
624 buf= &s->bitstream[s->bitstream_index];
625 init_get_bits(&s->gb, buf, buf_size*8);
626 skip_bits(&s->gb, bits_count);
632 for(i=0; i<metadata_size; i++)
633 skip_bits(&s->gb, 8);
636 } while(!metadata_last);
641 tmp = show_bits(&s->gb, 16);
643 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
644 while(get_bits_count(&s->gb)/8+2 < buf_size && show_bits(&s->gb, 16) != 0xFFF8)
645 skip_bits(&s->gb, 8);
646 goto end; // we may not have enough bits left to decode a frame, so try next time
648 skip_bits(&s->gb, 16);
649 if (decode_frame(s) < 0){
650 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
652 s->bitstream_index=0;
659 /* fix the channel order here */
660 if (s->order == MID_SIDE)
662 short *left = samples;
663 short *right = samples + s->blocksize;
664 for (i = 0; i < s->blocksize; i += 2)
666 uint32_t x = s->decoded[0][i];
667 uint32_t y = s->decoded[0][i+1];
669 right[i] = x - (y / 2);
670 left[i] = right[i] + y;
672 *data_size = 2 * s->blocksize;
676 for (i = 0; i < s->channels; i++)
681 for (j = 0; j < s->blocksize; j++)
682 samples[(s->blocksize*i)+j] = s->decoded[i][j];
687 for (j = 0; j < s->blocksize; j++)
688 samples[(s->blocksize*i)+j] = s->decoded[0][j];
690 for (j = 0; j < s->blocksize; j++)
691 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
694 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
696 *data_size += s->blocksize;
700 switch(s->decorrelation)
703 for (j = 0; j < s->blocksize; j++)
705 for (i = 0; i < s->channels; i++)
706 *(samples++) = shift_to_16_bits(s->decoded[i][j], s->bps);
710 assert(s->channels == 2);
711 for (i = 0; i < s->blocksize; i++)
713 *(samples++) = shift_to_16_bits(s->decoded[0][i], s->bps);
714 *(samples++) = shift_to_16_bits(s->decoded[0][i]
715 - s->decoded[1][i], s->bps);
719 assert(s->channels == 2);
720 for (i = 0; i < s->blocksize; i++)
722 *(samples++) = shift_to_16_bits(s->decoded[0][i]
723 + s->decoded[1][i], s->bps);
724 *(samples++) = shift_to_16_bits(s->decoded[1][i], s->bps);
728 assert(s->channels == 2);
729 for (i = 0; i < s->blocksize; i++)
732 mid = s->decoded[0][i];
733 side = s->decoded[1][i];
735 #if 1 //needs to be checked but IMHO it should be binary identical
737 *(samples++) = shift_to_16_bits(mid + side, s->bps);
738 *(samples++) = shift_to_16_bits(mid, s->bps);
744 *(samples++) = (mid + side) >> 1;
745 *(samples++) = (mid - side) >> 1;
752 *data_size = (int8_t *)samples - (int8_t *)data;
753 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
755 // s->last_blocksize = s->blocksize;
757 i= (get_bits_count(&s->gb)+7)/8;;
759 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
761 s->bitstream_index=0;
765 if(s->bitstream_size){
766 s->bitstream_index += i;
767 s->bitstream_size -= i;
768 return input_buf_size;
773 static int flac_decode_close(AVCodecContext *avctx)
775 FLACContext *s = avctx->priv_data;
778 for (i = 0; i < s->channels; i++)
780 av_freep(&s->decoded[i]);
782 av_freep(&s->bitstream);
787 static void flac_flush(AVCodecContext *avctx){
788 FLACContext *s = avctx->priv_data;
791 s->bitstream_index= 0;
794 AVCodec flac_decoder = {