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];
73 unsigned int allocated_bitstream_size;
76 #define METADATA_TYPE_STREAMINFO 0
78 static const int sample_rate_table[] =
80 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000,
83 static const int sample_size_table[] =
84 { 0, 8, 12, 0, 16, 20, 24, 0 };
86 static const 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 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, avctx->extradata);
111 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size*8);
119 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
121 av_log(avctx, AV_LOG_DEBUG, " Blocksize: %d .. %d\n", s->min_blocksize, s->max_blocksize);
122 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
123 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
124 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
125 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
128 static void allocate_buffers(FLACContext *s){
131 assert(s->max_blocksize);
133 if(s->max_framesize == 0 && s->max_blocksize){
134 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8; //FIXME header overhead
137 for (i = 0; i < s->channels; i++)
139 s->decoded[i] = av_realloc(s->decoded[i], sizeof(int32_t)*s->max_blocksize);
142 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
145 void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
146 const uint8_t *buffer)
149 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
151 /* mandatory streaminfo */
152 s->min_blocksize = get_bits(&gb, 16);
153 s->max_blocksize = get_bits(&gb, 16);
155 skip_bits(&gb, 24); /* skip min frame size */
156 s->max_framesize = get_bits_long(&gb, 24);
158 s->samplerate = get_bits_long(&gb, 20);
159 s->channels = get_bits(&gb, 3) + 1;
160 s->bps = get_bits(&gb, 5) + 1;
162 avctx->channels = s->channels;
163 avctx->sample_rate = s->samplerate;
165 skip_bits(&gb, 36); /* total num of samples */
167 skip_bits(&gb, 64); /* md5 sum */
168 skip_bits(&gb, 64); /* md5 sum */
170 dump_headers(avctx, s);
174 * Parse a list of metadata blocks. This list of blocks must begin with
176 * @param s the flac decoding context containing the gb bit reader used to
178 * @return 1 if some metadata was read, 0 if no fLaC marker was found
180 static int metadata_parse(FLACContext *s)
182 int i, metadata_last, metadata_type, metadata_size, streaminfo_updated=0;
184 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
185 skip_bits(&s->gb, 32);
187 av_log(s->avctx, AV_LOG_DEBUG, "STREAM HEADER\n");
189 metadata_last = get_bits1(&s->gb);
190 metadata_type = get_bits(&s->gb, 7);
191 metadata_size = get_bits_long(&s->gb, 24);
193 av_log(s->avctx, AV_LOG_DEBUG,
194 " metadata block: flag = %d, type = %d, size = %d\n",
195 metadata_last, metadata_type, metadata_size);
197 switch (metadata_type) {
198 case METADATA_TYPE_STREAMINFO:
199 ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, s->gb.buffer+get_bits_count(&s->gb)/8);
200 streaminfo_updated = 1;
203 for (i=0; i<metadata_size; i++)
204 skip_bits(&s->gb, 8);
207 } while (!metadata_last);
209 if (streaminfo_updated)
216 static int decode_residuals(FLACContext *s, int channel, int pred_order)
218 int i, tmp, partition, method_type, rice_order;
219 int sample = 0, samples;
221 method_type = get_bits(&s->gb, 2);
222 if (method_type > 1){
223 av_log(s->avctx, AV_LOG_DEBUG, "illegal residual coding method %d\n", method_type);
227 rice_order = get_bits(&s->gb, 4);
229 samples= s->blocksize >> rice_order;
230 if (pred_order > samples) {
231 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n", pred_order, samples);
237 for (partition = 0; partition < (1 << rice_order); partition++)
239 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
240 if (tmp == (method_type == 0 ? 15 : 31))
242 av_log(s->avctx, AV_LOG_DEBUG, "fixed len partition\n");
243 tmp = get_bits(&s->gb, 5);
244 for (; i < samples; i++, sample++)
245 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
249 // av_log(s->avctx, AV_LOG_DEBUG, "rice coded partition k=%d\n", tmp);
250 for (; i < samples; i++, sample++){
251 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
257 // av_log(s->avctx, AV_LOG_DEBUG, "partitions: %d, samples: %d\n", 1 << rice_order, sample);
262 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
264 const int blocksize = s->blocksize;
265 int32_t *decoded = s->decoded[channel];
268 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME FIXED\n");
270 /* warm up samples */
271 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
273 for (i = 0; i < pred_order; i++)
275 decoded[i] = get_sbits(&s->gb, s->curr_bps);
276 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
279 if (decode_residuals(s, channel, pred_order) < 0)
283 a = decoded[pred_order-1];
285 b = a - decoded[pred_order-2];
287 c = b - decoded[pred_order-2] + decoded[pred_order-3];
289 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
296 for (i = pred_order; i < blocksize; i++)
297 decoded[i] = a += decoded[i];
300 for (i = pred_order; i < blocksize; i++)
301 decoded[i] = a += b += decoded[i];
304 for (i = pred_order; i < blocksize; i++)
305 decoded[i] = a += b += c += decoded[i];
308 for (i = pred_order; i < blocksize; i++)
309 decoded[i] = a += b += c += d += decoded[i];
312 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
319 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
322 int coeff_prec, qlevel;
323 int coeffs[pred_order];
324 int32_t *decoded = s->decoded[channel];
326 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
328 /* warm up samples */
329 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
331 for (i = 0; i < pred_order; i++)
333 decoded[i] = get_sbits(&s->gb, s->curr_bps);
334 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, decoded[i]);
337 coeff_prec = get_bits(&s->gb, 4) + 1;
338 if (coeff_prec == 16)
340 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
343 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
344 qlevel = get_sbits(&s->gb, 5);
345 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
347 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
351 for (i = 0; i < pred_order; i++)
353 coeffs[i] = get_sbits(&s->gb, coeff_prec);
354 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
357 if (decode_residuals(s, channel, pred_order) < 0)
362 for (i = pred_order; i < s->blocksize; i++)
365 for (j = 0; j < pred_order; j++)
366 sum += (int64_t)coeffs[j] * decoded[i-j-1];
367 decoded[i] += sum >> qlevel;
370 for (i = pred_order; i < s->blocksize-1; i += 2)
373 int d = decoded[i-pred_order];
375 for (j = pred_order-1; j > 0; j--)
384 d = decoded[i] += s0 >> qlevel;
386 decoded[i+1] += s1 >> qlevel;
388 if (i < s->blocksize)
391 for (j = 0; j < pred_order; j++)
392 sum += coeffs[j] * decoded[i-j-1];
393 decoded[i] += sum >> qlevel;
400 static inline int decode_subframe(FLACContext *s, int channel)
402 int type, wasted = 0;
405 s->curr_bps = s->bps;
407 if(s->decorrelation == RIGHT_SIDE)
410 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
414 if (get_bits1(&s->gb))
416 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
419 type = get_bits(&s->gb, 6);
420 // wasted = get_bits1(&s->gb);
424 // while (!get_bits1(&s->gb))
428 // s->curr_bps -= wasted;
431 wasted= 16 - av_log2(show_bits(&s->gb, 17));
432 skip_bits(&s->gb, wasted+1);
433 s->curr_bps -= wasted;
435 if (get_bits1(&s->gb))
438 while (!get_bits1(&s->gb))
440 s->curr_bps -= wasted;
441 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
444 //FIXME use av_log2 for types
447 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
448 tmp = get_sbits(&s->gb, s->curr_bps);
449 for (i = 0; i < s->blocksize; i++)
450 s->decoded[channel][i] = tmp;
454 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
455 for (i = 0; i < s->blocksize; i++)
456 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
458 else if ((type >= 8) && (type <= 12))
460 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
461 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
466 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
467 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
472 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
479 for (i = 0; i < s->blocksize; i++)
480 s->decoded[channel][i] <<= wasted;
486 static int decode_frame(FLACContext *s, int alloc_data_size)
488 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
489 int decorrelation, bps, blocksize, samplerate;
491 blocksize_code = get_bits(&s->gb, 4);
493 sample_rate_code = get_bits(&s->gb, 4);
495 assignment = get_bits(&s->gb, 4); /* channel assignment */
496 if (assignment < 8 && s->channels == assignment+1)
497 decorrelation = INDEPENDENT;
498 else if (assignment >=8 && assignment < 11 && s->channels == 2)
499 decorrelation = LEFT_SIDE + assignment - 8;
502 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
506 sample_size_code = get_bits(&s->gb, 3);
507 if(sample_size_code == 0)
509 else if((sample_size_code != 3) && (sample_size_code != 7))
510 bps = sample_size_table[sample_size_code];
513 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
517 if (get_bits1(&s->gb))
519 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
523 if(get_utf8(&s->gb) < 0){
524 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
528 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
529 (s->min_blocksize != s->max_blocksize)){
534 if (blocksize_code == 0)
535 blocksize = s->min_blocksize;
536 else if (blocksize_code == 6)
537 blocksize = get_bits(&s->gb, 8)+1;
538 else if (blocksize_code == 7)
539 blocksize = get_bits(&s->gb, 16)+1;
541 blocksize = blocksize_table[blocksize_code];
543 if(blocksize > s->max_blocksize){
544 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
548 if(blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
551 if (sample_rate_code == 0){
552 samplerate= s->samplerate;
553 }else if ((sample_rate_code > 3) && (sample_rate_code < 12))
554 samplerate = sample_rate_table[sample_rate_code];
555 else if (sample_rate_code == 12)
556 samplerate = get_bits(&s->gb, 8) * 1000;
557 else if (sample_rate_code == 13)
558 samplerate = get_bits(&s->gb, 16);
559 else if (sample_rate_code == 14)
560 samplerate = get_bits(&s->gb, 16) * 10;
562 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
566 skip_bits(&s->gb, 8);
567 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
568 s->gb.buffer, get_bits_count(&s->gb)/8);
570 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
574 s->blocksize = blocksize;
575 s->samplerate = samplerate;
577 s->decorrelation= decorrelation;
579 // dump_headers(s->avctx, (FLACStreaminfo *)s);
582 for (i = 0; i < s->channels; i++)
584 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
585 if (decode_subframe(s, i) < 0)
589 align_get_bits(&s->gb);
592 skip_bits(&s->gb, 16); /* data crc */
597 static int flac_decode_frame(AVCodecContext *avctx,
598 void *data, int *data_size,
599 const uint8_t *buf, int buf_size)
601 FLACContext *s = avctx->priv_data;
602 int tmp = 0, i, j = 0, input_buf_size = 0;
603 int16_t *samples = data;
604 int alloc_data_size= *data_size;
608 if(s->max_framesize == 0){
609 s->max_framesize= 65536; // should hopefully be enough for the first header
610 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
613 if(1 && s->max_framesize){//FIXME truncated
614 buf_size= FFMAX(FFMIN(buf_size, s->max_framesize - s->bitstream_size), 0);
615 input_buf_size= buf_size;
617 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
618 // printf("memmove\n");
619 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
620 s->bitstream_index=0;
622 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
623 buf= &s->bitstream[s->bitstream_index];
624 buf_size += s->bitstream_size;
625 s->bitstream_size= buf_size;
627 if(buf_size < s->max_framesize && input_buf_size){
628 // printf("wanna more data ...\n");
629 return input_buf_size;
633 init_get_bits(&s->gb, buf, buf_size*8);
635 if (!metadata_parse(s))
637 tmp = show_bits(&s->gb, 16);
638 if((tmp & 0xFFFE) != 0xFFF8){
639 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
640 while(get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
641 skip_bits(&s->gb, 8);
642 goto end; // we may not have enough bits left to decode a frame, so try next time
644 skip_bits(&s->gb, 16);
645 if (decode_frame(s, alloc_data_size) < 0){
646 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
648 s->bitstream_index=0;
655 /* fix the channel order here */
656 if (s->order == MID_SIDE)
658 short *left = samples;
659 short *right = samples + s->blocksize;
660 for (i = 0; i < s->blocksize; i += 2)
662 uint32_t x = s->decoded[0][i];
663 uint32_t y = s->decoded[0][i+1];
665 right[i] = x - (y / 2);
666 left[i] = right[i] + y;
668 *data_size = 2 * s->blocksize;
672 for (i = 0; i < s->channels; i++)
677 for (j = 0; j < s->blocksize; j++)
678 samples[(s->blocksize*i)+j] = s->decoded[i][j];
683 for (j = 0; j < s->blocksize; j++)
684 samples[(s->blocksize*i)+j] = s->decoded[0][j];
686 for (j = 0; j < s->blocksize; j++)
687 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
690 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
692 *data_size += s->blocksize;
696 #define DECORRELATE(left, right)\
697 assert(s->channels == 2);\
698 for (i = 0; i < s->blocksize; i++)\
700 int a= s->decoded[0][i];\
701 int b= s->decoded[1][i];\
702 *samples++ = ((left) << (24 - s->bps)) >> 8;\
703 *samples++ = ((right) << (24 - s->bps)) >> 8;\
707 switch(s->decorrelation)
710 for (j = 0; j < s->blocksize; j++)
712 for (i = 0; i < s->channels; i++)
713 *samples++ = (s->decoded[i][j] << (24 - s->bps)) >> 8;
721 DECORRELATE( (a-=b>>1) + b, a)
725 *data_size = (int8_t *)samples - (int8_t *)data;
726 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
728 // s->last_blocksize = s->blocksize;
730 i= (get_bits_count(&s->gb)+7)/8;
732 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
734 s->bitstream_index=0;
738 if(s->bitstream_size){
739 s->bitstream_index += i;
740 s->bitstream_size -= i;
741 return input_buf_size;
746 static av_cold int flac_decode_close(AVCodecContext *avctx)
748 FLACContext *s = avctx->priv_data;
751 for (i = 0; i < s->channels; i++)
753 av_freep(&s->decoded[i]);
755 av_freep(&s->bitstream);
760 static void flac_flush(AVCodecContext *avctx){
761 FLACContext *s = avctx->priv_data;
764 s->bitstream_index= 0;
767 AVCodec flac_decoder = {
778 .long_name= NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),