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 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 if(s->allocated_bitstream_size < s->max_framesize)
143 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
146 void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
147 const uint8_t *buffer)
150 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
152 /* mandatory streaminfo */
153 s->min_blocksize = get_bits(&gb, 16);
154 s->max_blocksize = get_bits(&gb, 16);
156 skip_bits(&gb, 24); /* skip min frame size */
157 s->max_framesize = get_bits_long(&gb, 24);
159 s->samplerate = get_bits_long(&gb, 20);
160 s->channels = get_bits(&gb, 3) + 1;
161 s->bps = get_bits(&gb, 5) + 1;
163 avctx->channels = s->channels;
164 avctx->sample_rate = s->samplerate;
166 skip_bits(&gb, 36); /* total num of samples */
168 skip_bits(&gb, 64); /* md5 sum */
169 skip_bits(&gb, 64); /* md5 sum */
171 dump_headers(avctx, s);
175 * Parse a list of metadata blocks. This list of blocks must begin with
177 * @param s the flac decoding context containing the gb bit reader used to
179 * @return 1 if some metadata was read, 0 if no fLaC marker was found
181 static int metadata_parse(FLACContext *s)
183 int i, metadata_last, metadata_type, metadata_size, streaminfo_updated=0;
184 int initial_pos= get_bits_count(&s->gb);
186 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
187 skip_bits(&s->gb, 32);
189 av_log(s->avctx, AV_LOG_DEBUG, "STREAM HEADER\n");
191 metadata_last = get_bits1(&s->gb);
192 metadata_type = get_bits(&s->gb, 7);
193 metadata_size = get_bits_long(&s->gb, 24);
195 if(get_bits_count(&s->gb) + 8*metadata_size > s->gb.size_in_bits){
196 skip_bits_long(&s->gb, initial_pos - get_bits_count(&s->gb));
200 av_log(s->avctx, AV_LOG_DEBUG,
201 " metadata block: flag = %d, type = %d, size = %d\n",
202 metadata_last, metadata_type, metadata_size);
204 switch (metadata_type) {
205 case METADATA_TYPE_STREAMINFO:
206 ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, s->gb.buffer+get_bits_count(&s->gb)/8);
207 streaminfo_updated = 1;
210 for (i=0; i<metadata_size; i++)
211 skip_bits(&s->gb, 8);
214 } while (!metadata_last);
216 if (streaminfo_updated)
223 static int decode_residuals(FLACContext *s, int channel, int pred_order)
225 int i, tmp, partition, method_type, rice_order;
226 int sample = 0, samples;
228 method_type = get_bits(&s->gb, 2);
229 if (method_type > 1){
230 av_log(s->avctx, AV_LOG_DEBUG, "illegal residual coding method %d\n", method_type);
234 rice_order = get_bits(&s->gb, 4);
236 samples= s->blocksize >> rice_order;
237 if (pred_order > samples) {
238 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n", pred_order, samples);
244 for (partition = 0; partition < (1 << rice_order); partition++)
246 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
247 if (tmp == (method_type == 0 ? 15 : 31))
249 av_log(s->avctx, AV_LOG_DEBUG, "fixed len partition\n");
250 tmp = get_bits(&s->gb, 5);
251 for (; i < samples; i++, sample++)
252 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
256 // av_log(s->avctx, AV_LOG_DEBUG, "rice coded partition k=%d\n", tmp);
257 for (; i < samples; i++, sample++){
258 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
264 // av_log(s->avctx, AV_LOG_DEBUG, "partitions: %d, samples: %d\n", 1 << rice_order, sample);
269 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
271 const int blocksize = s->blocksize;
272 int32_t *decoded = s->decoded[channel];
275 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME FIXED\n");
277 /* warm up samples */
278 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
280 for (i = 0; i < pred_order; i++)
282 decoded[i] = get_sbits(&s->gb, s->curr_bps);
283 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, s->decoded[channel][i]);
286 if (decode_residuals(s, channel, pred_order) < 0)
290 a = decoded[pred_order-1];
292 b = a - decoded[pred_order-2];
294 c = b - decoded[pred_order-2] + decoded[pred_order-3];
296 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
303 for (i = pred_order; i < blocksize; i++)
304 decoded[i] = a += decoded[i];
307 for (i = pred_order; i < blocksize; i++)
308 decoded[i] = a += b += decoded[i];
311 for (i = pred_order; i < blocksize; i++)
312 decoded[i] = a += b += c += decoded[i];
315 for (i = pred_order; i < blocksize; i++)
316 decoded[i] = a += b += c += d += decoded[i];
319 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
326 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
329 int coeff_prec, qlevel;
330 int coeffs[pred_order];
331 int32_t *decoded = s->decoded[channel];
333 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
335 /* warm up samples */
336 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
338 for (i = 0; i < pred_order; i++)
340 decoded[i] = get_sbits(&s->gb, s->curr_bps);
341 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, decoded[i]);
344 coeff_prec = get_bits(&s->gb, 4) + 1;
345 if (coeff_prec == 16)
347 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
350 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
351 qlevel = get_sbits(&s->gb, 5);
352 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
354 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
358 for (i = 0; i < pred_order; i++)
360 coeffs[i] = get_sbits(&s->gb, coeff_prec);
361 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
364 if (decode_residuals(s, channel, pred_order) < 0)
369 for (i = pred_order; i < s->blocksize; i++)
372 for (j = 0; j < pred_order; j++)
373 sum += (int64_t)coeffs[j] * decoded[i-j-1];
374 decoded[i] += sum >> qlevel;
377 for (i = pred_order; i < s->blocksize-1; i += 2)
380 int d = decoded[i-pred_order];
382 for (j = pred_order-1; j > 0; j--)
391 d = decoded[i] += s0 >> qlevel;
393 decoded[i+1] += s1 >> qlevel;
395 if (i < s->blocksize)
398 for (j = 0; j < pred_order; j++)
399 sum += coeffs[j] * decoded[i-j-1];
400 decoded[i] += sum >> qlevel;
407 static inline int decode_subframe(FLACContext *s, int channel)
409 int type, wasted = 0;
412 s->curr_bps = s->bps;
414 if(s->decorrelation == RIGHT_SIDE)
417 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
421 if (get_bits1(&s->gb))
423 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
426 type = get_bits(&s->gb, 6);
427 // wasted = get_bits1(&s->gb);
431 // while (!get_bits1(&s->gb))
435 // s->curr_bps -= wasted;
438 wasted= 16 - av_log2(show_bits(&s->gb, 17));
439 skip_bits(&s->gb, wasted+1);
440 s->curr_bps -= wasted;
442 if (get_bits1(&s->gb))
445 while (!get_bits1(&s->gb))
447 s->curr_bps -= wasted;
448 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
451 //FIXME use av_log2 for types
454 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
455 tmp = get_sbits(&s->gb, s->curr_bps);
456 for (i = 0; i < s->blocksize; i++)
457 s->decoded[channel][i] = tmp;
461 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
462 for (i = 0; i < s->blocksize; i++)
463 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
465 else if ((type >= 8) && (type <= 12))
467 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
468 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
473 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
474 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
479 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
486 for (i = 0; i < s->blocksize; i++)
487 s->decoded[channel][i] <<= wasted;
493 static int decode_frame(FLACContext *s, int alloc_data_size)
495 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
496 int decorrelation, bps, blocksize, samplerate;
498 blocksize_code = get_bits(&s->gb, 4);
500 sample_rate_code = get_bits(&s->gb, 4);
502 assignment = get_bits(&s->gb, 4); /* channel assignment */
503 if (assignment < 8 && s->channels == assignment+1)
504 decorrelation = INDEPENDENT;
505 else if (assignment >=8 && assignment < 11 && s->channels == 2)
506 decorrelation = LEFT_SIDE + assignment - 8;
509 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
513 sample_size_code = get_bits(&s->gb, 3);
514 if(sample_size_code == 0)
516 else if((sample_size_code != 3) && (sample_size_code != 7))
517 bps = sample_size_table[sample_size_code];
520 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
524 if (get_bits1(&s->gb))
526 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
530 if(get_utf8(&s->gb) < 0){
531 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
535 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
536 (s->min_blocksize != s->max_blocksize)){
541 if (blocksize_code == 0)
542 blocksize = s->min_blocksize;
543 else if (blocksize_code == 6)
544 blocksize = get_bits(&s->gb, 8)+1;
545 else if (blocksize_code == 7)
546 blocksize = get_bits(&s->gb, 16)+1;
548 blocksize = blocksize_table[blocksize_code];
550 if(blocksize > s->max_blocksize){
551 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
555 if(blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
558 if (sample_rate_code == 0){
559 samplerate= s->samplerate;
560 }else if ((sample_rate_code > 3) && (sample_rate_code < 12))
561 samplerate = sample_rate_table[sample_rate_code];
562 else if (sample_rate_code == 12)
563 samplerate = get_bits(&s->gb, 8) * 1000;
564 else if (sample_rate_code == 13)
565 samplerate = get_bits(&s->gb, 16);
566 else if (sample_rate_code == 14)
567 samplerate = get_bits(&s->gb, 16) * 10;
569 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
573 skip_bits(&s->gb, 8);
574 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
575 s->gb.buffer, get_bits_count(&s->gb)/8);
577 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
581 s->blocksize = blocksize;
582 s->samplerate = samplerate;
584 s->decorrelation= decorrelation;
586 // dump_headers(s->avctx, (FLACStreaminfo *)s);
589 for (i = 0; i < s->channels; i++)
591 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
592 if (decode_subframe(s, i) < 0)
596 align_get_bits(&s->gb);
599 skip_bits(&s->gb, 16); /* data crc */
604 static int flac_decode_frame(AVCodecContext *avctx,
605 void *data, int *data_size,
606 const uint8_t *buf, int buf_size)
608 FLACContext *s = avctx->priv_data;
609 int tmp = 0, i, j = 0, input_buf_size = 0;
610 int16_t *samples = data;
611 int alloc_data_size= *data_size;
615 if(s->max_framesize == 0){
616 s->max_framesize= 65536; // should hopefully be enough for the first header
617 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
620 if(1 && s->max_framesize){//FIXME truncated
621 if(s->bitstream_size < 4 || AV_RL32(s->bitstream) != MKTAG('f','L','a','C'))
622 buf_size= FFMIN(buf_size, s->max_framesize - FFMIN(s->bitstream_size, s->max_framesize));
623 input_buf_size= buf_size;
625 if(s->bitstream_size + buf_size < buf_size || s->bitstream_index + s->bitstream_size + buf_size < s->bitstream_index)
628 if(s->allocated_bitstream_size < s->bitstream_size + buf_size)
629 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->bitstream_size + buf_size);
631 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
632 // printf("memmove\n");
633 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
634 s->bitstream_index=0;
636 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
637 buf= &s->bitstream[s->bitstream_index];
638 buf_size += s->bitstream_size;
639 s->bitstream_size= buf_size;
641 if(buf_size < s->max_framesize && input_buf_size){
642 // printf("wanna more data ...\n");
643 return input_buf_size;
647 init_get_bits(&s->gb, buf, buf_size*8);
649 if(metadata_parse(s))
652 tmp = show_bits(&s->gb, 16);
653 if((tmp & 0xFFFE) != 0xFFF8){
654 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
655 while(get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
656 skip_bits(&s->gb, 8);
657 goto end; // we may not have enough bits left to decode a frame, so try next time
659 skip_bits(&s->gb, 16);
660 if (decode_frame(s, alloc_data_size) < 0){
661 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
663 s->bitstream_index=0;
669 /* fix the channel order here */
670 if (s->order == MID_SIDE)
672 short *left = samples;
673 short *right = samples + s->blocksize;
674 for (i = 0; i < s->blocksize; i += 2)
676 uint32_t x = s->decoded[0][i];
677 uint32_t y = s->decoded[0][i+1];
679 right[i] = x - (y / 2);
680 left[i] = right[i] + y;
682 *data_size = 2 * s->blocksize;
686 for (i = 0; i < s->channels; i++)
691 for (j = 0; j < s->blocksize; j++)
692 samples[(s->blocksize*i)+j] = s->decoded[i][j];
697 for (j = 0; j < s->blocksize; j++)
698 samples[(s->blocksize*i)+j] = s->decoded[0][j];
700 for (j = 0; j < s->blocksize; j++)
701 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
704 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
706 *data_size += s->blocksize;
710 #define DECORRELATE(left, right)\
711 assert(s->channels == 2);\
712 for (i = 0; i < s->blocksize; i++)\
714 int a= s->decoded[0][i];\
715 int b= s->decoded[1][i];\
716 *samples++ = ((left) << (24 - s->bps)) >> 8;\
717 *samples++ = ((right) << (24 - s->bps)) >> 8;\
721 switch(s->decorrelation)
724 for (j = 0; j < s->blocksize; j++)
726 for (i = 0; i < s->channels; i++)
727 *samples++ = (s->decoded[i][j] << (24 - s->bps)) >> 8;
735 DECORRELATE( (a-=b>>1) + b, a)
739 *data_size = (int8_t *)samples - (int8_t *)data;
740 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
742 // s->last_blocksize = s->blocksize;
744 i= (get_bits_count(&s->gb)+7)/8;
746 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
748 s->bitstream_index=0;
752 if(s->bitstream_size){
753 s->bitstream_index += i;
754 s->bitstream_size -= i;
755 return input_buf_size;
760 static av_cold int flac_decode_close(AVCodecContext *avctx)
762 FLACContext *s = avctx->priv_data;
765 for (i = 0; i < s->channels; i++)
767 av_freep(&s->decoded[i]);
769 av_freep(&s->bitstream);
774 static void flac_flush(AVCodecContext *avctx){
775 FLACContext *s = avctx->priv_data;
778 s->bitstream_index= 0;
781 AVCodec flac_decoder = {
792 .long_name= NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),