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"
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 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 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)
282 a = decoded[pred_order-1];
283 b = a - decoded[pred_order-2];
284 c = b - decoded[pred_order-2] + decoded[pred_order-3];
285 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
292 for (i = pred_order; i < blocksize; i++)
293 decoded[i] = a += decoded[i];
296 for (i = pred_order; i < blocksize; i++)
297 decoded[i] = a += b += decoded[i];
300 for (i = pred_order; i < blocksize; i++)
301 decoded[i] = a += b += c += decoded[i];
304 for (i = pred_order; i < blocksize; i++)
305 decoded[i] = a += b += c += d += decoded[i];
308 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
315 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
318 int coeff_prec, qlevel;
319 int coeffs[pred_order];
320 int32_t *decoded = s->decoded[channel];
322 // av_log(s->avctx, AV_LOG_DEBUG, " SUBFRAME LPC\n");
324 /* warm up samples */
325 // av_log(s->avctx, AV_LOG_DEBUG, " warm up samples: %d\n", pred_order);
327 for (i = 0; i < pred_order; i++)
329 decoded[i] = get_sbits(&s->gb, s->curr_bps);
330 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, decoded[i]);
333 coeff_prec = get_bits(&s->gb, 4) + 1;
334 if (coeff_prec == 16)
336 av_log(s->avctx, AV_LOG_DEBUG, "invalid coeff precision\n");
339 // av_log(s->avctx, AV_LOG_DEBUG, " qlp coeff prec: %d\n", coeff_prec);
340 qlevel = get_sbits(&s->gb, 5);
341 // av_log(s->avctx, AV_LOG_DEBUG, " quant level: %d\n", qlevel);
343 av_log(s->avctx, AV_LOG_DEBUG, "qlevel %d not supported, maybe buggy stream\n", qlevel);
347 for (i = 0; i < pred_order; i++)
349 coeffs[i] = get_sbits(&s->gb, coeff_prec);
350 // av_log(s->avctx, AV_LOG_DEBUG, " %d: %d\n", i, coeffs[i]);
353 if (decode_residuals(s, channel, pred_order) < 0)
358 for (i = pred_order; i < s->blocksize; i++)
361 for (j = 0; j < pred_order; j++)
362 sum += (int64_t)coeffs[j] * decoded[i-j-1];
363 decoded[i] += sum >> qlevel;
366 for (i = pred_order; i < s->blocksize-1; i += 2)
369 int d = decoded[i-pred_order];
371 for (j = pred_order-1; j > 0; j--)
380 d = decoded[i] += s0 >> qlevel;
382 decoded[i+1] += s1 >> qlevel;
384 if (i < s->blocksize)
387 for (j = 0; j < pred_order; j++)
388 sum += coeffs[j] * decoded[i-j-1];
389 decoded[i] += sum >> qlevel;
396 static inline int decode_subframe(FLACContext *s, int channel)
398 int type, wasted = 0;
401 s->curr_bps = s->bps;
403 if(s->decorrelation == RIGHT_SIDE)
406 if(s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
410 if (get_bits1(&s->gb))
412 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
415 type = get_bits(&s->gb, 6);
416 // wasted = get_bits1(&s->gb);
420 // while (!get_bits1(&s->gb))
424 // s->curr_bps -= wasted;
427 wasted= 16 - av_log2(show_bits(&s->gb, 17));
428 skip_bits(&s->gb, wasted+1);
429 s->curr_bps -= wasted;
431 if (get_bits1(&s->gb))
434 while (!get_bits1(&s->gb))
436 s->curr_bps -= wasted;
437 av_log(s->avctx, AV_LOG_DEBUG, "%d wasted bits\n", wasted);
440 //FIXME use av_log2 for types
443 av_log(s->avctx, AV_LOG_DEBUG, "coding type: constant\n");
444 tmp = get_sbits(&s->gb, s->curr_bps);
445 for (i = 0; i < s->blocksize; i++)
446 s->decoded[channel][i] = tmp;
450 av_log(s->avctx, AV_LOG_DEBUG, "coding type: verbatim\n");
451 for (i = 0; i < s->blocksize; i++)
452 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
454 else if ((type >= 8) && (type <= 12))
456 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: fixed\n");
457 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
462 // av_log(s->avctx, AV_LOG_DEBUG, "coding type: lpc\n");
463 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
468 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
475 for (i = 0; i < s->blocksize; i++)
476 s->decoded[channel][i] <<= wasted;
482 static int decode_frame(FLACContext *s, int alloc_data_size)
484 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
485 int decorrelation, bps, blocksize, samplerate;
487 blocksize_code = get_bits(&s->gb, 4);
489 sample_rate_code = get_bits(&s->gb, 4);
491 assignment = get_bits(&s->gb, 4); /* channel assignment */
492 if (assignment < 8 && s->channels == assignment+1)
493 decorrelation = INDEPENDENT;
494 else if (assignment >=8 && assignment < 11 && s->channels == 2)
495 decorrelation = LEFT_SIDE + assignment - 8;
498 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n", assignment, s->channels);
502 sample_size_code = get_bits(&s->gb, 3);
503 if(sample_size_code == 0)
505 else if((sample_size_code != 3) && (sample_size_code != 7))
506 bps = sample_size_table[sample_size_code];
509 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n", sample_size_code);
513 if (get_bits1(&s->gb))
515 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
519 if(get_utf8(&s->gb) < 0){
520 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
524 if (/*((blocksize_code == 6) || (blocksize_code == 7)) &&*/
525 (s->min_blocksize != s->max_blocksize)){
530 if (blocksize_code == 0)
531 blocksize = s->min_blocksize;
532 else if (blocksize_code == 6)
533 blocksize = get_bits(&s->gb, 8)+1;
534 else if (blocksize_code == 7)
535 blocksize = get_bits(&s->gb, 16)+1;
537 blocksize = blocksize_table[blocksize_code];
539 if(blocksize > s->max_blocksize){
540 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize, s->max_blocksize);
544 if(blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
547 if (sample_rate_code == 0){
548 samplerate= s->samplerate;
549 }else if ((sample_rate_code > 3) && (sample_rate_code < 12))
550 samplerate = sample_rate_table[sample_rate_code];
551 else if (sample_rate_code == 12)
552 samplerate = get_bits(&s->gb, 8) * 1000;
553 else if (sample_rate_code == 13)
554 samplerate = get_bits(&s->gb, 16);
555 else if (sample_rate_code == 14)
556 samplerate = get_bits(&s->gb, 16) * 10;
558 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n", sample_rate_code);
562 skip_bits(&s->gb, 8);
563 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
564 s->gb.buffer, get_bits_count(&s->gb)/8);
566 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
570 s->blocksize = blocksize;
571 s->samplerate = samplerate;
573 s->decorrelation= decorrelation;
575 // dump_headers(s->avctx, (FLACStreaminfo *)s);
578 for (i = 0; i < s->channels; i++)
580 // av_log(s->avctx, AV_LOG_DEBUG, "decoded: %x residual: %x\n", s->decoded[i], s->residual[i]);
581 if (decode_subframe(s, i) < 0)
585 align_get_bits(&s->gb);
588 skip_bits(&s->gb, 16); /* data crc */
593 static int flac_decode_frame(AVCodecContext *avctx,
594 void *data, int *data_size,
595 const uint8_t *buf, int buf_size)
597 FLACContext *s = avctx->priv_data;
598 int tmp = 0, i, j = 0, input_buf_size = 0;
599 int16_t *samples = data;
600 int alloc_data_size= *data_size;
604 if(s->max_framesize == 0){
605 s->max_framesize= 65536; // should hopefully be enough for the first header
606 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
609 if(1 && s->max_framesize){//FIXME truncated
610 buf_size= FFMAX(FFMIN(buf_size, s->max_framesize - s->bitstream_size), 0);
611 input_buf_size= buf_size;
613 if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
614 // printf("memmove\n");
615 memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
616 s->bitstream_index=0;
618 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
619 buf= &s->bitstream[s->bitstream_index];
620 buf_size += s->bitstream_size;
621 s->bitstream_size= buf_size;
623 if(buf_size < s->max_framesize){
624 // printf("wanna more data ...\n");
625 return input_buf_size;
629 init_get_bits(&s->gb, buf, buf_size*8);
631 if (!metadata_parse(s))
633 tmp = show_bits(&s->gb, 16);
634 if((tmp & 0xFFFE) != 0xFFF8){
635 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
636 while(get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
637 skip_bits(&s->gb, 8);
638 goto end; // we may not have enough bits left to decode a frame, so try next time
640 skip_bits(&s->gb, 16);
641 if (decode_frame(s, alloc_data_size) < 0){
642 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
644 s->bitstream_index=0;
651 /* fix the channel order here */
652 if (s->order == MID_SIDE)
654 short *left = samples;
655 short *right = samples + s->blocksize;
656 for (i = 0; i < s->blocksize; i += 2)
658 uint32_t x = s->decoded[0][i];
659 uint32_t y = s->decoded[0][i+1];
661 right[i] = x - (y / 2);
662 left[i] = right[i] + y;
664 *data_size = 2 * s->blocksize;
668 for (i = 0; i < s->channels; i++)
673 for (j = 0; j < s->blocksize; j++)
674 samples[(s->blocksize*i)+j] = s->decoded[i][j];
679 for (j = 0; j < s->blocksize; j++)
680 samples[(s->blocksize*i)+j] = s->decoded[0][j];
682 for (j = 0; j < s->blocksize; j++)
683 samples[(s->blocksize*i)+j] = s->decoded[0][j] - s->decoded[i][j];
686 // av_log(s->avctx, AV_LOG_DEBUG, "mid-side unsupported\n");
688 *data_size += s->blocksize;
692 #define DECORRELATE(left, right)\
693 assert(s->channels == 2);\
694 for (i = 0; i < s->blocksize; i++)\
696 int a= s->decoded[0][i];\
697 int b= s->decoded[1][i];\
698 *samples++ = ((left) << (24 - s->bps)) >> 8;\
699 *samples++ = ((right) << (24 - s->bps)) >> 8;\
703 switch(s->decorrelation)
706 for (j = 0; j < s->blocksize; j++)
708 for (i = 0; i < s->channels; i++)
709 *samples++ = (s->decoded[i][j] << (24 - s->bps)) >> 8;
717 DECORRELATE( (a-=b>>1) + b, a)
721 *data_size = (int8_t *)samples - (int8_t *)data;
722 // av_log(s->avctx, AV_LOG_DEBUG, "data size: %d\n", *data_size);
724 // s->last_blocksize = s->blocksize;
726 i= (get_bits_count(&s->gb)+7)/8;
728 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
730 s->bitstream_index=0;
734 if(s->bitstream_size){
735 s->bitstream_index += i;
736 s->bitstream_size -= i;
737 return input_buf_size;
742 static av_cold int flac_decode_close(AVCodecContext *avctx)
744 FLACContext *s = avctx->priv_data;
747 for (i = 0; i < s->channels; i++)
749 av_freep(&s->decoded[i]);
751 av_freep(&s->bitstream);
756 static void flac_flush(AVCodecContext *avctx){
757 FLACContext *s = avctx->priv_data;
760 s->bitstream_index= 0;
763 AVCodec flac_decoder = {
773 .long_name= "FLAC (Free Lossless Audio Codec)"