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
23 * @file libavcodec/flacdec.c
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 #include "libavutil/crc.h"
41 #include "bitstream.h"
48 #define MAX_CHANNELS 8
49 #define MAX_BLOCKSIZE 65535
51 enum decorrelation_type {
58 typedef struct FLACContext {
61 AVCodecContext *avctx; ///< parent AVCodecContext
62 GetBitContext gb; ///< GetBitContext initialized to start at the current frame
64 int blocksize; ///< number of samples in the current frame
65 int curr_bps; ///< bps for current subframe, adjusted for channel correlation and wasted bits
66 int sample_shift; ///< shift required to make output samples 16-bit or 32-bit
67 int is32; ///< flag to indicate if output should be 32-bit instead of 16-bit
68 enum decorrelation_type decorrelation; ///< channel decorrelation type in the current frame
69 int got_streaminfo; ///< indicates if the STREAMINFO has been read
71 int32_t *decoded[MAX_CHANNELS]; ///< decoded samples
73 unsigned int bitstream_size;
74 unsigned int bitstream_index;
75 unsigned int allocated_bitstream_size;
78 static const int sample_rate_table[] =
80 88200, 176400, 192000,
81 8000, 16000, 22050, 24000, 32000, 44100, 48000, 96000,
84 static const int sample_size_table[] =
85 { 0, 8, 12, 0, 16, 20, 24, 0 };
87 static const int blocksize_table[] = {
88 0, 192, 576<<0, 576<<1, 576<<2, 576<<3, 0, 0,
89 256<<0, 256<<1, 256<<2, 256<<3, 256<<4, 256<<5, 256<<6, 256<<7
92 static int64_t get_utf8(GetBitContext *gb)
95 GET_UTF8(val, get_bits(gb, 8), return -1;)
99 static void allocate_buffers(FLACContext *s);
101 int ff_flac_is_extradata_valid(AVCodecContext *avctx,
102 enum FLACExtradataFormat *format,
103 uint8_t **streaminfo_start)
105 if (!avctx->extradata || avctx->extradata_size < FLAC_STREAMINFO_SIZE) {
106 av_log(avctx, AV_LOG_ERROR, "extradata NULL or too small.\n");
109 if (AV_RL32(avctx->extradata) != MKTAG('f','L','a','C')) {
110 /* extradata contains STREAMINFO only */
111 if (avctx->extradata_size != FLAC_STREAMINFO_SIZE) {
112 av_log(avctx, AV_LOG_WARNING, "extradata contains %d bytes too many.\n",
113 FLAC_STREAMINFO_SIZE-avctx->extradata_size);
115 *format = FLAC_EXTRADATA_FORMAT_STREAMINFO;
116 *streaminfo_start = avctx->extradata;
118 if (avctx->extradata_size < 8+FLAC_STREAMINFO_SIZE) {
119 av_log(avctx, AV_LOG_ERROR, "extradata too small.\n");
122 *format = FLAC_EXTRADATA_FORMAT_FULL_HEADER;
123 *streaminfo_start = &avctx->extradata[8];
128 static av_cold int flac_decode_init(AVCodecContext *avctx)
130 enum FLACExtradataFormat format;
132 FLACContext *s = avctx->priv_data;
135 avctx->sample_fmt = SAMPLE_FMT_S16;
137 /* for now, the raw FLAC header is allowed to be passed to the decoder as
138 frame data instead of extradata. */
139 if (!avctx->extradata)
142 if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
145 /* initialize based on the demuxer-supplied streamdata header */
146 ff_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s, streaminfo);
148 s->got_streaminfo = 1;
153 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
155 av_log(avctx, AV_LOG_DEBUG, " Max Blocksize: %d\n", s->max_blocksize);
156 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
157 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
158 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
159 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
162 static void allocate_buffers(FLACContext *s)
166 assert(s->max_blocksize);
168 if (s->max_framesize == 0 && s->max_blocksize) {
169 // FIXME header overhead
170 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8;
173 for (i = 0; i < s->channels; i++) {
174 s->decoded[i] = av_realloc(s->decoded[i],
175 sizeof(int32_t)*s->max_blocksize);
178 if (s->allocated_bitstream_size < s->max_framesize)
179 s->bitstream= av_fast_realloc(s->bitstream,
180 &s->allocated_bitstream_size,
184 void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
185 const uint8_t *buffer)
188 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
190 skip_bits(&gb, 16); /* skip min blocksize */
191 s->max_blocksize = get_bits(&gb, 16);
192 if (s->max_blocksize < 16) {
193 av_log(avctx, AV_LOG_WARNING, "invalid max blocksize: %d\n",
195 s->max_blocksize = 16;
198 skip_bits(&gb, 24); /* skip min frame size */
199 s->max_framesize = get_bits_long(&gb, 24);
201 s->samplerate = get_bits_long(&gb, 20);
202 s->channels = get_bits(&gb, 3) + 1;
203 s->bps = get_bits(&gb, 5) + 1;
205 avctx->channels = s->channels;
206 avctx->sample_rate = s->samplerate;
207 avctx->bits_per_raw_sample = s->bps;
209 avctx->sample_fmt = SAMPLE_FMT_S32;
211 avctx->sample_fmt = SAMPLE_FMT_S16;
213 s->samples = get_bits_long(&gb, 32) << 4;
214 s->samples |= get_bits(&gb, 4);
216 skip_bits_long(&gb, 64); /* md5 sum */
217 skip_bits_long(&gb, 64); /* md5 sum */
219 dump_headers(avctx, s);
223 * Parse a list of metadata blocks. This list of blocks must begin with
225 * @param s the flac decoding context containing the gb bit reader used to
227 * @return 1 if some metadata was read, 0 if no fLaC marker was found
229 static int metadata_parse(FLACContext *s)
231 int i, metadata_last, metadata_type, metadata_size;
232 int initial_pos= get_bits_count(&s->gb);
234 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
235 skip_bits_long(&s->gb, 32);
238 metadata_last = get_bits1(&s->gb);
239 metadata_type = get_bits(&s->gb, 7);
240 metadata_size = get_bits_long(&s->gb, 24);
242 if (get_bits_count(&s->gb) + 8*metadata_size > s->gb.size_in_bits) {
243 skip_bits_long(&s->gb, initial_pos - get_bits_count(&s->gb));
248 switch (metadata_type) {
249 case FLAC_METADATA_TYPE_STREAMINFO:
250 if (!s->got_streaminfo) {
251 ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s,
252 s->gb.buffer+get_bits_count(&s->gb)/8);
253 s->got_streaminfo = 1;
256 for (i = 0; i < metadata_size; i++)
257 skip_bits(&s->gb, 8);
260 } while (!metadata_last);
262 if (s->got_streaminfo)
269 static int decode_residuals(FLACContext *s, int channel, int pred_order)
271 int i, tmp, partition, method_type, rice_order;
272 int sample = 0, samples;
274 method_type = get_bits(&s->gb, 2);
275 if (method_type > 1) {
276 av_log(s->avctx, AV_LOG_ERROR, "illegal residual coding method %d\n",
281 rice_order = get_bits(&s->gb, 4);
283 samples= s->blocksize >> rice_order;
284 if (pred_order > samples) {
285 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n",
286 pred_order, samples);
292 for (partition = 0; partition < (1 << rice_order); partition++) {
293 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
294 if (tmp == (method_type == 0 ? 15 : 31)) {
295 tmp = get_bits(&s->gb, 5);
296 for (; i < samples; i++, sample++)
297 s->decoded[channel][sample] = get_sbits_long(&s->gb, tmp);
299 for (; i < samples; i++, sample++) {
300 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
309 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
311 const int blocksize = s->blocksize;
312 int32_t *decoded = s->decoded[channel];
313 int av_uninit(a), av_uninit(b), av_uninit(c), av_uninit(d), i;
315 /* warm up samples */
316 for (i = 0; i < pred_order; i++) {
317 decoded[i] = get_sbits_long(&s->gb, s->curr_bps);
320 if (decode_residuals(s, channel, pred_order) < 0)
324 a = decoded[pred_order-1];
326 b = a - decoded[pred_order-2];
328 c = b - decoded[pred_order-2] + decoded[pred_order-3];
330 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
332 switch (pred_order) {
336 for (i = pred_order; i < blocksize; i++)
337 decoded[i] = a += decoded[i];
340 for (i = pred_order; i < blocksize; i++)
341 decoded[i] = a += b += decoded[i];
344 for (i = pred_order; i < blocksize; i++)
345 decoded[i] = a += b += c += decoded[i];
348 for (i = pred_order; i < blocksize; i++)
349 decoded[i] = a += b += c += d += decoded[i];
352 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
359 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
362 int coeff_prec, qlevel;
363 int coeffs[pred_order];
364 int32_t *decoded = s->decoded[channel];
366 /* warm up samples */
367 for (i = 0; i < pred_order; i++) {
368 decoded[i] = get_sbits_long(&s->gb, s->curr_bps);
371 coeff_prec = get_bits(&s->gb, 4) + 1;
372 if (coeff_prec == 16) {
373 av_log(s->avctx, AV_LOG_ERROR, "invalid coeff precision\n");
376 qlevel = get_sbits(&s->gb, 5);
378 av_log(s->avctx, AV_LOG_ERROR, "qlevel %d not supported, maybe buggy stream\n",
383 for (i = 0; i < pred_order; i++) {
384 coeffs[i] = get_sbits(&s->gb, coeff_prec);
387 if (decode_residuals(s, channel, pred_order) < 0)
392 for (i = pred_order; i < s->blocksize; i++) {
394 for (j = 0; j < pred_order; j++)
395 sum += (int64_t)coeffs[j] * decoded[i-j-1];
396 decoded[i] += sum >> qlevel;
399 for (i = pred_order; i < s->blocksize-1; i += 2) {
401 int d = decoded[i-pred_order];
403 for (j = pred_order-1; j > 0; j--) {
411 d = decoded[i] += s0 >> qlevel;
413 decoded[i+1] += s1 >> qlevel;
415 if (i < s->blocksize) {
417 for (j = 0; j < pred_order; j++)
418 sum += coeffs[j] * decoded[i-j-1];
419 decoded[i] += sum >> qlevel;
426 static inline int decode_subframe(FLACContext *s, int channel)
428 int type, wasted = 0;
431 s->curr_bps = s->bps;
433 if (s->decorrelation == RIGHT_SIDE)
436 if (s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
439 if (s->curr_bps > 32) {
440 ff_log_missing_feature(s->avctx, "decorrelated bit depth > 32", 0);
444 if (get_bits1(&s->gb)) {
445 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
448 type = get_bits(&s->gb, 6);
450 if (get_bits1(&s->gb)) {
452 while (!get_bits1(&s->gb))
454 s->curr_bps -= wasted;
457 //FIXME use av_log2 for types
459 tmp = get_sbits_long(&s->gb, s->curr_bps);
460 for (i = 0; i < s->blocksize; i++)
461 s->decoded[channel][i] = tmp;
462 } else if (type == 1) {
463 for (i = 0; i < s->blocksize; i++)
464 s->decoded[channel][i] = get_sbits_long(&s->gb, s->curr_bps);
465 } else if ((type >= 8) && (type <= 12)) {
466 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
468 } else if (type >= 32) {
469 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
472 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
478 for (i = 0; i < s->blocksize; i++)
479 s->decoded[channel][i] <<= wasted;
485 static int decode_frame(FLACContext *s, int alloc_data_size)
487 int blocksize_code, sample_rate_code, sample_size_code, assignment, i, crc8;
488 int decorrelation, bps, blocksize, samplerate;
490 blocksize_code = get_bits(&s->gb, 4);
492 sample_rate_code = get_bits(&s->gb, 4);
494 assignment = get_bits(&s->gb, 4); /* channel assignment */
495 if (assignment < 8 && s->channels == assignment+1)
496 decorrelation = INDEPENDENT;
497 else if (assignment >=8 && assignment < 11 && s->channels == 2)
498 decorrelation = LEFT_SIDE + assignment - 8;
500 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n",
501 assignment, s->channels);
505 sample_size_code = get_bits(&s->gb, 3);
506 if (sample_size_code == 0)
508 else if ((sample_size_code != 3) && (sample_size_code != 7))
509 bps = sample_size_table[sample_size_code];
511 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n",
516 s->avctx->sample_fmt = SAMPLE_FMT_S32;
517 s->sample_shift = 32 - bps;
520 s->avctx->sample_fmt = SAMPLE_FMT_S16;
521 s->sample_shift = 16 - bps;
524 s->bps = s->avctx->bits_per_raw_sample = bps;
526 if (get_bits1(&s->gb)) {
527 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
531 if (get_utf8(&s->gb) < 0) {
532 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
536 if (blocksize_code == 0) {
537 av_log(s->avctx, AV_LOG_ERROR, "reserved blocksize code: 0\n");
539 } else if (blocksize_code == 6)
540 blocksize = get_bits(&s->gb, 8)+1;
541 else if (blocksize_code == 7)
542 blocksize = get_bits(&s->gb, 16)+1;
544 blocksize = blocksize_table[blocksize_code];
546 if (blocksize > s->max_blocksize) {
547 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize,
552 if (blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
555 if (sample_rate_code == 0)
556 samplerate= s->samplerate;
557 else if (sample_rate_code < 12)
558 samplerate = sample_rate_table[sample_rate_code];
559 else if (sample_rate_code == 12)
560 samplerate = get_bits(&s->gb, 8) * 1000;
561 else if (sample_rate_code == 13)
562 samplerate = get_bits(&s->gb, 16);
563 else if (sample_rate_code == 14)
564 samplerate = get_bits(&s->gb, 16) * 10;
566 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n",
571 skip_bits(&s->gb, 8);
572 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
573 s->gb.buffer, get_bits_count(&s->gb)/8);
575 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
579 s->blocksize = blocksize;
580 s->samplerate = samplerate;
582 s->decorrelation= decorrelation;
584 // dump_headers(s->avctx, (FLACStreaminfo *)s);
587 for (i = 0; i < s->channels; i++) {
588 if (decode_subframe(s, i) < 0)
592 align_get_bits(&s->gb);
595 skip_bits(&s->gb, 16); /* data crc */
600 static int flac_decode_frame(AVCodecContext *avctx,
601 void *data, int *data_size,
602 const uint8_t *buf, int buf_size)
604 FLACContext *s = avctx->priv_data;
605 int tmp = 0, i, j = 0, input_buf_size = 0;
606 int16_t *samples_16 = data;
607 int32_t *samples_32 = data;
608 int alloc_data_size= *data_size;
612 if (s->max_framesize == 0) {
613 s->max_framesize= FFMAX(4, buf_size); // should hopefully be enough for the first header
614 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
617 if (1 && s->max_framesize) { //FIXME truncated
618 if (s->bitstream_size < 4 || AV_RL32(s->bitstream) != MKTAG('f','L','a','C'))
619 buf_size= FFMIN(buf_size, s->max_framesize - FFMIN(s->bitstream_size, s->max_framesize));
620 input_buf_size= buf_size;
622 if (s->bitstream_size + buf_size < buf_size || s->bitstream_index + s->bitstream_size + buf_size < s->bitstream_index)
625 if (s->allocated_bitstream_size < s->bitstream_size + buf_size)
626 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->bitstream_size + buf_size);
628 if (s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size) {
629 memmove(s->bitstream, &s->bitstream[s->bitstream_index],
631 s->bitstream_index=0;
633 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size],
635 buf= &s->bitstream[s->bitstream_index];
636 buf_size += s->bitstream_size;
637 s->bitstream_size= buf_size;
639 if (buf_size < s->max_framesize && input_buf_size) {
640 return input_buf_size;
644 init_get_bits(&s->gb, buf, buf_size*8);
646 if (metadata_parse(s))
649 tmp = show_bits(&s->gb, 16);
650 if ((tmp & 0xFFFE) != 0xFFF8) {
651 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
652 while (get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
653 skip_bits(&s->gb, 8);
654 goto end; // we may not have enough bits left to decode a frame, so try next time
656 skip_bits(&s->gb, 16);
657 if (decode_frame(s, alloc_data_size) < 0) {
658 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
660 s->bitstream_index=0;
664 #define DECORRELATE(left, right)\
665 assert(s->channels == 2);\
666 for (i = 0; i < s->blocksize; i++) {\
667 int a= s->decoded[0][i];\
668 int b= s->decoded[1][i];\
670 *samples_32++ = (left) << s->sample_shift;\
671 *samples_32++ = (right) << s->sample_shift;\
673 *samples_16++ = (left) << s->sample_shift;\
674 *samples_16++ = (right) << s->sample_shift;\
679 switch (s->decorrelation) {
681 for (j = 0; j < s->blocksize; j++) {
682 for (i = 0; i < s->channels; i++) {
684 *samples_32++ = s->decoded[i][j] << s->sample_shift;
686 *samples_16++ = s->decoded[i][j] << s->sample_shift;
695 DECORRELATE( (a-=b>>1) + b, a)
698 *data_size = s->blocksize * s->channels * (s->is32 ? 4 : 2);
701 i= (get_bits_count(&s->gb)+7)/8;
703 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
705 s->bitstream_index=0;
709 if (s->bitstream_size) {
710 s->bitstream_index += i;
711 s->bitstream_size -= i;
712 return input_buf_size;
717 static av_cold int flac_decode_close(AVCodecContext *avctx)
719 FLACContext *s = avctx->priv_data;
722 for (i = 0; i < s->channels; i++) {
723 av_freep(&s->decoded[i]);
725 av_freep(&s->bitstream);
730 static void flac_flush(AVCodecContext *avctx)
732 FLACContext *s = avctx->priv_data;
735 s->bitstream_index= 0;
738 AVCodec flac_decoder = {
749 .long_name= NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),