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
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
70 int32_t *decoded[MAX_CHANNELS]; ///< decoded samples
72 unsigned int bitstream_size;
73 unsigned int bitstream_index;
74 unsigned int allocated_bitstream_size;
77 static const int sample_rate_table[] =
79 88200, 176400, 192000,
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)
94 GET_UTF8(val, get_bits(gb, 8), return -1;)
98 static void allocate_buffers(FLACContext *s);
99 static int metadata_parse(FLACContext *s);
101 static av_cold int flac_decode_init(AVCodecContext *avctx)
103 FLACContext *s = avctx->priv_data;
106 if (avctx->extradata_size > 4) {
107 /* initialize based on the demuxer-supplied streamdata header */
108 if (avctx->extradata_size == FLAC_STREAMINFO_SIZE) {
109 ff_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s,
113 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size*8);
118 avctx->sample_fmt = SAMPLE_FMT_S16;
122 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
124 av_log(avctx, AV_LOG_DEBUG, " Blocksize: %d .. %d\n", s->min_blocksize,
126 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
127 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
128 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
129 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
132 static void allocate_buffers(FLACContext *s)
136 assert(s->max_blocksize);
138 if (s->max_framesize == 0 && s->max_blocksize) {
139 // FIXME header overhead
140 s->max_framesize= (s->channels * s->bps * s->max_blocksize + 7)/ 8;
143 for (i = 0; i < s->channels; i++) {
144 s->decoded[i] = av_realloc(s->decoded[i],
145 sizeof(int32_t)*s->max_blocksize);
148 if (s->allocated_bitstream_size < s->max_framesize)
149 s->bitstream= av_fast_realloc(s->bitstream,
150 &s->allocated_bitstream_size,
154 void ff_flac_parse_streaminfo(AVCodecContext *avctx, struct FLACStreaminfo *s,
155 const uint8_t *buffer)
158 init_get_bits(&gb, buffer, FLAC_STREAMINFO_SIZE*8);
160 /* mandatory streaminfo */
161 s->min_blocksize = get_bits(&gb, 16);
162 s->max_blocksize = get_bits(&gb, 16);
164 skip_bits(&gb, 24); /* skip min frame size */
165 s->max_framesize = get_bits_long(&gb, 24);
167 s->samplerate = get_bits_long(&gb, 20);
168 s->channels = get_bits(&gb, 3) + 1;
169 s->bps = get_bits(&gb, 5) + 1;
171 avctx->channels = s->channels;
172 avctx->sample_rate = s->samplerate;
173 avctx->bits_per_raw_sample = s->bps;
175 avctx->sample_fmt = SAMPLE_FMT_S32;
177 avctx->sample_fmt = SAMPLE_FMT_S16;
179 s->samples = get_bits_long(&gb, 32) << 4;
180 s->samples |= get_bits_long(&gb, 4);
182 skip_bits(&gb, 64); /* md5 sum */
183 skip_bits(&gb, 64); /* md5 sum */
185 dump_headers(avctx, s);
189 * Parse a list of metadata blocks. This list of blocks must begin with
191 * @param s the flac decoding context containing the gb bit reader used to
193 * @return 1 if some metadata was read, 0 if no fLaC marker was found
195 static int metadata_parse(FLACContext *s)
197 int i, metadata_last, metadata_type, metadata_size, streaminfo_updated=0;
198 int initial_pos= get_bits_count(&s->gb);
200 if (show_bits_long(&s->gb, 32) == MKBETAG('f','L','a','C')) {
201 skip_bits(&s->gb, 32);
204 metadata_last = get_bits1(&s->gb);
205 metadata_type = get_bits(&s->gb, 7);
206 metadata_size = get_bits_long(&s->gb, 24);
208 if (get_bits_count(&s->gb) + 8*metadata_size > s->gb.size_in_bits) {
209 skip_bits_long(&s->gb, initial_pos - get_bits_count(&s->gb));
214 switch (metadata_type) {
215 case FLAC_METADATA_TYPE_STREAMINFO:
216 ff_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s,
217 s->gb.buffer+get_bits_count(&s->gb)/8);
218 streaminfo_updated = 1;
221 for (i = 0; i < metadata_size; i++)
222 skip_bits(&s->gb, 8);
225 } while (!metadata_last);
227 if (streaminfo_updated)
234 static int decode_residuals(FLACContext *s, int channel, int pred_order)
236 int i, tmp, partition, method_type, rice_order;
237 int sample = 0, samples;
239 method_type = get_bits(&s->gb, 2);
240 if (method_type > 1) {
241 av_log(s->avctx, AV_LOG_ERROR, "illegal residual coding method %d\n",
246 rice_order = get_bits(&s->gb, 4);
248 samples= s->blocksize >> rice_order;
249 if (pred_order > samples) {
250 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n",
251 pred_order, samples);
257 for (partition = 0; partition < (1 << rice_order); partition++) {
258 tmp = get_bits(&s->gb, method_type == 0 ? 4 : 5);
259 if (tmp == (method_type == 0 ? 15 : 31)) {
260 tmp = get_bits(&s->gb, 5);
261 for (; i < samples; i++, sample++)
262 s->decoded[channel][sample] = get_sbits(&s->gb, tmp);
264 for (; i < samples; i++, sample++) {
265 s->decoded[channel][sample] = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
274 static int decode_subframe_fixed(FLACContext *s, int channel, int pred_order)
276 const int blocksize = s->blocksize;
277 int32_t *decoded = s->decoded[channel];
280 /* warm up samples */
281 for (i = 0; i < pred_order; i++) {
282 decoded[i] = get_sbits(&s->gb, s->curr_bps);
285 if (decode_residuals(s, channel, pred_order) < 0)
289 a = decoded[pred_order-1];
291 b = a - decoded[pred_order-2];
293 c = b - decoded[pred_order-2] + decoded[pred_order-3];
295 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
297 switch (pred_order) {
301 for (i = pred_order; i < blocksize; i++)
302 decoded[i] = a += decoded[i];
305 for (i = pred_order; i < blocksize; i++)
306 decoded[i] = a += b += decoded[i];
309 for (i = pred_order; i < blocksize; i++)
310 decoded[i] = a += b += c += decoded[i];
313 for (i = pred_order; i < blocksize; i++)
314 decoded[i] = a += b += c += d += decoded[i];
317 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
324 static int decode_subframe_lpc(FLACContext *s, int channel, int pred_order)
327 int coeff_prec, qlevel;
328 int coeffs[pred_order];
329 int32_t *decoded = s->decoded[channel];
331 /* warm up samples */
332 for (i = 0; i < pred_order; i++) {
333 decoded[i] = get_sbits(&s->gb, s->curr_bps);
336 coeff_prec = get_bits(&s->gb, 4) + 1;
337 if (coeff_prec == 16) {
338 av_log(s->avctx, AV_LOG_ERROR, "invalid coeff precision\n");
341 qlevel = get_sbits(&s->gb, 5);
343 av_log(s->avctx, AV_LOG_ERROR, "qlevel %d not supported, maybe buggy stream\n",
348 for (i = 0; i < pred_order; i++) {
349 coeffs[i] = get_sbits(&s->gb, coeff_prec);
352 if (decode_residuals(s, channel, pred_order) < 0)
357 for (i = pred_order; i < s->blocksize; i++) {
359 for (j = 0; j < pred_order; j++)
360 sum += (int64_t)coeffs[j] * decoded[i-j-1];
361 decoded[i] += sum >> qlevel;
364 for (i = pred_order; i < s->blocksize-1; i += 2) {
366 int d = decoded[i-pred_order];
368 for (j = pred_order-1; j > 0; j--) {
376 d = decoded[i] += s0 >> qlevel;
378 decoded[i+1] += s1 >> qlevel;
380 if (i < s->blocksize) {
382 for (j = 0; j < pred_order; j++)
383 sum += coeffs[j] * decoded[i-j-1];
384 decoded[i] += sum >> qlevel;
391 static inline int decode_subframe(FLACContext *s, int channel)
393 int type, wasted = 0;
396 s->curr_bps = s->bps;
398 if (s->decorrelation == RIGHT_SIDE)
401 if (s->decorrelation == LEFT_SIDE || s->decorrelation == MID_SIDE)
405 if (get_bits1(&s->gb)) {
406 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
409 type = get_bits(&s->gb, 6);
411 if (get_bits1(&s->gb)) {
413 while (!get_bits1(&s->gb))
415 s->curr_bps -= wasted;
418 //FIXME use av_log2 for types
420 tmp = get_sbits(&s->gb, s->curr_bps);
421 for (i = 0; i < s->blocksize; i++)
422 s->decoded[channel][i] = tmp;
423 } else if (type == 1) {
424 for (i = 0; i < s->blocksize; i++)
425 s->decoded[channel][i] = get_sbits(&s->gb, s->curr_bps);
426 } else if ((type >= 8) && (type <= 12)) {
427 if (decode_subframe_fixed(s, channel, type & ~0x8) < 0)
429 } else if (type >= 32) {
430 if (decode_subframe_lpc(s, channel, (type & ~0x20)+1) < 0)
433 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, int alloc_data_size)
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;
461 av_log(s->avctx, AV_LOG_ERROR, "unsupported channel assignment %d (channels=%d)\n",
462 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];
472 av_log(s->avctx, AV_LOG_ERROR, "invalid sample size code (%d)\n",
477 s->avctx->sample_fmt = SAMPLE_FMT_S32;
478 s->sample_shift = 32 - bps;
481 s->avctx->sample_fmt = SAMPLE_FMT_S16;
482 s->sample_shift = 16 - bps;
485 s->bps = s->avctx->bits_per_raw_sample = bps;
487 if (get_bits1(&s->gb)) {
488 av_log(s->avctx, AV_LOG_ERROR, "broken stream, invalid padding\n");
492 if (get_utf8(&s->gb) < 0) {
493 av_log(s->avctx, AV_LOG_ERROR, "utf8 fscked\n");
497 if (blocksize_code == 0)
498 blocksize = s->min_blocksize;
499 else if (blocksize_code == 6)
500 blocksize = get_bits(&s->gb, 8)+1;
501 else if (blocksize_code == 7)
502 blocksize = get_bits(&s->gb, 16)+1;
504 blocksize = blocksize_table[blocksize_code];
506 if (blocksize > s->max_blocksize) {
507 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", blocksize,
512 if (blocksize * s->channels * sizeof(int16_t) > alloc_data_size)
515 if (sample_rate_code == 0)
516 samplerate= s->samplerate;
517 else if (sample_rate_code < 12)
518 samplerate = sample_rate_table[sample_rate_code];
519 else if (sample_rate_code == 12)
520 samplerate = get_bits(&s->gb, 8) * 1000;
521 else if (sample_rate_code == 13)
522 samplerate = get_bits(&s->gb, 16);
523 else if (sample_rate_code == 14)
524 samplerate = get_bits(&s->gb, 16) * 10;
526 av_log(s->avctx, AV_LOG_ERROR, "illegal sample rate code %d\n",
531 skip_bits(&s->gb, 8);
532 crc8 = av_crc(av_crc_get_table(AV_CRC_8_ATM), 0,
533 s->gb.buffer, get_bits_count(&s->gb)/8);
535 av_log(s->avctx, AV_LOG_ERROR, "header crc mismatch crc=%2X\n", crc8);
539 s->blocksize = blocksize;
540 s->samplerate = samplerate;
542 s->decorrelation= decorrelation;
544 // dump_headers(s->avctx, (FLACStreaminfo *)s);
547 for (i = 0; i < s->channels; i++) {
548 if (decode_subframe(s, i) < 0)
552 align_get_bits(&s->gb);
555 skip_bits(&s->gb, 16); /* data crc */
560 static int flac_decode_frame(AVCodecContext *avctx,
561 void *data, int *data_size,
562 const uint8_t *buf, int buf_size)
564 FLACContext *s = avctx->priv_data;
565 int tmp = 0, i, j = 0, input_buf_size = 0;
566 int16_t *samples_16 = data;
567 int32_t *samples_32 = data;
568 int alloc_data_size= *data_size;
572 if (s->max_framesize == 0) {
573 s->max_framesize= FFMAX(4, buf_size); // should hopefully be enough for the first header
574 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
577 if (1 && s->max_framesize) { //FIXME truncated
578 if (s->bitstream_size < 4 || AV_RL32(s->bitstream) != MKTAG('f','L','a','C'))
579 buf_size= FFMIN(buf_size, s->max_framesize - FFMIN(s->bitstream_size, s->max_framesize));
580 input_buf_size= buf_size;
582 if (s->bitstream_size + buf_size < buf_size || s->bitstream_index + s->bitstream_size + buf_size < s->bitstream_index)
585 if (s->allocated_bitstream_size < s->bitstream_size + buf_size)
586 s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->bitstream_size + buf_size);
588 if (s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size) {
589 memmove(s->bitstream, &s->bitstream[s->bitstream_index],
591 s->bitstream_index=0;
593 memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size],
595 buf= &s->bitstream[s->bitstream_index];
596 buf_size += s->bitstream_size;
597 s->bitstream_size= buf_size;
599 if (buf_size < s->max_framesize && input_buf_size) {
600 return input_buf_size;
604 init_get_bits(&s->gb, buf, buf_size*8);
606 if (metadata_parse(s))
609 tmp = show_bits(&s->gb, 16);
610 if ((tmp & 0xFFFE) != 0xFFF8) {
611 av_log(s->avctx, AV_LOG_ERROR, "FRAME HEADER not here\n");
612 while (get_bits_count(&s->gb)/8+2 < buf_size && (show_bits(&s->gb, 16) & 0xFFFE) != 0xFFF8)
613 skip_bits(&s->gb, 8);
614 goto end; // we may not have enough bits left to decode a frame, so try next time
616 skip_bits(&s->gb, 16);
617 if (decode_frame(s, alloc_data_size) < 0) {
618 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
620 s->bitstream_index=0;
624 #define DECORRELATE(left, right)\
625 assert(s->channels == 2);\
626 for (i = 0; i < s->blocksize; i++) {\
627 int a= s->decoded[0][i];\
628 int b= s->decoded[1][i];\
630 *samples_32++ = (left) << s->sample_shift;\
631 *samples_32++ = (right) << s->sample_shift;\
633 *samples_16++ = (left) << s->sample_shift;\
634 *samples_16++ = (right) << s->sample_shift;\
639 switch (s->decorrelation) {
641 for (j = 0; j < s->blocksize; j++) {
642 for (i = 0; i < s->channels; i++) {
644 *samples_32++ = s->decoded[i][j] << s->sample_shift;
646 *samples_16++ = s->decoded[i][j] << s->sample_shift;
655 DECORRELATE( (a-=b>>1) + b, a)
658 *data_size = s->blocksize * s->channels * (s->is32 ? 4 : 2);
661 i= (get_bits_count(&s->gb)+7)/8;
663 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
665 s->bitstream_index=0;
669 if (s->bitstream_size) {
670 s->bitstream_index += i;
671 s->bitstream_size -= i;
672 return input_buf_size;
677 static av_cold int flac_decode_close(AVCodecContext *avctx)
679 FLACContext *s = avctx->priv_data;
682 for (i = 0; i < s->channels; i++) {
683 av_freep(&s->decoded[i]);
685 av_freep(&s->bitstream);
690 static void flac_flush(AVCodecContext *avctx)
692 FLACContext *s = avctx->priv_data;
695 s->bitstream_index= 0;
698 AVCodec flac_decoder = {
709 .long_name= NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),