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
26 * @see http://flac.sourceforge.net/
28 * This decoder can be used in 1 of 2 ways: Either raw FLAC data can be fed
29 * through, starting from the initial 'fLaC' signature; or by passing the
30 * 34-byte streaminfo structure through avctx->extradata[_size] followed
31 * by data starting with the 0xFFF8 marker.
36 #include "libavutil/audioconvert.h"
37 #include "libavutil/avassert.h"
38 #include "libavutil/crc.h"
42 #include "bytestream.h"
48 typedef struct FLACContext {
51 AVCodecContext *avctx; ///< parent AVCodecContext
53 GetBitContext gb; ///< GetBitContext initialized to start at the current frame
55 int blocksize; ///< number of samples in the current frame
56 int sample_shift; ///< shift required to make output samples 16-bit or 32-bit
57 int ch_mode; ///< channel decorrelation type in the current frame
58 int got_streaminfo; ///< indicates if the STREAMINFO has been read
60 int32_t *decoded[FLAC_MAX_CHANNELS]; ///< decoded samples
65 static const int64_t flac_channel_layouts[6] = {
68 AV_CH_LAYOUT_SURROUND,
74 static void allocate_buffers(FLACContext *s);
76 static void flac_set_bps(FLACContext *s)
78 enum AVSampleFormat req = s->avctx->request_sample_fmt;
79 int need32 = s->bps > 16;
80 int want32 = av_get_bytes_per_sample(req) > 2;
81 int planar = av_sample_fmt_is_planar(req);
83 if (need32 || want32) {
85 s->avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
87 s->avctx->sample_fmt = AV_SAMPLE_FMT_S32;
88 s->sample_shift = 32 - s->bps;
91 s->avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
93 s->avctx->sample_fmt = AV_SAMPLE_FMT_S16;
94 s->sample_shift = 16 - s->bps;
98 static av_cold int flac_decode_init(AVCodecContext *avctx)
100 enum FLACExtradataFormat format;
102 FLACContext *s = avctx->priv_data;
105 /* for now, the raw FLAC header is allowed to be passed to the decoder as
106 frame data instead of extradata. */
107 if (!avctx->extradata)
110 if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
113 /* initialize based on the demuxer-supplied streamdata header */
114 avpriv_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s, streaminfo);
117 ff_flacdsp_init(&s->dsp, avctx->sample_fmt, s->bps);
118 s->got_streaminfo = 1;
120 avcodec_get_frame_defaults(&s->frame);
121 avctx->coded_frame = &s->frame;
123 if (avctx->channels <= FF_ARRAY_ELEMS(flac_channel_layouts))
124 avctx->channel_layout = flac_channel_layouts[avctx->channels - 1];
129 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
131 av_log(avctx, AV_LOG_DEBUG, " Max Blocksize: %d\n", s->max_blocksize);
132 av_log(avctx, AV_LOG_DEBUG, " Max Framesize: %d\n", s->max_framesize);
133 av_log(avctx, AV_LOG_DEBUG, " Samplerate: %d\n", s->samplerate);
134 av_log(avctx, AV_LOG_DEBUG, " Channels: %d\n", s->channels);
135 av_log(avctx, AV_LOG_DEBUG, " Bits: %d\n", s->bps);
138 static void allocate_buffers(FLACContext *s)
142 av_assert0(s->max_blocksize);
144 for (i = 0; i < s->channels; i++) {
145 s->decoded[i] = av_malloc(sizeof(int32_t)*s->max_blocksize);
150 * Parse the STREAMINFO from an inline header.
151 * @param s the flac decoding context
152 * @param buf input buffer, starting with the "fLaC" marker
153 * @param buf_size buffer size
154 * @return non-zero if metadata is invalid
156 static int parse_streaminfo(FLACContext *s, const uint8_t *buf, int buf_size)
158 int metadata_type, metadata_size;
160 if (buf_size < FLAC_STREAMINFO_SIZE+8) {
164 avpriv_flac_parse_block_header(&buf[4], NULL, &metadata_type, &metadata_size);
165 if (metadata_type != FLAC_METADATA_TYPE_STREAMINFO ||
166 metadata_size != FLAC_STREAMINFO_SIZE) {
167 return AVERROR_INVALIDDATA;
169 avpriv_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, &buf[8]);
172 ff_flacdsp_init(&s->dsp, s->avctx->sample_fmt, s->bps);
173 s->got_streaminfo = 1;
179 * Determine the size of an inline header.
180 * @param buf input buffer, starting with the "fLaC" marker
181 * @param buf_size buffer size
182 * @return number of bytes in the header, or 0 if more data is needed
184 static int get_metadata_size(const uint8_t *buf, int buf_size)
186 int metadata_last, metadata_size;
187 const uint8_t *buf_end = buf + buf_size;
191 if (buf_end - buf < 4)
193 avpriv_flac_parse_block_header(buf, &metadata_last, NULL, &metadata_size);
195 if (buf_end - buf < metadata_size) {
196 /* need more data in order to read the complete header */
199 buf += metadata_size;
200 } while (!metadata_last);
202 return buf_size - (buf_end - buf);
205 static int decode_residuals(FLACContext *s, int32_t *decoded, int pred_order)
207 int i, tmp, partition, method_type, rice_order;
208 int rice_bits, rice_esc;
211 method_type = get_bits(&s->gb, 2);
212 if (method_type > 1) {
213 av_log(s->avctx, AV_LOG_ERROR, "illegal residual coding method %d\n",
218 rice_order = get_bits(&s->gb, 4);
220 samples= s->blocksize >> rice_order;
221 if (pred_order > samples) {
222 av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n",
223 pred_order, samples);
227 rice_bits = 4 + method_type;
228 rice_esc = (1 << rice_bits) - 1;
230 decoded += pred_order;
232 for (partition = 0; partition < (1 << rice_order); partition++) {
233 tmp = get_bits(&s->gb, rice_bits);
234 if (tmp == rice_esc) {
235 tmp = get_bits(&s->gb, 5);
236 for (; i < samples; i++)
237 *decoded++ = get_sbits_long(&s->gb, tmp);
239 for (; i < samples; i++) {
240 *decoded++ = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
249 static int decode_subframe_fixed(FLACContext *s, int32_t *decoded,
250 int pred_order, int bps)
252 const int blocksize = s->blocksize;
255 /* warm up samples */
256 for (i = 0; i < pred_order; i++) {
257 decoded[i] = get_sbits_long(&s->gb, bps);
260 if (decode_residuals(s, decoded, pred_order) < 0)
264 a = decoded[pred_order-1];
266 b = a - decoded[pred_order-2];
268 c = b - decoded[pred_order-2] + decoded[pred_order-3];
270 d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
272 switch (pred_order) {
276 for (i = pred_order; i < blocksize; i++)
277 decoded[i] = a += decoded[i];
280 for (i = pred_order; i < blocksize; i++)
281 decoded[i] = a += b += decoded[i];
284 for (i = pred_order; i < blocksize; i++)
285 decoded[i] = a += b += c += decoded[i];
288 for (i = pred_order; i < blocksize; i++)
289 decoded[i] = a += b += c += d += decoded[i];
292 av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
299 static int decode_subframe_lpc(FLACContext *s, int32_t *decoded, int pred_order,
303 int coeff_prec, qlevel;
306 /* warm up samples */
307 for (i = 0; i < pred_order; i++) {
308 decoded[i] = get_sbits_long(&s->gb, bps);
311 coeff_prec = get_bits(&s->gb, 4) + 1;
312 if (coeff_prec == 16) {
313 av_log(s->avctx, AV_LOG_ERROR, "invalid coeff precision\n");
316 qlevel = get_sbits(&s->gb, 5);
318 av_log(s->avctx, AV_LOG_ERROR, "qlevel %d not supported, maybe buggy stream\n",
323 for (i = 0; i < pred_order; i++) {
324 coeffs[pred_order - i - 1] = get_sbits(&s->gb, coeff_prec);
327 if (decode_residuals(s, decoded, pred_order) < 0)
330 s->dsp.lpc(decoded, coeffs, pred_order, qlevel, s->blocksize);
335 static inline int decode_subframe(FLACContext *s, int channel)
337 int32_t *decoded = s->decoded[channel];
338 int type, wasted = 0;
343 if (s->ch_mode == FLAC_CHMODE_RIGHT_SIDE)
346 if (s->ch_mode == FLAC_CHMODE_LEFT_SIDE || s->ch_mode == FLAC_CHMODE_MID_SIDE)
350 if (get_bits1(&s->gb)) {
351 av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
354 type = get_bits(&s->gb, 6);
356 if (get_bits1(&s->gb)) {
357 int left = get_bits_left(&s->gb);
360 (left < bps && !show_bits_long(&s->gb, left)) ||
361 !show_bits_long(&s->gb, bps)) {
362 av_log(s->avctx, AV_LOG_ERROR,
363 "Invalid number of wasted bits > available bits (%d) - left=%d\n",
365 return AVERROR_INVALIDDATA;
367 while (!get_bits1(&s->gb))
372 av_log_missing_feature(s->avctx, "decorrelated bit depth > 32", 0);
376 //FIXME use av_log2 for types
378 tmp = get_sbits_long(&s->gb, bps);
379 for (i = 0; i < s->blocksize; i++)
381 } else if (type == 1) {
382 for (i = 0; i < s->blocksize; i++)
383 decoded[i] = get_sbits_long(&s->gb, bps);
384 } else if ((type >= 8) && (type <= 12)) {
385 if (decode_subframe_fixed(s, decoded, type & ~0x8, bps) < 0)
387 } else if (type >= 32) {
388 if (decode_subframe_lpc(s, decoded, (type & ~0x20)+1, bps) < 0)
391 av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
397 for (i = 0; i < s->blocksize; i++)
398 decoded[i] <<= wasted;
404 static int decode_frame(FLACContext *s)
407 GetBitContext *gb = &s->gb;
410 if (ff_flac_decode_frame_header(s->avctx, gb, &fi, 0)) {
411 av_log(s->avctx, AV_LOG_ERROR, "invalid frame header\n");
415 if (s->channels && fi.channels != s->channels) {
416 av_log(s->avctx, AV_LOG_ERROR, "switching channel layout mid-stream "
417 "is not supported\n");
420 s->channels = s->avctx->channels = fi.channels;
421 s->ch_mode = fi.ch_mode;
423 if (!s->bps && !fi.bps) {
424 av_log(s->avctx, AV_LOG_ERROR, "bps not found in STREAMINFO or frame header\n");
429 } else if (s->bps && fi.bps != s->bps) {
430 av_log(s->avctx, AV_LOG_ERROR, "switching bps mid-stream is not "
436 s->bps = s->avctx->bits_per_raw_sample = fi.bps;
440 if (!s->max_blocksize)
441 s->max_blocksize = FLAC_MAX_BLOCKSIZE;
442 if (fi.blocksize > s->max_blocksize) {
443 av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", fi.blocksize,
447 s->blocksize = fi.blocksize;
449 if (!s->samplerate && !fi.samplerate) {
450 av_log(s->avctx, AV_LOG_ERROR, "sample rate not found in STREAMINFO"
451 " or frame header\n");
454 if (fi.samplerate == 0) {
455 fi.samplerate = s->samplerate;
456 } else if (s->samplerate && fi.samplerate != s->samplerate) {
457 av_log(s->avctx, AV_LOG_WARNING, "sample rate changed from %d to %d\n",
458 s->samplerate, fi.samplerate);
460 s->samplerate = s->avctx->sample_rate = fi.samplerate;
462 if (!s->got_streaminfo) {
464 ff_flacdsp_init(&s->dsp, s->avctx->sample_fmt, s->bps);
465 s->got_streaminfo = 1;
466 dump_headers(s->avctx, (FLACStreaminfo *)s);
469 // dump_headers(s->avctx, (FLACStreaminfo *)s);
472 for (i = 0; i < s->channels; i++) {
473 if (decode_subframe(s, i) < 0)
480 skip_bits(gb, 16); /* data crc */
485 static int flac_decode_frame(AVCodecContext *avctx, void *data,
486 int *got_frame_ptr, AVPacket *avpkt)
488 const uint8_t *buf = avpkt->data;
489 int buf_size = avpkt->size;
490 FLACContext *s = avctx->priv_data;
496 if (s->max_framesize == 0) {
498 ff_flac_get_max_frame_size(s->max_blocksize ? s->max_blocksize : FLAC_MAX_BLOCKSIZE,
499 FLAC_MAX_CHANNELS, 32);
502 /* check that there is at least the smallest decodable amount of data.
503 this amount corresponds to the smallest valid FLAC frame possible.
504 FF F8 69 02 00 00 9A 00 00 34 46 */
505 if (buf_size < FLAC_MIN_FRAME_SIZE)
508 /* check for inline header */
509 if (AV_RB32(buf) == MKBETAG('f','L','a','C')) {
510 if (!s->got_streaminfo && parse_streaminfo(s, buf, buf_size)) {
511 av_log(s->avctx, AV_LOG_ERROR, "invalid header\n");
514 return get_metadata_size(buf, buf_size);
518 init_get_bits(&s->gb, buf, buf_size*8);
519 if (decode_frame(s) < 0) {
520 av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
523 bytes_read = (get_bits_count(&s->gb)+7)/8;
525 /* get output buffer */
526 s->frame.nb_samples = s->blocksize;
527 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
528 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
532 s->dsp.decorrelate[s->ch_mode](s->frame.data, s->decoded, s->channels,
533 s->blocksize, s->sample_shift);
535 if (bytes_read > buf_size) {
536 av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", bytes_read - buf_size);
539 if (bytes_read < buf_size) {
540 av_log(s->avctx, AV_LOG_DEBUG, "underread: %d orig size: %d\n",
541 buf_size - bytes_read, buf_size);
545 *(AVFrame *)data = s->frame;
550 static av_cold int flac_decode_close(AVCodecContext *avctx)
552 FLACContext *s = avctx->priv_data;
555 for (i = 0; i < s->channels; i++) {
556 av_freep(&s->decoded[i]);
562 AVCodec ff_flac_decoder = {
564 .type = AVMEDIA_TYPE_AUDIO,
565 .id = AV_CODEC_ID_FLAC,
566 .priv_data_size = sizeof(FLACContext),
567 .init = flac_decode_init,
568 .close = flac_decode_close,
569 .decode = flac_decode_frame,
570 .capabilities = CODEC_CAP_DR1,
571 .long_name = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),
572 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16,