summaryrefslogtreecommitdiff
path: root/libavcodec/alac.c (plain)
blob: d6bd21ba1351ac5294bd2f48e616603386ec2e75
1/*
2 * ALAC (Apple Lossless Audio Codec) decoder
3 * Copyright (c) 2005 David Hammerton
4 *
5 * This file is part of FFmpeg.
6 *
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.
11 *
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.
16 *
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
20 */
21
22/**
23 * @file
24 * ALAC (Apple Lossless Audio Codec) decoder
25 * @author 2005 David Hammerton
26 * @see http://crazney.net/programs/itunes/alac.html
27 *
28 * Note: This decoder expects a 36-byte QuickTime atom to be
29 * passed through the extradata[_size] fields. This atom is tacked onto
30 * the end of an 'alac' stsd atom and has the following format:
31 *
32 * 32 bits atom size
33 * 32 bits tag ("alac")
34 * 32 bits tag version (0)
35 * 32 bits samples per frame (used when not set explicitly in the frames)
36 * 8 bits compatible version (0)
37 * 8 bits sample size
38 * 8 bits history mult (40)
39 * 8 bits initial history (10)
40 * 8 bits rice param limit (14)
41 * 8 bits channels
42 * 16 bits maxRun (255)
43 * 32 bits max coded frame size (0 means unknown)
44 * 32 bits average bitrate (0 means unknown)
45 * 32 bits samplerate
46 */
47
48#include <inttypes.h>
49
50#include "libavutil/channel_layout.h"
51#include "libavutil/opt.h"
52#include "avcodec.h"
53#include "get_bits.h"
54#include "bytestream.h"
55#include "internal.h"
56#include "thread.h"
57#include "unary.h"
58#include "mathops.h"
59#include "alac_data.h"
60#include "alacdsp.h"
61
62#define ALAC_EXTRADATA_SIZE 36
63
64typedef struct ALACContext {
65 AVClass *class;
66 AVCodecContext *avctx;
67 GetBitContext gb;
68 int channels;
69
70 int32_t *predict_error_buffer[2];
71 int32_t *output_samples_buffer[2];
72 int32_t *extra_bits_buffer[2];
73
74 uint32_t max_samples_per_frame;
75 uint8_t sample_size;
76 uint8_t rice_history_mult;
77 uint8_t rice_initial_history;
78 uint8_t rice_limit;
79 int sample_rate;
80
81 int extra_bits; /**< number of extra bits beyond 16-bit */
82 int nb_samples; /**< number of samples in the current frame */
83
84 int direct_output;
85 int extra_bit_bug;
86
87 ALACDSPContext dsp;
88} ALACContext;
89
90static inline unsigned int decode_scalar(GetBitContext *gb, int k, int bps)
91{
92 unsigned int x = get_unary_0_9(gb);
93
94 if (x > 8) { /* RICE THRESHOLD */
95 /* use alternative encoding */
96 x = get_bits_long(gb, bps);
97 } else if (k != 1) {
98 int extrabits = show_bits(gb, k);
99
100 /* multiply x by 2^k - 1, as part of their strange algorithm */
101 x = (x << k) - x;
102
103 if (extrabits > 1) {
104 x += extrabits - 1;
105 skip_bits(gb, k);
106 } else
107 skip_bits(gb, k - 1);
108 }
109 return x;
110}
111
112static int rice_decompress(ALACContext *alac, int32_t *output_buffer,
113 int nb_samples, int bps, int rice_history_mult)
114{
115 int i;
116 unsigned int history = alac->rice_initial_history;
117 int sign_modifier = 0;
118
119 for (i = 0; i < nb_samples; i++) {
120 int k;
121 unsigned int x;
122
123 if(get_bits_left(&alac->gb) <= 0)
124 return -1;
125
126 /* calculate rice param and decode next value */
127 k = av_log2((history >> 9) + 3);
128 k = FFMIN(k, alac->rice_limit);
129 x = decode_scalar(&alac->gb, k, bps);
130 x += sign_modifier;
131 sign_modifier = 0;
132 output_buffer[i] = (x >> 1) ^ -(x & 1);
133
134 /* update the history */
135 if (x > 0xffff)
136 history = 0xffff;
137 else
138 history += x * rice_history_mult -
139 ((history * rice_history_mult) >> 9);
140
141 /* special case: there may be compressed blocks of 0 */
142 if ((history < 128) && (i + 1 < nb_samples)) {
143 int block_size;
144
145 /* calculate rice param and decode block size */
146 k = 7 - av_log2(history) + ((history + 16) >> 6);
147 k = FFMIN(k, alac->rice_limit);
148 block_size = decode_scalar(&alac->gb, k, 16);
149
150 if (block_size > 0) {
151 if (block_size >= nb_samples - i) {
152 av_log(alac->avctx, AV_LOG_ERROR,
153 "invalid zero block size of %d %d %d\n", block_size,
154 nb_samples, i);
155 block_size = nb_samples - i - 1;
156 }
157 memset(&output_buffer[i + 1], 0,
158 block_size * sizeof(*output_buffer));
159 i += block_size;
160 }
161 if (block_size <= 0xffff)
162 sign_modifier = 1;
163 history = 0;
164 }
165 }
166 return 0;
167}
168
169static inline int sign_only(int v)
170{
171 return v ? FFSIGN(v) : 0;
172}
173
174static void lpc_prediction(int32_t *error_buffer, int32_t *buffer_out,
175 int nb_samples, int bps, int16_t *lpc_coefs,
176 int lpc_order, int lpc_quant)
177{
178 int i;
179 int32_t *pred = buffer_out;
180
181 /* first sample always copies */
182 *buffer_out = *error_buffer;
183
184 if (nb_samples <= 1)
185 return;
186
187 if (!lpc_order) {
188 memcpy(&buffer_out[1], &error_buffer[1],
189 (nb_samples - 1) * sizeof(*buffer_out));
190 return;
191 }
192
193 if (lpc_order == 31) {
194 /* simple 1st-order prediction */
195 for (i = 1; i < nb_samples; i++) {
196 buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i],
197 bps);
198 }
199 return;
200 }
201
202 /* read warm-up samples */
203 for (i = 1; i <= lpc_order && i < nb_samples; i++)
204 buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i], bps);
205
206 /* NOTE: 4 and 8 are very common cases that could be optimized. */
207
208 for (; i < nb_samples; i++) {
209 int j;
210 int val = 0;
211 int error_val = error_buffer[i];
212 int error_sign;
213 int d = *pred++;
214
215 /* LPC prediction */
216 for (j = 0; j < lpc_order; j++)
217 val += (pred[j] - d) * lpc_coefs[j];
218 val = (val + (1 << (lpc_quant - 1))) >> lpc_quant;
219 val += d + error_val;
220 buffer_out[i] = sign_extend(val, bps);
221
222 /* adapt LPC coefficients */
223 error_sign = sign_only(error_val);
224 if (error_sign) {
225 for (j = 0; j < lpc_order && error_val * error_sign > 0; j++) {
226 int sign;
227 val = d - pred[j];
228 sign = sign_only(val) * error_sign;
229 lpc_coefs[j] -= sign;
230 val *= sign;
231 error_val -= (val >> lpc_quant) * (j + 1);
232 }
233 }
234 }
235}
236
237static int decode_element(AVCodecContext *avctx, AVFrame *frame, int ch_index,
238 int channels)
239{
240 ALACContext *alac = avctx->priv_data;
241 int has_size, bps, is_compressed, decorr_shift, decorr_left_weight, ret;
242 uint32_t output_samples;
243 int i, ch;
244
245 skip_bits(&alac->gb, 4); /* element instance tag */
246 skip_bits(&alac->gb, 12); /* unused header bits */
247
248 /* the number of output samples is stored in the frame */
249 has_size = get_bits1(&alac->gb);
250
251 alac->extra_bits = get_bits(&alac->gb, 2) << 3;
252 bps = alac->sample_size - alac->extra_bits + channels - 1;
253 if (bps > 32U) {
254 avpriv_report_missing_feature(avctx, "bps %d", bps);
255 return AVERROR_PATCHWELCOME;
256 }
257
258 /* whether the frame is compressed */
259 is_compressed = !get_bits1(&alac->gb);
260
261 if (has_size)
262 output_samples = get_bits_long(&alac->gb, 32);
263 else
264 output_samples = alac->max_samples_per_frame;
265 if (!output_samples || output_samples > alac->max_samples_per_frame) {
266 av_log(avctx, AV_LOG_ERROR, "invalid samples per frame: %"PRIu32"\n",
267 output_samples);
268 return AVERROR_INVALIDDATA;
269 }
270 if (!alac->nb_samples) {
271 ThreadFrame tframe = { .f = frame };
272 /* get output buffer */
273 frame->nb_samples = output_samples;
274 if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
275 return ret;
276 } else if (output_samples != alac->nb_samples) {
277 av_log(avctx, AV_LOG_ERROR, "sample count mismatch: %"PRIu32" != %d\n",
278 output_samples, alac->nb_samples);
279 return AVERROR_INVALIDDATA;
280 }
281 alac->nb_samples = output_samples;
282 if (alac->direct_output) {
283 for (ch = 0; ch < channels; ch++)
284 alac->output_samples_buffer[ch] = (int32_t *)frame->extended_data[ch_index + ch];
285 }
286
287 if (is_compressed) {
288 int16_t lpc_coefs[2][32];
289 int lpc_order[2];
290 int prediction_type[2];
291 int lpc_quant[2];
292 int rice_history_mult[2];
293
294 if (!alac->rice_limit) {
295 avpriv_request_sample(alac->avctx,
296 "Compression with rice limit 0");
297 return AVERROR(ENOSYS);
298 }
299
300 decorr_shift = get_bits(&alac->gb, 8);
301 decorr_left_weight = get_bits(&alac->gb, 8);
302
303 for (ch = 0; ch < channels; ch++) {
304 prediction_type[ch] = get_bits(&alac->gb, 4);
305 lpc_quant[ch] = get_bits(&alac->gb, 4);
306 rice_history_mult[ch] = get_bits(&alac->gb, 3);
307 lpc_order[ch] = get_bits(&alac->gb, 5);
308
309 if (lpc_order[ch] >= alac->max_samples_per_frame)
310 return AVERROR_INVALIDDATA;
311
312 /* read the predictor table */
313 for (i = lpc_order[ch] - 1; i >= 0; i--)
314 lpc_coefs[ch][i] = get_sbits(&alac->gb, 16);
315 }
316
317 if (alac->extra_bits) {
318 for (i = 0; i < alac->nb_samples; i++) {
319 if(get_bits_left(&alac->gb) <= 0)
320 return -1;
321 for (ch = 0; ch < channels; ch++)
322 alac->extra_bits_buffer[ch][i] = get_bits(&alac->gb, alac->extra_bits);
323 }
324 }
325 for (ch = 0; ch < channels; ch++) {
326 int ret=rice_decompress(alac, alac->predict_error_buffer[ch],
327 alac->nb_samples, bps,
328 rice_history_mult[ch] * alac->rice_history_mult / 4);
329 if(ret<0)
330 return ret;
331
332 /* adaptive FIR filter */
333 if (prediction_type[ch] == 15) {
334 /* Prediction type 15 runs the adaptive FIR twice.
335 * The first pass uses the special-case coef_num = 31, while
336 * the second pass uses the coefs from the bitstream.
337 *
338 * However, this prediction type is not currently used by the
339 * reference encoder.
340 */
341 lpc_prediction(alac->predict_error_buffer[ch],
342 alac->predict_error_buffer[ch],
343 alac->nb_samples, bps, NULL, 31, 0);
344 } else if (prediction_type[ch] > 0) {
345 av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n",
346 prediction_type[ch]);
347 }
348 lpc_prediction(alac->predict_error_buffer[ch],
349 alac->output_samples_buffer[ch], alac->nb_samples,
350 bps, lpc_coefs[ch], lpc_order[ch], lpc_quant[ch]);
351 }
352 } else {
353 /* not compressed, easy case */
354 for (i = 0; i < alac->nb_samples; i++) {
355 if(get_bits_left(&alac->gb) <= 0)
356 return -1;
357 for (ch = 0; ch < channels; ch++) {
358 alac->output_samples_buffer[ch][i] =
359 get_sbits_long(&alac->gb, alac->sample_size);
360 }
361 }
362 alac->extra_bits = 0;
363 decorr_shift = 0;
364 decorr_left_weight = 0;
365 }
366
367 if (channels == 2) {
368 if (alac->extra_bits && alac->extra_bit_bug) {
369 alac->dsp.append_extra_bits[1](alac->output_samples_buffer, alac->extra_bits_buffer,
370 alac->extra_bits, channels, alac->nb_samples);
371 }
372
373 if (decorr_left_weight) {
374 alac->dsp.decorrelate_stereo(alac->output_samples_buffer, alac->nb_samples,
375 decorr_shift, decorr_left_weight);
376 }
377
378 if (alac->extra_bits && !alac->extra_bit_bug) {
379 alac->dsp.append_extra_bits[1](alac->output_samples_buffer, alac->extra_bits_buffer,
380 alac->extra_bits, channels, alac->nb_samples);
381 }
382 } else if (alac->extra_bits) {
383 alac->dsp.append_extra_bits[0](alac->output_samples_buffer, alac->extra_bits_buffer,
384 alac->extra_bits, channels, alac->nb_samples);
385 }
386
387 switch(alac->sample_size) {
388 case 16: {
389 for (ch = 0; ch < channels; ch++) {
390 int16_t *outbuffer = (int16_t *)frame->extended_data[ch_index + ch];
391 for (i = 0; i < alac->nb_samples; i++)
392 *outbuffer++ = alac->output_samples_buffer[ch][i];
393 }}
394 break;
395 case 20: {
396 for (ch = 0; ch < channels; ch++) {
397 for (i = 0; i < alac->nb_samples; i++)
398 alac->output_samples_buffer[ch][i] <<= 12;
399 }}
400 break;
401 case 24: {
402 for (ch = 0; ch < channels; ch++) {
403 for (i = 0; i < alac->nb_samples; i++)
404 alac->output_samples_buffer[ch][i] <<= 8;
405 }}
406 break;
407 }
408
409 return 0;
410}
411
412static int alac_decode_frame(AVCodecContext *avctx, void *data,
413 int *got_frame_ptr, AVPacket *avpkt)
414{
415 ALACContext *alac = avctx->priv_data;
416 AVFrame *frame = data;
417 enum AlacRawDataBlockType element;
418 int channels;
419 int ch, ret, got_end;
420
421 if ((ret = init_get_bits8(&alac->gb, avpkt->data, avpkt->size)) < 0)
422 return ret;
423
424 got_end = 0;
425 alac->nb_samples = 0;
426 ch = 0;
427 while (get_bits_left(&alac->gb) >= 3) {
428 element = get_bits(&alac->gb, 3);
429 if (element == TYPE_END) {
430 got_end = 1;
431 break;
432 }
433 if (element > TYPE_CPE && element != TYPE_LFE) {
434 avpriv_report_missing_feature(avctx, "Syntax element %d", element);
435 return AVERROR_PATCHWELCOME;
436 }
437
438 channels = (element == TYPE_CPE) ? 2 : 1;
439 if (ch + channels > alac->channels ||
440 ff_alac_channel_layout_offsets[alac->channels - 1][ch] + channels > alac->channels) {
441 av_log(avctx, AV_LOG_ERROR, "invalid element channel count\n");
442 return AVERROR_INVALIDDATA;
443 }
444
445 ret = decode_element(avctx, frame,
446 ff_alac_channel_layout_offsets[alac->channels - 1][ch],
447 channels);
448 if (ret < 0 && get_bits_left(&alac->gb))
449 return ret;
450
451 ch += channels;
452 }
453 if (!got_end) {
454 av_log(avctx, AV_LOG_ERROR, "no end tag found. incomplete packet.\n");
455 return AVERROR_INVALIDDATA;
456 }
457
458 if (avpkt->size * 8 - get_bits_count(&alac->gb) > 8) {
459 av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n",
460 avpkt->size * 8 - get_bits_count(&alac->gb));
461 }
462
463 if (alac->channels == ch && alac->nb_samples)
464 *got_frame_ptr = 1;
465 else
466 av_log(avctx, AV_LOG_WARNING, "Failed to decode all channels\n");
467
468 return avpkt->size;
469}
470
471static av_cold int alac_decode_close(AVCodecContext *avctx)
472{
473 ALACContext *alac = avctx->priv_data;
474
475 int ch;
476 for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
477 av_freep(&alac->predict_error_buffer[ch]);
478 if (!alac->direct_output)
479 av_freep(&alac->output_samples_buffer[ch]);
480 av_freep(&alac->extra_bits_buffer[ch]);
481 }
482
483 return 0;
484}
485
486static int allocate_buffers(ALACContext *alac)
487{
488 int ch;
489 int buf_size = alac->max_samples_per_frame * sizeof(int32_t);
490
491 for (ch = 0; ch < 2; ch++) {
492 alac->predict_error_buffer[ch] = NULL;
493 alac->output_samples_buffer[ch] = NULL;
494 alac->extra_bits_buffer[ch] = NULL;
495 }
496
497 for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
498 FF_ALLOC_OR_GOTO(alac->avctx, alac->predict_error_buffer[ch],
499 buf_size, buf_alloc_fail);
500
501 alac->direct_output = alac->sample_size > 16;
502 if (!alac->direct_output) {
503 FF_ALLOC_OR_GOTO(alac->avctx, alac->output_samples_buffer[ch],
504 buf_size + AV_INPUT_BUFFER_PADDING_SIZE, buf_alloc_fail);
505 }
506
507 FF_ALLOC_OR_GOTO(alac->avctx, alac->extra_bits_buffer[ch],
508 buf_size + AV_INPUT_BUFFER_PADDING_SIZE, buf_alloc_fail);
509 }
510 return 0;
511buf_alloc_fail:
512 alac_decode_close(alac->avctx);
513 return AVERROR(ENOMEM);
514}
515
516static int alac_set_info(ALACContext *alac)
517{
518 GetByteContext gb;
519
520 bytestream2_init(&gb, alac->avctx->extradata,
521 alac->avctx->extradata_size);
522
523 bytestream2_skipu(&gb, 12); // size:4, alac:4, version:4
524
525 alac->max_samples_per_frame = bytestream2_get_be32u(&gb);
526 if (!alac->max_samples_per_frame ||
527 alac->max_samples_per_frame > INT_MAX / sizeof(int32_t)) {
528 av_log(alac->avctx, AV_LOG_ERROR,
529 "max samples per frame invalid: %"PRIu32"\n",
530 alac->max_samples_per_frame);
531 return AVERROR_INVALIDDATA;
532 }
533 bytestream2_skipu(&gb, 1); // compatible version
534 alac->sample_size = bytestream2_get_byteu(&gb);
535 alac->rice_history_mult = bytestream2_get_byteu(&gb);
536 alac->rice_initial_history = bytestream2_get_byteu(&gb);
537 alac->rice_limit = bytestream2_get_byteu(&gb);
538 alac->channels = bytestream2_get_byteu(&gb);
539 bytestream2_get_be16u(&gb); // maxRun
540 bytestream2_get_be32u(&gb); // max coded frame size
541 bytestream2_get_be32u(&gb); // average bitrate
542 alac->sample_rate = bytestream2_get_be32u(&gb);
543
544 return 0;
545}
546
547static av_cold int alac_decode_init(AVCodecContext * avctx)
548{
549 int ret;
550 ALACContext *alac = avctx->priv_data;
551 alac->avctx = avctx;
552
553 /* initialize from the extradata */
554 if (alac->avctx->extradata_size < ALAC_EXTRADATA_SIZE) {
555 av_log(avctx, AV_LOG_ERROR, "extradata is too small\n");
556 return AVERROR_INVALIDDATA;
557 }
558 if (alac_set_info(alac)) {
559 av_log(avctx, AV_LOG_ERROR, "set_info failed\n");
560 return -1;
561 }
562
563 switch (alac->sample_size) {
564 case 16: avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
565 break;
566 case 20:
567 case 24:
568 case 32: avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
569 break;
570 default: avpriv_request_sample(avctx, "Sample depth %d", alac->sample_size);
571 return AVERROR_PATCHWELCOME;
572 }
573 avctx->bits_per_raw_sample = alac->sample_size;
574 avctx->sample_rate = alac->sample_rate;
575
576 if (alac->channels < 1) {
577 av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n");
578 alac->channels = avctx->channels;
579 } else {
580 if (alac->channels > ALAC_MAX_CHANNELS)
581 alac->channels = avctx->channels;
582 else
583 avctx->channels = alac->channels;
584 }
585 if (avctx->channels > ALAC_MAX_CHANNELS || avctx->channels <= 0 ) {
586 avpriv_report_missing_feature(avctx, "Channel count %d",
587 avctx->channels);
588 return AVERROR_PATCHWELCOME;
589 }
590 avctx->channel_layout = ff_alac_channel_layouts[alac->channels - 1];
591
592 if ((ret = allocate_buffers(alac)) < 0) {
593 av_log(avctx, AV_LOG_ERROR, "Error allocating buffers\n");
594 return ret;
595 }
596
597 ff_alacdsp_init(&alac->dsp);
598
599 return 0;
600}
601
602#if HAVE_THREADS
603static int init_thread_copy(AVCodecContext *avctx)
604{
605 ALACContext *alac = avctx->priv_data;
606 alac->avctx = avctx;
607 return allocate_buffers(alac);
608}
609#endif
610
611static const AVOption options[] = {
612 { "extra_bits_bug", "Force non-standard decoding process",
613 offsetof(ALACContext, extra_bit_bug), AV_OPT_TYPE_BOOL, { .i64 = 0 },
614 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM },
615 { NULL },
616};
617
618static const AVClass alac_class = {
619 .class_name = "alac",
620 .item_name = av_default_item_name,
621 .option = options,
622 .version = LIBAVUTIL_VERSION_INT,
623};
624
625AVCodec ff_alac_decoder = {
626 .name = "alac",
627 .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
628 .type = AVMEDIA_TYPE_AUDIO,
629 .id = AV_CODEC_ID_ALAC,
630 .priv_data_size = sizeof(ALACContext),
631 .init = alac_decode_init,
632 .close = alac_decode_close,
633 .decode = alac_decode_frame,
634 .init_thread_copy = ONLY_IF_THREADS_ENABLED(init_thread_copy),
635 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS,
636 .priv_class = &alac_class
637};
638
generated by cgit at 2024-05-17 06:34:01 (GMT)