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
64 typedef 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
decode_scalar(GetBitContext * gb,int k,int bps)90 static 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
rice_decompress(ALACContext * alac,int32_t * output_buffer,int nb_samples,int bps,int rice_history_mult)112 static 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 AVERROR_INVALIDDATA;
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
sign_only(int v)169 static inline int sign_only(int v)
170 {
171 return v ? FFSIGN(v) : 0;
172 }
173
lpc_prediction(int32_t * error_buffer,uint32_t * buffer_out,int nb_samples,int bps,int16_t * lpc_coefs,int lpc_order,int lpc_quant)174 static void lpc_prediction(int32_t *error_buffer, uint32_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 uint32_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 unsigned 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 + (1LL << (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 && (int)(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 *= (unsigned)sign;
231 error_val -= (val >> lpc_quant) * (j + 1U);
232 }
233 }
234 }
235 }
236
decode_element(AVCodecContext * avctx,AVFrame * frame,int ch_index,int channels)237 static 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 > 32) {
254 avpriv_report_missing_feature(avctx, "bps %d", bps);
255 return AVERROR_PATCHWELCOME;
256 }
257 if (bps < 1)
258 return AVERROR_INVALIDDATA;
259
260 /* whether the frame is compressed */
261 is_compressed = !get_bits1(&alac->gb);
262
263 if (has_size)
264 output_samples = get_bits_long(&alac->gb, 32);
265 else
266 output_samples = alac->max_samples_per_frame;
267 if (!output_samples || output_samples > alac->max_samples_per_frame) {
268 av_log(avctx, AV_LOG_ERROR, "invalid samples per frame: %"PRIu32"\n",
269 output_samples);
270 return AVERROR_INVALIDDATA;
271 }
272 if (!alac->nb_samples) {
273 ThreadFrame tframe = { .f = frame };
274 /* get output buffer */
275 frame->nb_samples = output_samples;
276 if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0)
277 return ret;
278 } else if (output_samples != alac->nb_samples) {
279 av_log(avctx, AV_LOG_ERROR, "sample count mismatch: %"PRIu32" != %d\n",
280 output_samples, alac->nb_samples);
281 return AVERROR_INVALIDDATA;
282 }
283 alac->nb_samples = output_samples;
284 if (alac->direct_output) {
285 for (ch = 0; ch < channels; ch++)
286 alac->output_samples_buffer[ch] = (int32_t *)frame->extended_data[ch_index + ch];
287 }
288
289 if (is_compressed) {
290 int16_t lpc_coefs[2][32];
291 int lpc_order[2];
292 int prediction_type[2];
293 int lpc_quant[2];
294 int rice_history_mult[2];
295
296 if (!alac->rice_limit) {
297 avpriv_request_sample(alac->avctx,
298 "Compression with rice limit 0");
299 return AVERROR(ENOSYS);
300 }
301
302 decorr_shift = get_bits(&alac->gb, 8);
303 decorr_left_weight = get_bits(&alac->gb, 8);
304
305 if (channels == 2 && decorr_left_weight && decorr_shift > 31)
306 return AVERROR_INVALIDDATA;
307
308 for (ch = 0; ch < channels; ch++) {
309 prediction_type[ch] = get_bits(&alac->gb, 4);
310 lpc_quant[ch] = get_bits(&alac->gb, 4);
311 rice_history_mult[ch] = get_bits(&alac->gb, 3);
312 lpc_order[ch] = get_bits(&alac->gb, 5);
313
314 if (lpc_order[ch] >= alac->max_samples_per_frame || !lpc_quant[ch])
315 return AVERROR_INVALIDDATA;
316
317 /* read the predictor table */
318 for (i = lpc_order[ch] - 1; i >= 0; i--)
319 lpc_coefs[ch][i] = get_sbits(&alac->gb, 16);
320 }
321
322 if (alac->extra_bits) {
323 for (i = 0; i < alac->nb_samples; i++) {
324 if(get_bits_left(&alac->gb) <= 0)
325 return AVERROR_INVALIDDATA;
326 for (ch = 0; ch < channels; ch++)
327 alac->extra_bits_buffer[ch][i] = get_bits(&alac->gb, alac->extra_bits);
328 }
329 }
330 for (ch = 0; ch < channels; ch++) {
331 int ret=rice_decompress(alac, alac->predict_error_buffer[ch],
332 alac->nb_samples, bps,
333 rice_history_mult[ch] * alac->rice_history_mult / 4);
334 if(ret<0)
335 return ret;
336
337 /* adaptive FIR filter */
338 if (prediction_type[ch] == 15) {
339 /* Prediction type 15 runs the adaptive FIR twice.
340 * The first pass uses the special-case coef_num = 31, while
341 * the second pass uses the coefs from the bitstream.
342 *
343 * However, this prediction type is not currently used by the
344 * reference encoder.
345 */
346 lpc_prediction(alac->predict_error_buffer[ch],
347 alac->predict_error_buffer[ch],
348 alac->nb_samples, bps, NULL, 31, 0);
349 } else if (prediction_type[ch] > 0) {
350 av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n",
351 prediction_type[ch]);
352 }
353 lpc_prediction(alac->predict_error_buffer[ch],
354 alac->output_samples_buffer[ch], alac->nb_samples,
355 bps, lpc_coefs[ch], lpc_order[ch], lpc_quant[ch]);
356 }
357 } else {
358 /* not compressed, easy case */
359 for (i = 0; i < alac->nb_samples; i++) {
360 if(get_bits_left(&alac->gb) <= 0)
361 return AVERROR_INVALIDDATA;
362 for (ch = 0; ch < channels; ch++) {
363 alac->output_samples_buffer[ch][i] =
364 get_sbits_long(&alac->gb, alac->sample_size);
365 }
366 }
367 alac->extra_bits = 0;
368 decorr_shift = 0;
369 decorr_left_weight = 0;
370 }
371
372 if (channels == 2) {
373 if (alac->extra_bits && alac->extra_bit_bug) {
374 alac->dsp.append_extra_bits[1](alac->output_samples_buffer, alac->extra_bits_buffer,
375 alac->extra_bits, channels, alac->nb_samples);
376 }
377
378 if (decorr_left_weight) {
379 alac->dsp.decorrelate_stereo(alac->output_samples_buffer, alac->nb_samples,
380 decorr_shift, decorr_left_weight);
381 }
382
383 if (alac->extra_bits && !alac->extra_bit_bug) {
384 alac->dsp.append_extra_bits[1](alac->output_samples_buffer, alac->extra_bits_buffer,
385 alac->extra_bits, channels, alac->nb_samples);
386 }
387 } else if (alac->extra_bits) {
388 alac->dsp.append_extra_bits[0](alac->output_samples_buffer, alac->extra_bits_buffer,
389 alac->extra_bits, channels, alac->nb_samples);
390 }
391
392 switch(alac->sample_size) {
393 case 16: {
394 for (ch = 0; ch < channels; ch++) {
395 int16_t *outbuffer = (int16_t *)frame->extended_data[ch_index + ch];
396 for (i = 0; i < alac->nb_samples; i++)
397 *outbuffer++ = alac->output_samples_buffer[ch][i];
398 }}
399 break;
400 case 20: {
401 for (ch = 0; ch < channels; ch++) {
402 for (i = 0; i < alac->nb_samples; i++)
403 alac->output_samples_buffer[ch][i] *= 1U << 12;
404 }}
405 break;
406 case 24: {
407 for (ch = 0; ch < channels; ch++) {
408 for (i = 0; i < alac->nb_samples; i++)
409 alac->output_samples_buffer[ch][i] *= 1U << 8;
410 }}
411 break;
412 }
413
414 return 0;
415 }
416
alac_decode_frame(AVCodecContext * avctx,void * data,int * got_frame_ptr,AVPacket * avpkt)417 static int alac_decode_frame(AVCodecContext *avctx, void *data,
418 int *got_frame_ptr, AVPacket *avpkt)
419 {
420 ALACContext *alac = avctx->priv_data;
421 AVFrame *frame = data;
422 enum AlacRawDataBlockType element;
423 int channels;
424 int ch, ret, got_end;
425
426 if ((ret = init_get_bits8(&alac->gb, avpkt->data, avpkt->size)) < 0)
427 return ret;
428
429 got_end = 0;
430 alac->nb_samples = 0;
431 ch = 0;
432 while (get_bits_left(&alac->gb) >= 3) {
433 element = get_bits(&alac->gb, 3);
434 if (element == TYPE_END) {
435 got_end = 1;
436 break;
437 }
438 if (element > TYPE_CPE && element != TYPE_LFE) {
439 avpriv_report_missing_feature(avctx, "Syntax element %d", element);
440 return AVERROR_PATCHWELCOME;
441 }
442
443 channels = (element == TYPE_CPE) ? 2 : 1;
444 if (ch + channels > alac->channels ||
445 ff_alac_channel_layout_offsets[alac->channels - 1][ch] + channels > alac->channels) {
446 av_log(avctx, AV_LOG_ERROR, "invalid element channel count\n");
447 return AVERROR_INVALIDDATA;
448 }
449
450 ret = decode_element(avctx, frame,
451 ff_alac_channel_layout_offsets[alac->channels - 1][ch],
452 channels);
453 if (ret < 0 && get_bits_left(&alac->gb))
454 return ret;
455
456 ch += channels;
457 }
458 if (!got_end) {
459 av_log(avctx, AV_LOG_ERROR, "no end tag found. incomplete packet.\n");
460 return AVERROR_INVALIDDATA;
461 }
462
463 if (avpkt->size * 8 - get_bits_count(&alac->gb) > 8) {
464 av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n",
465 avpkt->size * 8 - get_bits_count(&alac->gb));
466 }
467
468 if (alac->channels == ch && alac->nb_samples)
469 *got_frame_ptr = 1;
470 else
471 av_log(avctx, AV_LOG_WARNING, "Failed to decode all channels\n");
472
473 return avpkt->size;
474 }
475
alac_decode_close(AVCodecContext * avctx)476 static av_cold int alac_decode_close(AVCodecContext *avctx)
477 {
478 ALACContext *alac = avctx->priv_data;
479
480 int ch;
481 for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
482 av_freep(&alac->predict_error_buffer[ch]);
483 if (!alac->direct_output)
484 av_freep(&alac->output_samples_buffer[ch]);
485 av_freep(&alac->extra_bits_buffer[ch]);
486 }
487
488 return 0;
489 }
490
allocate_buffers(ALACContext * alac)491 static int allocate_buffers(ALACContext *alac)
492 {
493 int ch;
494 unsigned buf_size = alac->max_samples_per_frame * sizeof(int32_t);
495 unsigned extra_buf_size = buf_size + AV_INPUT_BUFFER_PADDING_SIZE;
496
497 for (ch = 0; ch < 2; ch++) {
498 alac->predict_error_buffer[ch] = NULL;
499 alac->output_samples_buffer[ch] = NULL;
500 alac->extra_bits_buffer[ch] = NULL;
501 }
502
503 for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
504 if (!(alac->predict_error_buffer[ch] = av_malloc(buf_size)))
505 return AVERROR(ENOMEM);
506
507 alac->direct_output = alac->sample_size > 16;
508 if (!alac->direct_output) {
509 if (!(alac->output_samples_buffer[ch] = av_malloc(extra_buf_size)))
510 return AVERROR(ENOMEM);
511 }
512
513 if (!(alac->extra_bits_buffer[ch] = av_malloc(extra_buf_size)))
514 return AVERROR(ENOMEM);
515 }
516 return 0;
517 }
518
alac_set_info(ALACContext * alac)519 static int alac_set_info(ALACContext *alac)
520 {
521 GetByteContext gb;
522
523 bytestream2_init(&gb, alac->avctx->extradata,
524 alac->avctx->extradata_size);
525
526 bytestream2_skipu(&gb, 12); // size:4, alac:4, version:4
527
528 alac->max_samples_per_frame = bytestream2_get_be32u(&gb);
529 if (!alac->max_samples_per_frame ||
530 alac->max_samples_per_frame > 4096 * 4096) {
531 av_log(alac->avctx, AV_LOG_ERROR,
532 "max samples per frame invalid: %"PRIu32"\n",
533 alac->max_samples_per_frame);
534 return AVERROR_INVALIDDATA;
535 }
536 bytestream2_skipu(&gb, 1); // compatible version
537 alac->sample_size = bytestream2_get_byteu(&gb);
538 alac->rice_history_mult = bytestream2_get_byteu(&gb);
539 alac->rice_initial_history = bytestream2_get_byteu(&gb);
540 alac->rice_limit = bytestream2_get_byteu(&gb);
541 alac->channels = bytestream2_get_byteu(&gb);
542 bytestream2_get_be16u(&gb); // maxRun
543 bytestream2_get_be32u(&gb); // max coded frame size
544 bytestream2_get_be32u(&gb); // average bitrate
545 alac->sample_rate = bytestream2_get_be32u(&gb);
546
547 return 0;
548 }
549
alac_decode_init(AVCodecContext * avctx)550 static av_cold int alac_decode_init(AVCodecContext * avctx)
551 {
552 int ret;
553 ALACContext *alac = avctx->priv_data;
554 alac->avctx = avctx;
555
556 /* initialize from the extradata */
557 if (alac->avctx->extradata_size < ALAC_EXTRADATA_SIZE) {
558 av_log(avctx, AV_LOG_ERROR, "extradata is too small\n");
559 return AVERROR_INVALIDDATA;
560 }
561 if ((ret = alac_set_info(alac)) < 0) {
562 av_log(avctx, AV_LOG_ERROR, "set_info failed\n");
563 return ret;
564 }
565
566 switch (alac->sample_size) {
567 case 16: avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
568 break;
569 case 20:
570 case 24:
571 case 32: avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
572 break;
573 default: avpriv_request_sample(avctx, "Sample depth %d", alac->sample_size);
574 return AVERROR_PATCHWELCOME;
575 }
576 avctx->bits_per_raw_sample = alac->sample_size;
577 avctx->sample_rate = alac->sample_rate;
578
579 if (alac->channels < 1) {
580 av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n");
581 alac->channels = avctx->channels;
582 } else {
583 if (alac->channels > ALAC_MAX_CHANNELS)
584 alac->channels = avctx->channels;
585 else
586 avctx->channels = alac->channels;
587 }
588 if (avctx->channels > ALAC_MAX_CHANNELS || avctx->channels <= 0 ) {
589 avpriv_report_missing_feature(avctx, "Channel count %d",
590 avctx->channels);
591 return AVERROR_PATCHWELCOME;
592 }
593 avctx->channel_layout = ff_alac_channel_layouts[alac->channels - 1];
594
595 if ((ret = allocate_buffers(alac)) < 0) {
596 av_log(avctx, AV_LOG_ERROR, "Error allocating buffers\n");
597 return ret;
598 }
599
600 ff_alacdsp_init(&alac->dsp);
601
602 return 0;
603 }
604
605 static const AVOption options[] = {
606 { "extra_bits_bug", "Force non-standard decoding process",
607 offsetof(ALACContext, extra_bit_bug), AV_OPT_TYPE_BOOL, { .i64 = 0 },
608 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM },
609 { NULL },
610 };
611
612 static const AVClass alac_class = {
613 .class_name = "alac",
614 .item_name = av_default_item_name,
615 .option = options,
616 .version = LIBAVUTIL_VERSION_INT,
617 };
618
619 AVCodec ff_alac_decoder = {
620 .name = "alac",
621 .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
622 .type = AVMEDIA_TYPE_AUDIO,
623 .id = AV_CODEC_ID_ALAC,
624 .priv_data_size = sizeof(ALACContext),
625 .init = alac_decode_init,
626 .close = alac_decode_close,
627 .decode = alac_decode_frame,
628 .capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_FRAME_THREADS | AV_CODEC_CAP_CHANNEL_CONF,
629 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
630 .priv_class = &alac_class
631 };
632