Mercurial > libavcodec.hg
comparison alsdec.c @ 10522:9b56cd1e37b7 libavcodec
Add MPEG-4 Audio Lossless Coding (ALS) decoder.
Patch by Thilo Borgmann, thilo D borgmann A googlemail
author | cehoyos |
---|---|
date | Wed, 11 Nov 2009 22:36:17 +0000 |
parents | |
children | af172182963b |
comparison
equal
deleted
inserted
replaced
10521:33890a382b36 | 10522:9b56cd1e37b7 |
---|---|
1 /* | |
2 * MPEG-4 ALS decoder | |
3 * Copyright (c) 2009 Thilo Borgmann <thilo.borgmann _at_ googlemail.com> | |
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 libavcodec/alsdec.c | |
24 * MPEG-4 ALS decoder | |
25 * @author Thilo Borgmann <thilo.borgmann _at_ googlemail.com> | |
26 */ | |
27 | |
28 | |
29 //#define DEBUG | |
30 | |
31 | |
32 #include "avcodec.h" | |
33 #include "get_bits.h" | |
34 #include "unary.h" | |
35 #include "mpeg4audio.h" | |
36 #include "bytestream.h" | |
37 | |
38 #include "als_data.h" | |
39 | |
40 enum RA_Flag { | |
41 RA_FLAG_NONE, | |
42 RA_FLAG_FRAMES, | |
43 RA_FLAG_HEADER | |
44 }; | |
45 | |
46 | |
47 typedef struct { | |
48 uint32_t samples; ///< number of samples, 0xFFFFFFFF if unknown | |
49 int resolution; ///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit | |
50 int floating; ///< 1 = IEEE 32-bit floating-point, 0 = integer | |
51 int frame_length; ///< frame length for each frame (last frame may differ) | |
52 int ra_distance; ///< distance between RA frames (in frames, 0...255) | |
53 enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored | |
54 int adapt_order; ///< adaptive order: 1 = on, 0 = off | |
55 int coef_table; ///< table index of Rice code parameters | |
56 int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off | |
57 int max_order; ///< maximum prediction order (0..1023) | |
58 int block_switching; ///< number of block switching levels | |
59 int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only) | |
60 int sb_part; ///< sub-block partition | |
61 int joint_stereo; ///< joint stereo: 1 = on, 0 = off | |
62 int mc_coding; ///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off | |
63 int chan_config; ///< indicates that a chan_config_info field is present | |
64 int chan_sort; ///< channel rearrangement: 1 = on, 0 = off | |
65 int rlslms; ///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off | |
66 int chan_config_info; ///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented. | |
67 int *chan_pos; ///< original channel positions | |
68 uint32_t header_size; ///< header size of original audio file in bytes, provided for debugging | |
69 uint32_t trailer_size; ///< trailer size of original audio file in bytes, provided for debugging | |
70 } ALSSpecificConfig; | |
71 | |
72 | |
73 typedef struct { | |
74 AVCodecContext *avctx; | |
75 ALSSpecificConfig sconf; | |
76 GetBitContext gb; | |
77 unsigned int cur_frame_length; ///< length of the current frame to decode | |
78 unsigned int frame_id; ///< the frame ID / number of the current frame | |
79 unsigned int js_switch; ///< if true, joint-stereo decoding is enforced | |
80 unsigned int num_blocks; ///< number of blocks used in the current frame | |
81 int32_t *quant_cof; ///< quantized parcor coefficients | |
82 int32_t *lpc_cof; ///< coefficients of the direct form prediction filter | |
83 int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block | |
84 int32_t **raw_samples; ///< decoded raw samples for each channel | |
85 int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples | |
86 } ALSDecContext; | |
87 | |
88 | |
89 static av_cold void dprint_specific_config(ALSDecContext *ctx) | |
90 { | |
91 #ifdef DEBUG | |
92 AVCodecContext *avctx = ctx->avctx; | |
93 ALSSpecificConfig *sconf = &ctx->sconf; | |
94 | |
95 dprintf(avctx, "resolution = %i\n", sconf->resolution); | |
96 dprintf(avctx, "floating = %i\n", sconf->floating); | |
97 dprintf(avctx, "frame_length = %i\n", sconf->frame_length); | |
98 dprintf(avctx, "ra_distance = %i\n", sconf->ra_distance); | |
99 dprintf(avctx, "ra_flag = %i\n", sconf->ra_flag); | |
100 dprintf(avctx, "adapt_order = %i\n", sconf->adapt_order); | |
101 dprintf(avctx, "coef_table = %i\n", sconf->coef_table); | |
102 dprintf(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction); | |
103 dprintf(avctx, "max_order = %i\n", sconf->max_order); | |
104 dprintf(avctx, "block_switching = %i\n", sconf->block_switching); | |
105 dprintf(avctx, "bgmc = %i\n", sconf->bgmc); | |
106 dprintf(avctx, "sb_part = %i\n", sconf->sb_part); | |
107 dprintf(avctx, "joint_stereo = %i\n", sconf->joint_stereo); | |
108 dprintf(avctx, "mc_coding = %i\n", sconf->mc_coding); | |
109 dprintf(avctx, "chan_config = %i\n", sconf->chan_config); | |
110 dprintf(avctx, "chan_sort = %i\n", sconf->chan_sort); | |
111 dprintf(avctx, "RLSLMS = %i\n", sconf->rlslms); | |
112 dprintf(avctx, "chan_config_info = %i\n", sconf->chan_config_info); | |
113 dprintf(avctx, "header_size = %i\n", sconf->header_size); | |
114 dprintf(avctx, "trailer_size = %i\n", sconf->trailer_size); | |
115 #endif | |
116 } | |
117 | |
118 | |
119 /** Reads an ALSSpecificConfig from a buffer into the output struct. | |
120 */ | |
121 static av_cold int read_specific_config(ALSDecContext *ctx) | |
122 { | |
123 GetBitContext gb; | |
124 uint64_t ht_size; | |
125 int i, config_offset, crc_enabled; | |
126 MPEG4AudioConfig m4ac; | |
127 ALSSpecificConfig *sconf = &ctx->sconf; | |
128 AVCodecContext *avctx = ctx->avctx; | |
129 uint32_t als_id; | |
130 | |
131 init_get_bits(&gb, avctx->extradata, avctx->extradata_size * 8); | |
132 | |
133 config_offset = ff_mpeg4audio_get_config(&m4ac, avctx->extradata, | |
134 avctx->extradata_size); | |
135 | |
136 if (config_offset < 0) | |
137 return -1; | |
138 | |
139 skip_bits_long(&gb, config_offset); | |
140 | |
141 if (get_bits_left(&gb) < (30 << 3)) | |
142 return -1; | |
143 | |
144 // read the fixed items | |
145 als_id = get_bits_long(&gb, 32); | |
146 avctx->sample_rate = m4ac.sample_rate; | |
147 skip_bits_long(&gb, 32); // sample rate already known | |
148 sconf->samples = get_bits_long(&gb, 32); | |
149 avctx->channels = m4ac.channels; | |
150 skip_bits(&gb, 16); // number of channels already knwon | |
151 skip_bits(&gb, 3); // skip file_type | |
152 sconf->resolution = get_bits(&gb, 3); | |
153 sconf->floating = get_bits1(&gb); | |
154 skip_bits1(&gb); // skip msb_first | |
155 sconf->frame_length = get_bits(&gb, 16) + 1; | |
156 sconf->ra_distance = get_bits(&gb, 8); | |
157 sconf->ra_flag = get_bits(&gb, 2); | |
158 sconf->adapt_order = get_bits1(&gb); | |
159 sconf->coef_table = get_bits(&gb, 2); | |
160 sconf->long_term_prediction = get_bits1(&gb); | |
161 sconf->max_order = get_bits(&gb, 10); | |
162 sconf->block_switching = get_bits(&gb, 2); | |
163 sconf->bgmc = get_bits1(&gb); | |
164 sconf->sb_part = get_bits1(&gb); | |
165 sconf->joint_stereo = get_bits1(&gb); | |
166 sconf->mc_coding = get_bits1(&gb); | |
167 sconf->chan_config = get_bits1(&gb); | |
168 sconf->chan_sort = get_bits1(&gb); | |
169 crc_enabled = get_bits1(&gb); | |
170 sconf->rlslms = get_bits1(&gb); | |
171 skip_bits(&gb, 5); // skip 5 reserved bits | |
172 skip_bits1(&gb); // skip aux_data_enabled | |
173 | |
174 | |
175 // check for ALSSpecificConfig struct | |
176 if (als_id != MKBETAG('A','L','S','\0')) | |
177 return -1; | |
178 | |
179 ctx->cur_frame_length = sconf->frame_length; | |
180 | |
181 // allocate quantized parcor coefficient buffer | |
182 if (!(ctx->quant_cof = av_malloc(sizeof(*ctx->quant_cof) * sconf->max_order)) || | |
183 !(ctx->lpc_cof = av_malloc(sizeof(*ctx->lpc_cof) * sconf->max_order))) { | |
184 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
185 return AVERROR(ENOMEM); | |
186 } | |
187 | |
188 // read channel config | |
189 if (sconf->chan_config) | |
190 sconf->chan_config_info = get_bits(&gb, 16); | |
191 // TODO: use this to set avctx->channel_layout | |
192 | |
193 | |
194 // read channel sorting | |
195 if (sconf->chan_sort && avctx->channels > 1) { | |
196 int chan_pos_bits = av_ceil_log2(avctx->channels); | |
197 int bits_needed = avctx->channels * chan_pos_bits + 7; | |
198 if (get_bits_left(&gb) < bits_needed) | |
199 return -1; | |
200 | |
201 if (!(sconf->chan_pos = av_malloc(avctx->channels * sizeof(*sconf->chan_pos)))) | |
202 return AVERROR(ENOMEM); | |
203 | |
204 for (i = 0; i < avctx->channels; i++) | |
205 sconf->chan_pos[i] = get_bits(&gb, chan_pos_bits); | |
206 | |
207 align_get_bits(&gb); | |
208 // TODO: use this to actually do channel sorting | |
209 } else { | |
210 sconf->chan_sort = 0; | |
211 } | |
212 | |
213 | |
214 // read fixed header and trailer sizes, | |
215 // if size = 0xFFFFFFFF then there is no data field! | |
216 if (get_bits_left(&gb) < 64) | |
217 return -1; | |
218 | |
219 sconf->header_size = get_bits_long(&gb, 32); | |
220 sconf->trailer_size = get_bits_long(&gb, 32); | |
221 if (sconf->header_size == 0xFFFFFFFF) | |
222 sconf->header_size = 0; | |
223 if (sconf->trailer_size == 0xFFFFFFFF) | |
224 sconf->trailer_size = 0; | |
225 | |
226 ht_size = ((int64_t)(sconf->header_size) + (int64_t)(sconf->trailer_size)) << 3; | |
227 | |
228 | |
229 // skip the header and trailer data | |
230 if (get_bits_left(&gb) < ht_size) | |
231 return -1; | |
232 | |
233 if (ht_size > INT32_MAX) | |
234 return -1; | |
235 | |
236 skip_bits_long(&gb, ht_size); | |
237 | |
238 | |
239 // skip the crc data | |
240 if (crc_enabled) { | |
241 if (get_bits_left(&gb) < 32) | |
242 return -1; | |
243 | |
244 skip_bits_long(&gb, 32); | |
245 } | |
246 | |
247 | |
248 // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data) | |
249 | |
250 dprint_specific_config(ctx); | |
251 | |
252 return 0; | |
253 } | |
254 | |
255 | |
256 /** Checks the ALSSpecificConfig for unsupported features. | |
257 */ | |
258 static int check_specific_config(ALSDecContext *ctx) | |
259 { | |
260 ALSSpecificConfig *sconf = &ctx->sconf; | |
261 int error = 0; | |
262 | |
263 // report unsupported feature and set error value | |
264 #define MISSING_ERR(cond, str, errval) \ | |
265 { \ | |
266 if (cond) { \ | |
267 av_log_missing_feature(ctx->avctx, str, 0); \ | |
268 error = errval; \ | |
269 } \ | |
270 } | |
271 | |
272 MISSING_ERR(sconf->floating, "Floating point decoding", -1); | |
273 MISSING_ERR(sconf->long_term_prediction, "Long-term prediction", -1); | |
274 MISSING_ERR(sconf->bgmc, "BGMC entropy decoding", -1); | |
275 MISSING_ERR(sconf->mc_coding, "Multi-channel correlation", -1); | |
276 MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", -1); | |
277 MISSING_ERR(sconf->chan_sort, "Channel sorting", 0); | |
278 | |
279 return error; | |
280 } | |
281 | |
282 | |
283 /** Parses the bs_info field to extract the block partitioning used in | |
284 * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2. | |
285 */ | |
286 static void parse_bs_info(const uint32_t bs_info, unsigned int n, | |
287 unsigned int div, unsigned int **div_blocks, | |
288 unsigned int *num_blocks) | |
289 { | |
290 if (n < 31 && ((bs_info << n) & 0x40000000)) { | |
291 // if the level is valid and the investigated bit n is set | |
292 // then recursively check both children at bits (2n+1) and (2n+2) | |
293 n *= 2; | |
294 div += 1; | |
295 parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks); | |
296 parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks); | |
297 } else { | |
298 // else the bit is not set or the last level has been reached | |
299 // (bit implicitly not set) | |
300 **div_blocks = div; | |
301 (*div_blocks)++; | |
302 (*num_blocks)++; | |
303 } | |
304 } | |
305 | |
306 | |
307 /** Reads and decodes a Rice codeword. | |
308 */ | |
309 static int32_t decode_rice(GetBitContext *gb, unsigned int k) | |
310 { | |
311 int max = gb->size_in_bits - get_bits_count(gb) - k; | |
312 int q = get_unary(gb, 0, max); | |
313 int r = k ? get_bits1(gb) : !(q & 1); | |
314 | |
315 if (k > 1) { | |
316 q <<= (k - 1); | |
317 q += get_bits_long(gb, k - 1); | |
318 } else if (!k) { | |
319 q >>= 1; | |
320 } | |
321 return r ? q : ~q; | |
322 } | |
323 | |
324 | |
325 /** Converts PARCOR coefficient k to direct filter coefficient. | |
326 */ | |
327 static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof) | |
328 { | |
329 int i, j; | |
330 | |
331 for (i = 0, j = k - 1; i < j; i++, j--) { | |
332 int tmp1 = ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); | |
333 cof[j] += ((MUL64(par[k], cof[i]) + (1 << 19)) >> 20); | |
334 cof[i] += tmp1; | |
335 } | |
336 if (i == j) | |
337 cof[i] += ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); | |
338 | |
339 cof[k] = par[k]; | |
340 } | |
341 | |
342 | |
343 /** Reads block switching field if necessary and sets actual block sizes. | |
344 * Also assures that the block sizes of the last frame correspond to the | |
345 * actual number of samples. | |
346 */ | |
347 static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks, | |
348 uint32_t *bs_info) | |
349 { | |
350 ALSSpecificConfig *sconf = &ctx->sconf; | |
351 GetBitContext *gb = &ctx->gb; | |
352 unsigned int *ptr_div_blocks = div_blocks; | |
353 unsigned int b; | |
354 | |
355 if (sconf->block_switching) { | |
356 unsigned int bs_info_len = 1 << (sconf->block_switching + 2); | |
357 *bs_info = get_bits_long(gb, bs_info_len); | |
358 *bs_info <<= (32 - bs_info_len); | |
359 } | |
360 | |
361 ctx->num_blocks = 0; | |
362 parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks); | |
363 | |
364 // The last frame may have an overdetermined block structure given in | |
365 // the bitstream. In that case the defined block structure would need | |
366 // more samples than available to be consistent. | |
367 // The block structure is actually used but the block sizes are adapted | |
368 // to fit the actual number of available samples. | |
369 // Example: 5 samples, 2nd level block sizes: 2 2 2 2. | |
370 // This results in the actual block sizes: 2 2 1 0. | |
371 // This is not specified in 14496-3 but actually done by the reference | |
372 // codec RM22 revision 2. | |
373 // This appears to happen in case of an odd number of samples in the last | |
374 // frame which is actually not allowed by the block length switching part | |
375 // of 14496-3. | |
376 // The ALS conformance files feature an odd number of samples in the last | |
377 // frame. | |
378 | |
379 for (b = 0; b < ctx->num_blocks; b++) | |
380 div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b]; | |
381 | |
382 if (ctx->cur_frame_length != ctx->sconf.frame_length) { | |
383 unsigned int remaining = ctx->cur_frame_length; | |
384 | |
385 for (b = 0; b < ctx->num_blocks; b++) { | |
386 if (remaining < div_blocks[b]) { | |
387 div_blocks[b] = remaining; | |
388 ctx->num_blocks = b + 1; | |
389 break; | |
390 } | |
391 | |
392 remaining -= div_blocks[b]; | |
393 } | |
394 } | |
395 } | |
396 | |
397 | |
398 /** Reads the block data for a constant block | |
399 */ | |
400 static void read_const_block(ALSDecContext *ctx, int32_t *raw_samples, | |
401 unsigned int block_length, unsigned int *js_blocks) | |
402 { | |
403 ALSSpecificConfig *sconf = &ctx->sconf; | |
404 AVCodecContext *avctx = ctx->avctx; | |
405 GetBitContext *gb = &ctx->gb; | |
406 int32_t const_val = 0; | |
407 unsigned int const_block, k; | |
408 | |
409 const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence) | |
410 *js_blocks = get_bits1(gb); | |
411 | |
412 // skip 5 reserved bits | |
413 skip_bits(gb, 5); | |
414 | |
415 if (const_block) { | |
416 unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample; | |
417 const_val = get_sbits_long(gb, const_val_bits); | |
418 } | |
419 | |
420 // write raw samples into buffer | |
421 for (k = 0; k < block_length; k++) | |
422 raw_samples[k] = const_val; | |
423 } | |
424 | |
425 | |
426 /** Reads the block data for a non-constant block | |
427 */ | |
428 static int read_var_block(ALSDecContext *ctx, unsigned int ra_block, | |
429 int32_t *raw_samples, unsigned int block_length, | |
430 unsigned int *js_blocks, int32_t *raw_other, | |
431 unsigned int *shift_lsbs) | |
432 { | |
433 ALSSpecificConfig *sconf = &ctx->sconf; | |
434 AVCodecContext *avctx = ctx->avctx; | |
435 GetBitContext *gb = &ctx->gb; | |
436 unsigned int k; | |
437 unsigned int s[8]; | |
438 unsigned int sub_blocks, log2_sub_blocks, sb_length; | |
439 unsigned int opt_order = 1; | |
440 int32_t *quant_cof = ctx->quant_cof; | |
441 int32_t *lpc_cof = ctx->lpc_cof; | |
442 unsigned int start = 0; | |
443 int smp = 0; | |
444 int sb, store_prev_samples; | |
445 int64_t y; | |
446 | |
447 *js_blocks = get_bits1(gb); | |
448 | |
449 // determine the number of subblocks for entropy decoding | |
450 if (!sconf->bgmc && !sconf->sb_part) { | |
451 log2_sub_blocks = 0; | |
452 } else { | |
453 if (sconf->bgmc && sconf->sb_part) | |
454 log2_sub_blocks = get_bits(gb, 2); | |
455 else | |
456 log2_sub_blocks = 2 * get_bits1(gb); | |
457 } | |
458 | |
459 sub_blocks = 1 << log2_sub_blocks; | |
460 | |
461 // do not continue in case of a damaged stream since | |
462 // block_length must be evenly divisible by sub_blocks | |
463 if (block_length & (sub_blocks - 1)) { | |
464 av_log(avctx, AV_LOG_WARNING, | |
465 "Block length is not evenly divisible by the number of subblocks.\n"); | |
466 return -1; | |
467 } | |
468 | |
469 sb_length = block_length >> log2_sub_blocks; | |
470 | |
471 | |
472 if (sconf->bgmc) { | |
473 // TODO: BGMC mode | |
474 } else { | |
475 s[0] = get_bits(gb, 4 + (sconf->resolution > 1)); | |
476 for (k = 1; k < sub_blocks; k++) | |
477 s[k] = s[k - 1] + decode_rice(gb, 0); | |
478 } | |
479 | |
480 if (get_bits1(gb)) | |
481 *shift_lsbs = get_bits(gb, 4) + 1; | |
482 | |
483 store_prev_samples = (*js_blocks && raw_other) || *shift_lsbs; | |
484 | |
485 | |
486 if (!sconf->rlslms) { | |
487 if (sconf->adapt_order) { | |
488 int opt_order_length = av_ceil_log2(av_clip((block_length >> 3) - 1, | |
489 2, sconf->max_order + 1)); | |
490 opt_order = get_bits(gb, opt_order_length); | |
491 } else { | |
492 opt_order = sconf->max_order; | |
493 } | |
494 | |
495 if (opt_order) { | |
496 int add_base; | |
497 | |
498 if (sconf->coef_table == 3) { | |
499 add_base = 0x7F; | |
500 | |
501 // read coefficient 0 | |
502 quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)]; | |
503 | |
504 // read coefficient 1 | |
505 if (opt_order > 1) | |
506 quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)]; | |
507 | |
508 // read coefficients 2 to opt_order | |
509 for (k = 2; k < opt_order; k++) | |
510 quant_cof[k] = get_bits(gb, 7); | |
511 } else { | |
512 int k_max; | |
513 add_base = 1; | |
514 | |
515 // read coefficient 0 to 19 | |
516 k_max = FFMIN(opt_order, 20); | |
517 for (k = 0; k < k_max; k++) { | |
518 int rice_param = parcor_rice_table[sconf->coef_table][k][1]; | |
519 int offset = parcor_rice_table[sconf->coef_table][k][0]; | |
520 quant_cof[k] = decode_rice(gb, rice_param) + offset; | |
521 } | |
522 | |
523 // read coefficients 20 to 126 | |
524 k_max = FFMIN(opt_order, 127); | |
525 for (; k < k_max; k++) | |
526 quant_cof[k] = decode_rice(gb, 2) + (k & 1); | |
527 | |
528 // read coefficients 127 to opt_order | |
529 for (; k < opt_order; k++) | |
530 quant_cof[k] = decode_rice(gb, 1); | |
531 | |
532 quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64]; | |
533 | |
534 if (opt_order > 1) | |
535 quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64]; | |
536 } | |
537 | |
538 for (k = 2; k < opt_order; k++) | |
539 quant_cof[k] = (quant_cof[k] << 14) + (add_base << 13); | |
540 } | |
541 } | |
542 | |
543 // TODO: LTP mode | |
544 | |
545 // read first value and residuals in case of a random access block | |
546 if (ra_block) { | |
547 if (opt_order) | |
548 raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4); | |
549 if (opt_order > 1) | |
550 raw_samples[1] = decode_rice(gb, s[0] + 3); | |
551 if (opt_order > 2) | |
552 raw_samples[2] = decode_rice(gb, s[0] + 1); | |
553 | |
554 start = FFMIN(opt_order, 3); | |
555 } | |
556 | |
557 // read all residuals | |
558 if (sconf->bgmc) { | |
559 // TODO: BGMC mode | |
560 } else { | |
561 int32_t *current_res = raw_samples + start; | |
562 | |
563 for (sb = 0; sb < sub_blocks; sb++, start = 0) | |
564 for (; start < sb_length; start++) | |
565 *current_res++ = decode_rice(gb, s[sb]); | |
566 } | |
567 | |
568 // reconstruct all samples from residuals | |
569 if (ra_block) { | |
570 for (smp = 0; smp < opt_order; smp++) { | |
571 y = 1 << 19; | |
572 | |
573 for (sb = 0; sb < smp; sb++) | |
574 y += MUL64(lpc_cof[sb],raw_samples[smp - (sb + 1)]); | |
575 | |
576 raw_samples[smp] -= y >> 20; | |
577 parcor_to_lpc(smp, quant_cof, lpc_cof); | |
578 } | |
579 } else { | |
580 for (k = 0; k < opt_order; k++) | |
581 parcor_to_lpc(k, quant_cof, lpc_cof); | |
582 | |
583 // store previous samples in case that they have to be altered | |
584 if (store_prev_samples) | |
585 memcpy(ctx->prev_raw_samples, raw_samples - sconf->max_order, | |
586 sizeof(*ctx->prev_raw_samples) * sconf->max_order); | |
587 | |
588 // reconstruct difference signal for prediction (joint-stereo) | |
589 if (*js_blocks && raw_other) { | |
590 int32_t *left, *right; | |
591 | |
592 if (raw_other > raw_samples) { // D = R - L | |
593 left = raw_samples; | |
594 right = raw_other; | |
595 } else { // D = R - L | |
596 left = raw_other; | |
597 right = raw_samples; | |
598 } | |
599 | |
600 for (sb = -1; sb >= -sconf->max_order; sb--) | |
601 raw_samples[sb] = right[sb] - left[sb]; | |
602 } | |
603 | |
604 // reconstruct shifted signal | |
605 if (*shift_lsbs) | |
606 for (sb = -1; sb >= -sconf->max_order; sb--) | |
607 raw_samples[sb] >>= *shift_lsbs; | |
608 } | |
609 | |
610 // reconstruct raw samples | |
611 for (; smp < block_length; smp++) { | |
612 y = 1 << 19; | |
613 | |
614 for (sb = 0; sb < opt_order; sb++) | |
615 y += MUL64(lpc_cof[sb],raw_samples[smp - (sb + 1)]); | |
616 | |
617 raw_samples[smp] -= y >> 20; | |
618 } | |
619 | |
620 // restore previous samples in case that they have been altered | |
621 if (store_prev_samples) | |
622 memcpy(raw_samples - sconf->max_order, ctx->prev_raw_samples, | |
623 sizeof(*raw_samples) * sconf->max_order); | |
624 | |
625 return 0; | |
626 } | |
627 | |
628 | |
629 /** Reads the block data. | |
630 */ | |
631 static int read_block_data(ALSDecContext *ctx, unsigned int ra_block, | |
632 int32_t *raw_samples, unsigned int block_length, | |
633 unsigned int *js_blocks, int32_t *raw_other) | |
634 { | |
635 ALSSpecificConfig *sconf = &ctx->sconf; | |
636 GetBitContext *gb = &ctx->gb; | |
637 unsigned int shift_lsbs = 0; | |
638 unsigned int k; | |
639 | |
640 // read block type flag and read the samples accordingly | |
641 if (get_bits1(gb)) { | |
642 if (read_var_block(ctx, ra_block, raw_samples, block_length, js_blocks, | |
643 raw_other, &shift_lsbs)) | |
644 return -1; | |
645 } else { | |
646 read_const_block(ctx, raw_samples, block_length, js_blocks); | |
647 } | |
648 | |
649 // TODO: read RLSLMS extension data | |
650 | |
651 if (!sconf->mc_coding || ctx->js_switch) | |
652 align_get_bits(gb); | |
653 | |
654 if (shift_lsbs) | |
655 for (k = 0; k < block_length; k++) | |
656 raw_samples[k] <<= shift_lsbs; | |
657 | |
658 return 0; | |
659 } | |
660 | |
661 | |
662 /** Computes the number of samples left to decode for the current frame and | |
663 * sets these samples to zero. | |
664 */ | |
665 static void zero_remaining(unsigned int b, unsigned int b_max, | |
666 const unsigned int *div_blocks, int32_t *buf) | |
667 { | |
668 unsigned int count = 0; | |
669 | |
670 while (b < b_max) | |
671 count += div_blocks[b]; | |
672 | |
673 memset(buf, 0, sizeof(*buf) * count); | |
674 } | |
675 | |
676 | |
677 /** Decodes blocks independently. | |
678 */ | |
679 static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame, | |
680 unsigned int c, const unsigned int *div_blocks, | |
681 unsigned int *js_blocks) | |
682 { | |
683 int32_t *raw_sample; | |
684 unsigned int b; | |
685 raw_sample = ctx->raw_samples[c]; | |
686 | |
687 for (b = 0; b < ctx->num_blocks; b++) { | |
688 if (read_block_data(ctx, ra_frame, raw_sample, | |
689 div_blocks[b], &js_blocks[0], NULL)) { | |
690 // damaged block, write zero for the rest of the frame | |
691 zero_remaining(b, ctx->num_blocks, div_blocks, raw_sample); | |
692 return -1; | |
693 } | |
694 raw_sample += div_blocks[b]; | |
695 ra_frame = 0; | |
696 } | |
697 | |
698 return 0; | |
699 } | |
700 | |
701 | |
702 /** Decodes blocks dependently. | |
703 */ | |
704 static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame, | |
705 unsigned int c, const unsigned int *div_blocks, | |
706 unsigned int *js_blocks) | |
707 { | |
708 ALSSpecificConfig *sconf = &ctx->sconf; | |
709 unsigned int offset = 0; | |
710 int32_t *raw_samples_R; | |
711 int32_t *raw_samples_L; | |
712 unsigned int b; | |
713 | |
714 // decode all blocks | |
715 for (b = 0; b < ctx->num_blocks; b++) { | |
716 unsigned int s; | |
717 raw_samples_L = ctx->raw_samples[c ] + offset; | |
718 raw_samples_R = ctx->raw_samples[c + 1] + offset; | |
719 if (read_block_data(ctx, ra_frame, raw_samples_L, div_blocks[b], | |
720 &js_blocks[0], raw_samples_R) || | |
721 read_block_data(ctx, ra_frame, raw_samples_R, div_blocks[b], | |
722 &js_blocks[1], raw_samples_L)) { | |
723 // damaged block, write zero for the rest of the frame | |
724 zero_remaining(b, ctx->num_blocks, div_blocks, raw_samples_L); | |
725 zero_remaining(b, ctx->num_blocks, div_blocks, raw_samples_R); | |
726 return -1; | |
727 } | |
728 | |
729 // reconstruct joint-stereo blocks | |
730 if (js_blocks[0]) { | |
731 if (js_blocks[1]) | |
732 av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair!\n"); | |
733 | |
734 for (s = 0; s < div_blocks[b]; s++) | |
735 raw_samples_L[s] = raw_samples_R[s] - raw_samples_L[s]; | |
736 } else if (js_blocks[1]) { | |
737 for (s = 0; s < div_blocks[b]; s++) | |
738 raw_samples_R[s] = raw_samples_R[s] + raw_samples_L[s]; | |
739 } | |
740 | |
741 offset += div_blocks[b]; | |
742 ra_frame = 0; | |
743 } | |
744 | |
745 // store carryover raw samples, | |
746 // the others channel raw samples are stored by the calling function. | |
747 memmove(ctx->raw_samples[c] - sconf->max_order, | |
748 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
749 sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
750 | |
751 return 0; | |
752 } | |
753 | |
754 | |
755 /** Reads the frame data. | |
756 */ | |
757 static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame) | |
758 { | |
759 ALSSpecificConfig *sconf = &ctx->sconf; | |
760 AVCodecContext *avctx = ctx->avctx; | |
761 GetBitContext *gb = &ctx->gb; | |
762 unsigned int div_blocks[32]; ///< block sizes. | |
763 unsigned int c; | |
764 unsigned int js_blocks[2]; | |
765 | |
766 uint32_t bs_info = 0; | |
767 | |
768 // skip the size of the ra unit if present in the frame | |
769 if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame) | |
770 skip_bits_long(gb, 32); | |
771 | |
772 if (sconf->mc_coding && sconf->joint_stereo) { | |
773 ctx->js_switch = get_bits1(gb); | |
774 align_get_bits(gb); | |
775 } | |
776 | |
777 if (!sconf->mc_coding || ctx->js_switch) { | |
778 int independent_bs = !sconf->joint_stereo; | |
779 | |
780 for (c = 0; c < avctx->channels; c++) { | |
781 js_blocks[0] = 0; | |
782 js_blocks[1] = 0; | |
783 | |
784 get_block_sizes(ctx, div_blocks, &bs_info); | |
785 | |
786 // if joint_stereo and block_switching is set, independent decoding | |
787 // is signaled via the first bit of bs_info | |
788 if (sconf->joint_stereo && sconf->block_switching) | |
789 if (bs_info >> 31) | |
790 independent_bs = 2; | |
791 | |
792 // if this is the last channel, it has to be decoded independently | |
793 if (c == avctx->channels - 1) | |
794 independent_bs = 1; | |
795 | |
796 if (independent_bs) { | |
797 if (decode_blocks_ind(ctx, ra_frame, c, div_blocks, js_blocks)) | |
798 return -1; | |
799 | |
800 independent_bs--; | |
801 } else { | |
802 if (decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks)) | |
803 return -1; | |
804 | |
805 c++; | |
806 } | |
807 | |
808 // store carryover raw samples | |
809 memmove(ctx->raw_samples[c] - sconf->max_order, | |
810 ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
811 sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
812 } | |
813 } else { // multi-channel coding | |
814 get_block_sizes(ctx, div_blocks, &bs_info); | |
815 | |
816 // TODO: multi channel coding might use a temporary buffer instead as | |
817 // the actual channel is not known when read_block-data is called | |
818 if (decode_blocks_ind(ctx, ra_frame, 0, div_blocks, js_blocks)) | |
819 return -1; | |
820 // TODO: read_channel_data | |
821 } | |
822 | |
823 // TODO: read_diff_float_data | |
824 | |
825 return 0; | |
826 } | |
827 | |
828 | |
829 /** Decodes an ALS frame. | |
830 */ | |
831 static int decode_frame(AVCodecContext *avctx, | |
832 void *data, int *data_size, | |
833 AVPacket *avpkt) | |
834 { | |
835 ALSDecContext *ctx = avctx->priv_data; | |
836 ALSSpecificConfig *sconf = &ctx->sconf; | |
837 const uint8_t *buffer = avpkt->data; | |
838 int buffer_size = avpkt->size; | |
839 int invalid_frame, size; | |
840 unsigned int c, sample, ra_frame, bytes_read, shift; | |
841 | |
842 init_get_bits(&ctx->gb, buffer, buffer_size * 8); | |
843 | |
844 // In the case that the distance between random access frames is set to zero | |
845 // (sconf->ra_distance == 0) no frame is treated as a random access frame. | |
846 // For the first frame, if prediction is used, all samples used from the | |
847 // previous frame are assumed to be zero. | |
848 ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance); | |
849 | |
850 // the last frame to decode might have a different length | |
851 if (sconf->samples != 0xFFFFFFFF) | |
852 ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length, | |
853 sconf->frame_length); | |
854 else | |
855 ctx->cur_frame_length = sconf->frame_length; | |
856 | |
857 // decode the frame data | |
858 if ((invalid_frame = read_frame_data(ctx, ra_frame) < 0)) | |
859 av_log(ctx->avctx, AV_LOG_WARNING, | |
860 "Reading frame data failed. Skipping RA unit.\n"); | |
861 | |
862 ctx->frame_id++; | |
863 | |
864 // check for size of decoded data | |
865 size = ctx->cur_frame_length * avctx->channels * | |
866 (av_get_bits_per_sample_format(avctx->sample_fmt) >> 3); | |
867 | |
868 if (size > *data_size) { | |
869 av_log(avctx, AV_LOG_ERROR, "Decoded data exceeds buffer size.\n"); | |
870 return -1; | |
871 } | |
872 | |
873 *data_size = size; | |
874 | |
875 // transform decoded frame into output format | |
876 #define INTERLEAVE_OUTPUT(bps) \ | |
877 { \ | |
878 int##bps##_t *dest = (int##bps##_t*) data; \ | |
879 shift = bps - ctx->avctx->bits_per_raw_sample; \ | |
880 for (sample = 0; sample < ctx->cur_frame_length; sample++) \ | |
881 for (c = 0; c < avctx->channels; c++) \ | |
882 *dest++ = ctx->raw_samples[c][sample] << shift; \ | |
883 } | |
884 | |
885 if (ctx->avctx->bits_per_raw_sample <= 16) { | |
886 INTERLEAVE_OUTPUT(16) | |
887 } else { | |
888 INTERLEAVE_OUTPUT(32) | |
889 } | |
890 | |
891 bytes_read = invalid_frame ? buffer_size : | |
892 (get_bits_count(&ctx->gb) + 7) >> 3; | |
893 | |
894 return bytes_read; | |
895 } | |
896 | |
897 | |
898 /** Uninitializes the ALS decoder. | |
899 */ | |
900 static av_cold int decode_end(AVCodecContext *avctx) | |
901 { | |
902 ALSDecContext *ctx = avctx->priv_data; | |
903 | |
904 av_freep(&ctx->sconf.chan_pos); | |
905 | |
906 av_freep(&ctx->quant_cof); | |
907 av_freep(&ctx->lpc_cof); | |
908 av_freep(&ctx->prev_raw_samples); | |
909 av_freep(&ctx->raw_samples); | |
910 av_freep(&ctx->raw_buffer); | |
911 | |
912 return 0; | |
913 } | |
914 | |
915 | |
916 /** Initializes the ALS decoder. | |
917 */ | |
918 static av_cold int decode_init(AVCodecContext *avctx) | |
919 { | |
920 unsigned int c; | |
921 unsigned int channel_size; | |
922 ALSDecContext *ctx = avctx->priv_data; | |
923 ALSSpecificConfig *sconf = &ctx->sconf; | |
924 ctx->avctx = avctx; | |
925 | |
926 if (!avctx->extradata) { | |
927 av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n"); | |
928 return -1; | |
929 } | |
930 | |
931 if (read_specific_config(ctx)) { | |
932 av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n"); | |
933 decode_end(avctx); | |
934 return -1; | |
935 } | |
936 | |
937 if (check_specific_config(ctx)) { | |
938 decode_end(avctx); | |
939 return -1; | |
940 } | |
941 | |
942 if (sconf->floating) { | |
943 avctx->sample_fmt = SAMPLE_FMT_FLT; | |
944 avctx->bits_per_raw_sample = 32; | |
945 } else { | |
946 avctx->sample_fmt = sconf->resolution > 1 | |
947 ? SAMPLE_FMT_S32 : SAMPLE_FMT_S16; | |
948 avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8; | |
949 } | |
950 | |
951 avctx->frame_size = sconf->frame_length; | |
952 channel_size = sconf->frame_length + sconf->max_order; | |
953 | |
954 ctx->prev_raw_samples = av_malloc (sizeof(*ctx->prev_raw_samples) * sconf->max_order); | |
955 ctx->raw_buffer = av_mallocz(sizeof(*ctx-> raw_buffer) * avctx->channels * channel_size); | |
956 ctx->raw_samples = av_malloc (sizeof(*ctx-> raw_samples) * avctx->channels); | |
957 | |
958 // allocate previous raw sample buffer | |
959 if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) { | |
960 av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
961 decode_end(avctx); | |
962 return AVERROR(ENOMEM); | |
963 } | |
964 | |
965 // assign raw samples buffers | |
966 ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order; | |
967 for (c = 1; c < avctx->channels; c++) | |
968 ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size; | |
969 | |
970 return 0; | |
971 } | |
972 | |
973 | |
974 /** Flushes (resets) the frame ID after seeking. | |
975 */ | |
976 static av_cold void flush(AVCodecContext *avctx) | |
977 { | |
978 ALSDecContext *ctx = avctx->priv_data; | |
979 | |
980 ctx->frame_id = 0; | |
981 } | |
982 | |
983 | |
984 AVCodec als_decoder = { | |
985 "als", | |
986 CODEC_TYPE_AUDIO, | |
987 CODEC_ID_MP4ALS, | |
988 sizeof(ALSDecContext), | |
989 decode_init, | |
990 NULL, | |
991 decode_end, | |
992 decode_frame, | |
993 .flush = flush, | |
994 .capabilities = CODEC_CAP_SUBFRAMES, | |
995 .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"), | |
996 }; | |
997 |