Commit | Line | Data |
---|---|---|
99971952 TB |
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 | ||
25c4fdda TB |
38 | #include <stdint.h> |
39 | ||
40 | /** Rice parameters and corresponding index offsets for decoding the | |
41 | * indices of scaled PARCOR values. The table choosen is set globally | |
42 | * by the encoder and stored in ALSSpecificConfig. | |
43 | */ | |
44 | static const int8_t parcor_rice_table[3][20][2] = { | |
45 | { {-52, 4}, {-29, 5}, {-31, 4}, { 19, 4}, {-16, 4}, | |
46 | { 12, 3}, { -7, 3}, { 9, 3}, { -5, 3}, { 6, 3}, | |
47 | { -4, 3}, { 3, 3}, { -3, 2}, { 3, 2}, { -2, 2}, | |
48 | { 3, 2}, { -1, 2}, { 2, 2}, { -1, 2}, { 2, 2} }, | |
49 | { {-58, 3}, {-42, 4}, {-46, 4}, { 37, 5}, {-36, 4}, | |
50 | { 29, 4}, {-29, 4}, { 25, 4}, {-23, 4}, { 20, 4}, | |
51 | {-17, 4}, { 16, 4}, {-12, 4}, { 12, 3}, {-10, 4}, | |
52 | { 7, 3}, { -4, 4}, { 3, 3}, { -1, 3}, { 1, 3} }, | |
53 | { {-59, 3}, {-45, 5}, {-50, 4}, { 38, 4}, {-39, 4}, | |
54 | { 32, 4}, {-30, 4}, { 25, 3}, {-23, 3}, { 20, 3}, | |
55 | {-20, 3}, { 16, 3}, {-13, 3}, { 10, 3}, { -7, 3}, | |
56 | { 3, 3}, { 0, 3}, { -1, 3}, { 2, 3}, { -1, 2} } | |
57 | }; | |
58 | ||
59 | ||
60 | /** Scaled PARCOR values used for the first two PARCOR coefficients. | |
61 | * To be indexed by the Rice coded indices. | |
62 | * Generated by: parcor_scaled_values[i] = 32 + ((i * (i+1)) << 7) - (1 << 20) | |
63 | * Actual values are divided by 32 in order to be stored in 16 bits. | |
64 | */ | |
65 | static const int16_t parcor_scaled_values[] = { | |
66 | -1048544 / 32, -1048288 / 32, -1047776 / 32, -1047008 / 32, | |
67 | -1045984 / 32, -1044704 / 32, -1043168 / 32, -1041376 / 32, | |
68 | -1039328 / 32, -1037024 / 32, -1034464 / 32, -1031648 / 32, | |
69 | -1028576 / 32, -1025248 / 32, -1021664 / 32, -1017824 / 32, | |
70 | -1013728 / 32, -1009376 / 32, -1004768 / 32, -999904 / 32, | |
71 | -994784 / 32, -989408 / 32, -983776 / 32, -977888 / 32, | |
72 | -971744 / 32, -965344 / 32, -958688 / 32, -951776 / 32, | |
73 | -944608 / 32, -937184 / 32, -929504 / 32, -921568 / 32, | |
74 | -913376 / 32, -904928 / 32, -896224 / 32, -887264 / 32, | |
75 | -878048 / 32, -868576 / 32, -858848 / 32, -848864 / 32, | |
76 | -838624 / 32, -828128 / 32, -817376 / 32, -806368 / 32, | |
77 | -795104 / 32, -783584 / 32, -771808 / 32, -759776 / 32, | |
78 | -747488 / 32, -734944 / 32, -722144 / 32, -709088 / 32, | |
79 | -695776 / 32, -682208 / 32, -668384 / 32, -654304 / 32, | |
80 | -639968 / 32, -625376 / 32, -610528 / 32, -595424 / 32, | |
81 | -580064 / 32, -564448 / 32, -548576 / 32, -532448 / 32, | |
82 | -516064 / 32, -499424 / 32, -482528 / 32, -465376 / 32, | |
83 | -447968 / 32, -430304 / 32, -412384 / 32, -394208 / 32, | |
84 | -375776 / 32, -357088 / 32, -338144 / 32, -318944 / 32, | |
85 | -299488 / 32, -279776 / 32, -259808 / 32, -239584 / 32, | |
86 | -219104 / 32, -198368 / 32, -177376 / 32, -156128 / 32, | |
87 | -134624 / 32, -112864 / 32, -90848 / 32, -68576 / 32, | |
88 | -46048 / 32, -23264 / 32, -224 / 32, 23072 / 32, | |
89 | 46624 / 32, 70432 / 32, 94496 / 32, 118816 / 32, | |
90 | 143392 / 32, 168224 / 32, 193312 / 32, 218656 / 32, | |
91 | 244256 / 32, 270112 / 32, 296224 / 32, 322592 / 32, | |
92 | 349216 / 32, 376096 / 32, 403232 / 32, 430624 / 32, | |
93 | 458272 / 32, 486176 / 32, 514336 / 32, 542752 / 32, | |
94 | 571424 / 32, 600352 / 32, 629536 / 32, 658976 / 32, | |
95 | 688672 / 32, 718624 / 32, 748832 / 32, 779296 / 32, | |
96 | 810016 / 32, 840992 / 32, 872224 / 32, 903712 / 32, | |
97 | 935456 / 32, 967456 / 32, 999712 / 32, 1032224 / 32 | |
98 | }; | |
99 | ||
100 | ||
101 | /** Gain values of p(0) for long-term prediction. | |
102 | * To be indexed by the Rice coded indices. | |
103 | */ | |
104 | static const uint8_t ltp_gain_values [4][4] = { | |
105 | { 0, 8, 16, 24}, | |
106 | {32, 40, 48, 56}, | |
107 | {64, 70, 76, 82}, | |
108 | {88, 92, 96, 100} | |
109 | }; | |
110 | ||
99971952 TB |
111 | |
112 | enum RA_Flag { | |
113 | RA_FLAG_NONE, | |
114 | RA_FLAG_FRAMES, | |
115 | RA_FLAG_HEADER | |
116 | }; | |
117 | ||
118 | ||
119 | typedef struct { | |
120 | uint32_t samples; ///< number of samples, 0xFFFFFFFF if unknown | |
121 | int resolution; ///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit | |
122 | int floating; ///< 1 = IEEE 32-bit floating-point, 0 = integer | |
123 | int frame_length; ///< frame length for each frame (last frame may differ) | |
124 | int ra_distance; ///< distance between RA frames (in frames, 0...255) | |
125 | enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored | |
126 | int adapt_order; ///< adaptive order: 1 = on, 0 = off | |
127 | int coef_table; ///< table index of Rice code parameters | |
128 | int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off | |
129 | int max_order; ///< maximum prediction order (0..1023) | |
130 | int block_switching; ///< number of block switching levels | |
131 | int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only) | |
132 | int sb_part; ///< sub-block partition | |
133 | int joint_stereo; ///< joint stereo: 1 = on, 0 = off | |
134 | int mc_coding; ///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off | |
135 | int chan_config; ///< indicates that a chan_config_info field is present | |
136 | int chan_sort; ///< channel rearrangement: 1 = on, 0 = off | |
137 | int rlslms; ///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off | |
138 | int chan_config_info; ///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented. | |
139 | int *chan_pos; ///< original channel positions | |
140 | uint32_t header_size; ///< header size of original audio file in bytes, provided for debugging | |
141 | uint32_t trailer_size; ///< trailer size of original audio file in bytes, provided for debugging | |
142 | } ALSSpecificConfig; | |
143 | ||
144 | ||
145 | typedef struct { | |
146 | AVCodecContext *avctx; | |
147 | ALSSpecificConfig sconf; | |
148 | GetBitContext gb; | |
149 | unsigned int cur_frame_length; ///< length of the current frame to decode | |
150 | unsigned int frame_id; ///< the frame ID / number of the current frame | |
151 | unsigned int js_switch; ///< if true, joint-stereo decoding is enforced | |
152 | unsigned int num_blocks; ///< number of blocks used in the current frame | |
93d38cf6 | 153 | int ltp_lag_length; ///< number of bits used for ltp lag value |
1261b07f TB |
154 | int *use_ltp; ///< contains use_ltp flags for all channels |
155 | int *ltp_lag; ///< contains ltp lag values for all channels | |
156 | int **ltp_gain; ///< gain values for ltp 5-tap filter for a channel | |
157 | int *ltp_gain_buffer; ///< contains all gain values for ltp 5-tap filter | |
99971952 TB |
158 | int32_t *quant_cof; ///< quantized parcor coefficients |
159 | int32_t *lpc_cof; ///< coefficients of the direct form prediction filter | |
160 | int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block | |
161 | int32_t **raw_samples; ///< decoded raw samples for each channel | |
162 | int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples | |
163 | } ALSDecContext; | |
164 | ||
165 | ||
1261b07f TB |
166 | typedef struct { |
167 | unsigned int block_length; ///< number of samples within the block | |
168 | unsigned int ra_block; ///< if true, this is a random access block | |
169 | int const_block; ///< if true, this is a constant value block | |
170 | int32_t const_val; ///< the sample value of a constant block | |
171 | int js_blocks; ///< true if this block contains a difference signal | |
172 | unsigned int shift_lsbs; ///< shift of values for this block | |
173 | unsigned int opt_order; ///< prediction order of this block | |
174 | int store_prev_samples;///< if true, carryover samples have to be stored | |
175 | int *use_ltp; ///< if true, long-term prediction is used | |
176 | int *ltp_lag; ///< lag value for long-term prediction | |
177 | int *ltp_gain; ///< gain values for ltp 5-tap filter | |
178 | int32_t *quant_cof; ///< quantized parcor coefficients | |
179 | int32_t *lpc_cof; ///< coefficients of the direct form prediction | |
180 | int32_t *raw_samples; ///< decoded raw samples / residuals for this block | |
181 | int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block | |
182 | int32_t *raw_other; ///< decoded raw samples of the other channel of a channel pair | |
183 | } ALSBlockData; | |
184 | ||
185 | ||
99971952 TB |
186 | static av_cold void dprint_specific_config(ALSDecContext *ctx) |
187 | { | |
188 | #ifdef DEBUG | |
189 | AVCodecContext *avctx = ctx->avctx; | |
190 | ALSSpecificConfig *sconf = &ctx->sconf; | |
191 | ||
192 | dprintf(avctx, "resolution = %i\n", sconf->resolution); | |
193 | dprintf(avctx, "floating = %i\n", sconf->floating); | |
194 | dprintf(avctx, "frame_length = %i\n", sconf->frame_length); | |
195 | dprintf(avctx, "ra_distance = %i\n", sconf->ra_distance); | |
196 | dprintf(avctx, "ra_flag = %i\n", sconf->ra_flag); | |
197 | dprintf(avctx, "adapt_order = %i\n", sconf->adapt_order); | |
198 | dprintf(avctx, "coef_table = %i\n", sconf->coef_table); | |
199 | dprintf(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction); | |
200 | dprintf(avctx, "max_order = %i\n", sconf->max_order); | |
201 | dprintf(avctx, "block_switching = %i\n", sconf->block_switching); | |
202 | dprintf(avctx, "bgmc = %i\n", sconf->bgmc); | |
203 | dprintf(avctx, "sb_part = %i\n", sconf->sb_part); | |
204 | dprintf(avctx, "joint_stereo = %i\n", sconf->joint_stereo); | |
205 | dprintf(avctx, "mc_coding = %i\n", sconf->mc_coding); | |
206 | dprintf(avctx, "chan_config = %i\n", sconf->chan_config); | |
207 | dprintf(avctx, "chan_sort = %i\n", sconf->chan_sort); | |
208 | dprintf(avctx, "RLSLMS = %i\n", sconf->rlslms); | |
209 | dprintf(avctx, "chan_config_info = %i\n", sconf->chan_config_info); | |
210 | dprintf(avctx, "header_size = %i\n", sconf->header_size); | |
211 | dprintf(avctx, "trailer_size = %i\n", sconf->trailer_size); | |
212 | #endif | |
213 | } | |
214 | ||
215 | ||
216 | /** Reads an ALSSpecificConfig from a buffer into the output struct. | |
217 | */ | |
218 | static av_cold int read_specific_config(ALSDecContext *ctx) | |
219 | { | |
220 | GetBitContext gb; | |
221 | uint64_t ht_size; | |
222 | int i, config_offset, crc_enabled; | |
223 | MPEG4AudioConfig m4ac; | |
224 | ALSSpecificConfig *sconf = &ctx->sconf; | |
225 | AVCodecContext *avctx = ctx->avctx; | |
226 | uint32_t als_id; | |
227 | ||
228 | init_get_bits(&gb, avctx->extradata, avctx->extradata_size * 8); | |
229 | ||
230 | config_offset = ff_mpeg4audio_get_config(&m4ac, avctx->extradata, | |
231 | avctx->extradata_size); | |
232 | ||
233 | if (config_offset < 0) | |
234 | return -1; | |
235 | ||
236 | skip_bits_long(&gb, config_offset); | |
237 | ||
238 | if (get_bits_left(&gb) < (30 << 3)) | |
239 | return -1; | |
240 | ||
241 | // read the fixed items | |
242 | als_id = get_bits_long(&gb, 32); | |
243 | avctx->sample_rate = m4ac.sample_rate; | |
244 | skip_bits_long(&gb, 32); // sample rate already known | |
245 | sconf->samples = get_bits_long(&gb, 32); | |
246 | avctx->channels = m4ac.channels; | |
247 | skip_bits(&gb, 16); // number of channels already knwon | |
248 | skip_bits(&gb, 3); // skip file_type | |
249 | sconf->resolution = get_bits(&gb, 3); | |
250 | sconf->floating = get_bits1(&gb); | |
251 | skip_bits1(&gb); // skip msb_first | |
252 | sconf->frame_length = get_bits(&gb, 16) + 1; | |
253 | sconf->ra_distance = get_bits(&gb, 8); | |
254 | sconf->ra_flag = get_bits(&gb, 2); | |
255 | sconf->adapt_order = get_bits1(&gb); | |
256 | sconf->coef_table = get_bits(&gb, 2); | |
257 | sconf->long_term_prediction = get_bits1(&gb); | |
258 | sconf->max_order = get_bits(&gb, 10); | |
259 | sconf->block_switching = get_bits(&gb, 2); | |
260 | sconf->bgmc = get_bits1(&gb); | |
261 | sconf->sb_part = get_bits1(&gb); | |
262 | sconf->joint_stereo = get_bits1(&gb); | |
263 | sconf->mc_coding = get_bits1(&gb); | |
264 | sconf->chan_config = get_bits1(&gb); | |
265 | sconf->chan_sort = get_bits1(&gb); | |
266 | crc_enabled = get_bits1(&gb); | |
267 | sconf->rlslms = get_bits1(&gb); | |
268 | skip_bits(&gb, 5); // skip 5 reserved bits | |
269 | skip_bits1(&gb); // skip aux_data_enabled | |
270 | ||
271 | ||
272 | // check for ALSSpecificConfig struct | |
273 | if (als_id != MKBETAG('A','L','S','\0')) | |
274 | return -1; | |
275 | ||
276 | ctx->cur_frame_length = sconf->frame_length; | |
277 | ||
278 | // allocate quantized parcor coefficient buffer | |
279 | if (!(ctx->quant_cof = av_malloc(sizeof(*ctx->quant_cof) * sconf->max_order)) || | |
280 | !(ctx->lpc_cof = av_malloc(sizeof(*ctx->lpc_cof) * sconf->max_order))) { | |
281 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
282 | return AVERROR(ENOMEM); | |
283 | } | |
284 | ||
285 | // read channel config | |
286 | if (sconf->chan_config) | |
287 | sconf->chan_config_info = get_bits(&gb, 16); | |
288 | // TODO: use this to set avctx->channel_layout | |
289 | ||
290 | ||
291 | // read channel sorting | |
292 | if (sconf->chan_sort && avctx->channels > 1) { | |
293 | int chan_pos_bits = av_ceil_log2(avctx->channels); | |
294 | int bits_needed = avctx->channels * chan_pos_bits + 7; | |
295 | if (get_bits_left(&gb) < bits_needed) | |
296 | return -1; | |
297 | ||
298 | if (!(sconf->chan_pos = av_malloc(avctx->channels * sizeof(*sconf->chan_pos)))) | |
299 | return AVERROR(ENOMEM); | |
300 | ||
301 | for (i = 0; i < avctx->channels; i++) | |
302 | sconf->chan_pos[i] = get_bits(&gb, chan_pos_bits); | |
303 | ||
304 | align_get_bits(&gb); | |
305 | // TODO: use this to actually do channel sorting | |
306 | } else { | |
307 | sconf->chan_sort = 0; | |
308 | } | |
309 | ||
310 | ||
311 | // read fixed header and trailer sizes, | |
312 | // if size = 0xFFFFFFFF then there is no data field! | |
313 | if (get_bits_left(&gb) < 64) | |
314 | return -1; | |
315 | ||
316 | sconf->header_size = get_bits_long(&gb, 32); | |
317 | sconf->trailer_size = get_bits_long(&gb, 32); | |
318 | if (sconf->header_size == 0xFFFFFFFF) | |
319 | sconf->header_size = 0; | |
320 | if (sconf->trailer_size == 0xFFFFFFFF) | |
321 | sconf->trailer_size = 0; | |
322 | ||
323 | ht_size = ((int64_t)(sconf->header_size) + (int64_t)(sconf->trailer_size)) << 3; | |
324 | ||
325 | ||
326 | // skip the header and trailer data | |
327 | if (get_bits_left(&gb) < ht_size) | |
328 | return -1; | |
329 | ||
330 | if (ht_size > INT32_MAX) | |
331 | return -1; | |
332 | ||
333 | skip_bits_long(&gb, ht_size); | |
334 | ||
335 | ||
336 | // skip the crc data | |
337 | if (crc_enabled) { | |
338 | if (get_bits_left(&gb) < 32) | |
339 | return -1; | |
340 | ||
341 | skip_bits_long(&gb, 32); | |
342 | } | |
343 | ||
344 | ||
345 | // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data) | |
346 | ||
347 | dprint_specific_config(ctx); | |
348 | ||
349 | return 0; | |
350 | } | |
351 | ||
352 | ||
353 | /** Checks the ALSSpecificConfig for unsupported features. | |
354 | */ | |
355 | static int check_specific_config(ALSDecContext *ctx) | |
356 | { | |
357 | ALSSpecificConfig *sconf = &ctx->sconf; | |
358 | int error = 0; | |
359 | ||
360 | // report unsupported feature and set error value | |
361 | #define MISSING_ERR(cond, str, errval) \ | |
362 | { \ | |
363 | if (cond) { \ | |
364 | av_log_missing_feature(ctx->avctx, str, 0); \ | |
365 | error = errval; \ | |
366 | } \ | |
367 | } | |
368 | ||
369 | MISSING_ERR(sconf->floating, "Floating point decoding", -1); | |
99971952 TB |
370 | MISSING_ERR(sconf->bgmc, "BGMC entropy decoding", -1); |
371 | MISSING_ERR(sconf->mc_coding, "Multi-channel correlation", -1); | |
372 | MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", -1); | |
373 | MISSING_ERR(sconf->chan_sort, "Channel sorting", 0); | |
374 | ||
375 | return error; | |
376 | } | |
377 | ||
378 | ||
379 | /** Parses the bs_info field to extract the block partitioning used in | |
380 | * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2. | |
381 | */ | |
382 | static void parse_bs_info(const uint32_t bs_info, unsigned int n, | |
383 | unsigned int div, unsigned int **div_blocks, | |
384 | unsigned int *num_blocks) | |
385 | { | |
386 | if (n < 31 && ((bs_info << n) & 0x40000000)) { | |
387 | // if the level is valid and the investigated bit n is set | |
388 | // then recursively check both children at bits (2n+1) and (2n+2) | |
389 | n *= 2; | |
390 | div += 1; | |
391 | parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks); | |
392 | parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks); | |
393 | } else { | |
394 | // else the bit is not set or the last level has been reached | |
395 | // (bit implicitly not set) | |
396 | **div_blocks = div; | |
397 | (*div_blocks)++; | |
398 | (*num_blocks)++; | |
399 | } | |
400 | } | |
401 | ||
402 | ||
403 | /** Reads and decodes a Rice codeword. | |
404 | */ | |
405 | static int32_t decode_rice(GetBitContext *gb, unsigned int k) | |
406 | { | |
6e44ba15 | 407 | int max = get_bits_left(gb) - k; |
99971952 TB |
408 | int q = get_unary(gb, 0, max); |
409 | int r = k ? get_bits1(gb) : !(q & 1); | |
410 | ||
411 | if (k > 1) { | |
412 | q <<= (k - 1); | |
413 | q += get_bits_long(gb, k - 1); | |
414 | } else if (!k) { | |
415 | q >>= 1; | |
416 | } | |
417 | return r ? q : ~q; | |
418 | } | |
419 | ||
420 | ||
421 | /** Converts PARCOR coefficient k to direct filter coefficient. | |
422 | */ | |
423 | static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof) | |
424 | { | |
425 | int i, j; | |
426 | ||
427 | for (i = 0, j = k - 1; i < j; i++, j--) { | |
428 | int tmp1 = ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); | |
429 | cof[j] += ((MUL64(par[k], cof[i]) + (1 << 19)) >> 20); | |
430 | cof[i] += tmp1; | |
431 | } | |
432 | if (i == j) | |
433 | cof[i] += ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20); | |
434 | ||
435 | cof[k] = par[k]; | |
436 | } | |
437 | ||
438 | ||
439 | /** Reads block switching field if necessary and sets actual block sizes. | |
440 | * Also assures that the block sizes of the last frame correspond to the | |
441 | * actual number of samples. | |
442 | */ | |
443 | static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks, | |
444 | uint32_t *bs_info) | |
445 | { | |
446 | ALSSpecificConfig *sconf = &ctx->sconf; | |
447 | GetBitContext *gb = &ctx->gb; | |
448 | unsigned int *ptr_div_blocks = div_blocks; | |
449 | unsigned int b; | |
450 | ||
451 | if (sconf->block_switching) { | |
452 | unsigned int bs_info_len = 1 << (sconf->block_switching + 2); | |
453 | *bs_info = get_bits_long(gb, bs_info_len); | |
454 | *bs_info <<= (32 - bs_info_len); | |
455 | } | |
456 | ||
457 | ctx->num_blocks = 0; | |
458 | parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks); | |
459 | ||
460 | // The last frame may have an overdetermined block structure given in | |
461 | // the bitstream. In that case the defined block structure would need | |
462 | // more samples than available to be consistent. | |
463 | // The block structure is actually used but the block sizes are adapted | |
464 | // to fit the actual number of available samples. | |
465 | // Example: 5 samples, 2nd level block sizes: 2 2 2 2. | |
466 | // This results in the actual block sizes: 2 2 1 0. | |
467 | // This is not specified in 14496-3 but actually done by the reference | |
468 | // codec RM22 revision 2. | |
469 | // This appears to happen in case of an odd number of samples in the last | |
470 | // frame which is actually not allowed by the block length switching part | |
471 | // of 14496-3. | |
472 | // The ALS conformance files feature an odd number of samples in the last | |
473 | // frame. | |
474 | ||
475 | for (b = 0; b < ctx->num_blocks; b++) | |
476 | div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b]; | |
477 | ||
478 | if (ctx->cur_frame_length != ctx->sconf.frame_length) { | |
479 | unsigned int remaining = ctx->cur_frame_length; | |
480 | ||
481 | for (b = 0; b < ctx->num_blocks; b++) { | |
482 | if (remaining < div_blocks[b]) { | |
483 | div_blocks[b] = remaining; | |
484 | ctx->num_blocks = b + 1; | |
485 | break; | |
486 | } | |
487 | ||
488 | remaining -= div_blocks[b]; | |
489 | } | |
490 | } | |
491 | } | |
492 | ||
493 | ||
494 | /** Reads the block data for a constant block | |
495 | */ | |
1261b07f | 496 | static void read_const_block_data(ALSDecContext *ctx, ALSBlockData *bd) |
99971952 TB |
497 | { |
498 | ALSSpecificConfig *sconf = &ctx->sconf; | |
499 | AVCodecContext *avctx = ctx->avctx; | |
500 | GetBitContext *gb = &ctx->gb; | |
99971952 | 501 | |
1261b07f TB |
502 | bd->const_val = 0; |
503 | bd->const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence) | |
504 | bd->js_blocks = get_bits1(gb); | |
99971952 TB |
505 | |
506 | // skip 5 reserved bits | |
507 | skip_bits(gb, 5); | |
508 | ||
1261b07f | 509 | if (bd->const_block) { |
99971952 | 510 | unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample; |
1261b07f | 511 | bd->const_val = get_sbits_long(gb, const_val_bits); |
99971952 TB |
512 | } |
513 | ||
1261b07f TB |
514 | // ensure constant block decoding by reusing this field |
515 | bd->const_block = 1; | |
516 | } | |
517 | ||
518 | ||
519 | /** Decodes the block data for a constant block | |
520 | */ | |
521 | static void decode_const_block_data(ALSDecContext *ctx, ALSBlockData *bd) | |
522 | { | |
523 | int smp = bd->block_length; | |
524 | int32_t val = bd->const_val; | |
525 | int32_t *dst = bd->raw_samples; | |
526 | ||
99971952 | 527 | // write raw samples into buffer |
1261b07f TB |
528 | for (; smp; smp--) |
529 | *dst++ = val; | |
99971952 TB |
530 | } |
531 | ||
532 | ||
533 | /** Reads the block data for a non-constant block | |
534 | */ | |
1261b07f | 535 | static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd) |
99971952 TB |
536 | { |
537 | ALSSpecificConfig *sconf = &ctx->sconf; | |
538 | AVCodecContext *avctx = ctx->avctx; | |
539 | GetBitContext *gb = &ctx->gb; | |
540 | unsigned int k; | |
541 | unsigned int s[8]; | |
542 | unsigned int sub_blocks, log2_sub_blocks, sb_length; | |
99971952 | 543 | unsigned int start = 0; |
1261b07f TB |
544 | unsigned int opt_order; |
545 | int sb; | |
546 | int32_t *quant_cof = bd->quant_cof; | |
99971952 | 547 | |
1261b07f TB |
548 | |
549 | // ensure variable block decoding by reusing this field | |
550 | bd->const_block = 0; | |
551 | ||
552 | bd->opt_order = 1; | |
553 | bd->js_blocks = get_bits1(gb); | |
554 | ||
555 | opt_order = bd->opt_order; | |
99971952 TB |
556 | |
557 | // determine the number of subblocks for entropy decoding | |
558 | if (!sconf->bgmc && !sconf->sb_part) { | |
559 | log2_sub_blocks = 0; | |
560 | } else { | |
561 | if (sconf->bgmc && sconf->sb_part) | |
562 | log2_sub_blocks = get_bits(gb, 2); | |
563 | else | |
564 | log2_sub_blocks = 2 * get_bits1(gb); | |
565 | } | |
566 | ||
567 | sub_blocks = 1 << log2_sub_blocks; | |
568 | ||
569 | // do not continue in case of a damaged stream since | |
570 | // block_length must be evenly divisible by sub_blocks | |
1261b07f | 571 | if (bd->block_length & (sub_blocks - 1)) { |
99971952 TB |
572 | av_log(avctx, AV_LOG_WARNING, |
573 | "Block length is not evenly divisible by the number of subblocks.\n"); | |
574 | return -1; | |
575 | } | |
576 | ||
1261b07f | 577 | sb_length = bd->block_length >> log2_sub_blocks; |
99971952 TB |
578 | |
579 | ||
580 | if (sconf->bgmc) { | |
581 | // TODO: BGMC mode | |
582 | } else { | |
583 | s[0] = get_bits(gb, 4 + (sconf->resolution > 1)); | |
584 | for (k = 1; k < sub_blocks; k++) | |
585 | s[k] = s[k - 1] + decode_rice(gb, 0); | |
586 | } | |
587 | ||
588 | if (get_bits1(gb)) | |
1261b07f | 589 | bd->shift_lsbs = get_bits(gb, 4) + 1; |
99971952 | 590 | |
1261b07f | 591 | bd->store_prev_samples = (bd->js_blocks && bd->raw_other) || bd->shift_lsbs; |
99971952 TB |
592 | |
593 | ||
594 | if (!sconf->rlslms) { | |
595 | if (sconf->adapt_order) { | |
1261b07f | 596 | int opt_order_length = av_ceil_log2(av_clip((bd->block_length >> 3) - 1, |
99971952 | 597 | 2, sconf->max_order + 1)); |
1261b07f | 598 | bd->opt_order = get_bits(gb, opt_order_length); |
99971952 | 599 | } else { |
1261b07f | 600 | bd->opt_order = sconf->max_order; |
99971952 TB |
601 | } |
602 | ||
1261b07f TB |
603 | opt_order = bd->opt_order; |
604 | ||
99971952 TB |
605 | if (opt_order) { |
606 | int add_base; | |
607 | ||
608 | if (sconf->coef_table == 3) { | |
609 | add_base = 0x7F; | |
610 | ||
611 | // read coefficient 0 | |
612 | quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)]; | |
613 | ||
614 | // read coefficient 1 | |
615 | if (opt_order > 1) | |
616 | quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)]; | |
617 | ||
618 | // read coefficients 2 to opt_order | |
619 | for (k = 2; k < opt_order; k++) | |
620 | quant_cof[k] = get_bits(gb, 7); | |
621 | } else { | |
622 | int k_max; | |
623 | add_base = 1; | |
624 | ||
625 | // read coefficient 0 to 19 | |
626 | k_max = FFMIN(opt_order, 20); | |
627 | for (k = 0; k < k_max; k++) { | |
628 | int rice_param = parcor_rice_table[sconf->coef_table][k][1]; | |
629 | int offset = parcor_rice_table[sconf->coef_table][k][0]; | |
630 | quant_cof[k] = decode_rice(gb, rice_param) + offset; | |
631 | } | |
632 | ||
633 | // read coefficients 20 to 126 | |
634 | k_max = FFMIN(opt_order, 127); | |
635 | for (; k < k_max; k++) | |
636 | quant_cof[k] = decode_rice(gb, 2) + (k & 1); | |
637 | ||
638 | // read coefficients 127 to opt_order | |
639 | for (; k < opt_order; k++) | |
640 | quant_cof[k] = decode_rice(gb, 1); | |
641 | ||
642 | quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64]; | |
643 | ||
644 | if (opt_order > 1) | |
645 | quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64]; | |
646 | } | |
647 | ||
648 | for (k = 2; k < opt_order; k++) | |
649 | quant_cof[k] = (quant_cof[k] << 14) + (add_base << 13); | |
650 | } | |
651 | } | |
652 | ||
93d38cf6 TB |
653 | // read LTP gain and lag values |
654 | if (sconf->long_term_prediction) { | |
1261b07f | 655 | *bd->use_ltp = get_bits1(gb); |
93d38cf6 | 656 | |
1261b07f TB |
657 | if (*bd->use_ltp) { |
658 | bd->ltp_gain[0] = decode_rice(gb, 1) << 3; | |
659 | bd->ltp_gain[1] = decode_rice(gb, 2) << 3; | |
93d38cf6 | 660 | |
1261b07f | 661 | bd->ltp_gain[2] = ltp_gain_values[get_unary(gb, 0, 4)][get_bits(gb, 2)]; |
93d38cf6 | 662 | |
1261b07f TB |
663 | bd->ltp_gain[3] = decode_rice(gb, 2) << 3; |
664 | bd->ltp_gain[4] = decode_rice(gb, 1) << 3; | |
93d38cf6 | 665 | |
1261b07f TB |
666 | *bd->ltp_lag = get_bits(gb, ctx->ltp_lag_length); |
667 | *bd->ltp_lag += FFMAX(4, opt_order + 1); | |
93d38cf6 TB |
668 | } |
669 | } | |
99971952 TB |
670 | |
671 | // read first value and residuals in case of a random access block | |
1261b07f | 672 | if (bd->ra_block) { |
99971952 | 673 | if (opt_order) |
1261b07f | 674 | bd->raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4); |
99971952 | 675 | if (opt_order > 1) |
1261b07f | 676 | bd->raw_samples[1] = decode_rice(gb, s[0] + 3); |
99971952 | 677 | if (opt_order > 2) |
1261b07f | 678 | bd->raw_samples[2] = decode_rice(gb, s[0] + 1); |
99971952 TB |
679 | |
680 | start = FFMIN(opt_order, 3); | |
681 | } | |
682 | ||
683 | // read all residuals | |
684 | if (sconf->bgmc) { | |
685 | // TODO: BGMC mode | |
686 | } else { | |
1261b07f | 687 | int32_t *current_res = bd->raw_samples + start; |
99971952 TB |
688 | |
689 | for (sb = 0; sb < sub_blocks; sb++, start = 0) | |
690 | for (; start < sb_length; start++) | |
691 | *current_res++ = decode_rice(gb, s[sb]); | |
692 | } | |
693 | ||
1261b07f TB |
694 | if (!sconf->mc_coding || ctx->js_switch) |
695 | align_get_bits(gb); | |
696 | ||
697 | return 0; | |
698 | } | |
699 | ||
700 | ||
701 | /** Decodes the block data for a non-constant block | |
702 | */ | |
703 | static int decode_var_block_data(ALSDecContext *ctx, ALSBlockData *bd) | |
704 | { | |
705 | ALSSpecificConfig *sconf = &ctx->sconf; | |
706 | unsigned int block_length = bd->block_length; | |
707 | unsigned int smp = 0; | |
708 | unsigned int k; | |
709 | unsigned int opt_order = bd->opt_order; | |
710 | int sb; | |
711 | int64_t y; | |
712 | int32_t *quant_cof = bd->quant_cof; | |
713 | int32_t *lpc_cof = bd->lpc_cof; | |
714 | int32_t *raw_samples = bd->raw_samples; | |
99c5f5cc TB |
715 | int32_t *raw_samples_end = bd->raw_samples + bd->block_length; |
716 | int32_t lpc_cof_reversed[opt_order]; | |
1261b07f | 717 | |
93d38cf6 | 718 | // reverse long-term prediction |
1261b07f | 719 | if (*bd->use_ltp) { |
93d38cf6 TB |
720 | int ltp_smp; |
721 | ||
1261b07f TB |
722 | for (ltp_smp = FFMAX(*bd->ltp_lag - 2, 0); ltp_smp < block_length; ltp_smp++) { |
723 | int center = ltp_smp - *bd->ltp_lag; | |
93d38cf6 TB |
724 | int begin = FFMAX(0, center - 2); |
725 | int end = center + 3; | |
726 | int tab = 5 - (end - begin); | |
727 | int base; | |
728 | ||
729 | y = 1 << 6; | |
730 | ||
731 | for (base = begin; base < end; base++, tab++) | |
1261b07f | 732 | y += MUL64(bd->ltp_gain[tab], raw_samples[base]); |
93d38cf6 TB |
733 | |
734 | raw_samples[ltp_smp] += y >> 7; | |
735 | } | |
736 | } | |
737 | ||
99971952 | 738 | // reconstruct all samples from residuals |
1261b07f | 739 | if (bd->ra_block) { |
99971952 TB |
740 | for (smp = 0; smp < opt_order; smp++) { |
741 | y = 1 << 19; | |
742 | ||
743 | for (sb = 0; sb < smp; sb++) | |
99c5f5cc | 744 | y += MUL64(lpc_cof[sb], raw_samples[-(sb + 1)]); |
99971952 | 745 | |
99c5f5cc | 746 | *raw_samples++ -= y >> 20; |
99971952 TB |
747 | parcor_to_lpc(smp, quant_cof, lpc_cof); |
748 | } | |
749 | } else { | |
750 | for (k = 0; k < opt_order; k++) | |
751 | parcor_to_lpc(k, quant_cof, lpc_cof); | |
752 | ||
753 | // store previous samples in case that they have to be altered | |
1261b07f TB |
754 | if (bd->store_prev_samples) |
755 | memcpy(bd->prev_raw_samples, raw_samples - sconf->max_order, | |
756 | sizeof(*bd->prev_raw_samples) * sconf->max_order); | |
99971952 TB |
757 | |
758 | // reconstruct difference signal for prediction (joint-stereo) | |
1261b07f | 759 | if (bd->js_blocks && bd->raw_other) { |
99971952 TB |
760 | int32_t *left, *right; |
761 | ||
1261b07f | 762 | if (bd->raw_other > raw_samples) { // D = R - L |
99971952 | 763 | left = raw_samples; |
1261b07f | 764 | right = bd->raw_other; |
99971952 | 765 | } else { // D = R - L |
1261b07f | 766 | left = bd->raw_other; |
99971952 TB |
767 | right = raw_samples; |
768 | } | |
769 | ||
770 | for (sb = -1; sb >= -sconf->max_order; sb--) | |
771 | raw_samples[sb] = right[sb] - left[sb]; | |
772 | } | |
773 | ||
774 | // reconstruct shifted signal | |
1261b07f | 775 | if (bd->shift_lsbs) |
99971952 | 776 | for (sb = -1; sb >= -sconf->max_order; sb--) |
1261b07f | 777 | raw_samples[sb] >>= bd->shift_lsbs; |
99971952 TB |
778 | } |
779 | ||
99c5f5cc TB |
780 | // reverse linear prediction coefficients for efficiency |
781 | lpc_cof = lpc_cof + opt_order; | |
782 | ||
783 | for (sb = 0; sb < opt_order; sb++) | |
784 | lpc_cof_reversed[sb] = lpc_cof[-(sb + 1)]; | |
785 | ||
99971952 | 786 | // reconstruct raw samples |
99c5f5cc TB |
787 | raw_samples = bd->raw_samples + smp; |
788 | lpc_cof = lpc_cof_reversed + opt_order; | |
789 | ||
790 | for (; raw_samples < raw_samples_end; raw_samples++) { | |
99971952 TB |
791 | y = 1 << 19; |
792 | ||
99c5f5cc TB |
793 | for (sb = -opt_order; sb < 0; sb++) |
794 | y += MUL64(lpc_cof[sb], raw_samples[sb]); | |
99971952 | 795 | |
99c5f5cc | 796 | *raw_samples -= y >> 20; |
99971952 TB |
797 | } |
798 | ||
99c5f5cc TB |
799 | raw_samples = bd->raw_samples; |
800 | ||
99971952 | 801 | // restore previous samples in case that they have been altered |
1261b07f TB |
802 | if (bd->store_prev_samples) |
803 | memcpy(raw_samples - sconf->max_order, bd->prev_raw_samples, | |
99971952 TB |
804 | sizeof(*raw_samples) * sconf->max_order); |
805 | ||
806 | return 0; | |
807 | } | |
808 | ||
809 | ||
810 | /** Reads the block data. | |
811 | */ | |
1261b07f | 812 | static int read_block(ALSDecContext *ctx, ALSBlockData *bd) |
99971952 | 813 | { |
99971952 | 814 | GetBitContext *gb = &ctx->gb; |
99971952 TB |
815 | |
816 | // read block type flag and read the samples accordingly | |
817 | if (get_bits1(gb)) { | |
1261b07f | 818 | if (read_var_block_data(ctx, bd)) |
99971952 TB |
819 | return -1; |
820 | } else { | |
1261b07f | 821 | read_const_block_data(ctx, bd); |
99971952 TB |
822 | } |
823 | ||
1261b07f TB |
824 | return 0; |
825 | } | |
99971952 | 826 | |
99971952 | 827 | |
1261b07f TB |
828 | /** Decodes the block data. |
829 | */ | |
830 | static int decode_block(ALSDecContext *ctx, ALSBlockData *bd) | |
831 | { | |
832 | unsigned int smp; | |
833 | ||
834 | // read block type flag and read the samples accordingly | |
835 | if (bd->const_block) | |
836 | decode_const_block_data(ctx, bd); | |
837 | else if (decode_var_block_data(ctx, bd)) | |
838 | return -1; | |
839 | ||
840 | // TODO: read RLSLMS extension data | |
841 | ||
842 | if (bd->shift_lsbs) | |
843 | for (smp = 0; smp < bd->block_length; smp++) | |
844 | bd->raw_samples[smp] <<= bd->shift_lsbs; | |
99971952 TB |
845 | |
846 | return 0; | |
847 | } | |
848 | ||
849 | ||
1261b07f TB |
850 | /** Reads and decodes block data successively. |
851 | */ | |
852 | static int read_decode_block(ALSDecContext *ctx, ALSBlockData *bd) | |
853 | { | |
854 | int ret; | |
855 | ||
856 | ret = read_block(ctx, bd); | |
857 | ||
858 | if (ret) | |
859 | return ret; | |
860 | ||
861 | ret = decode_block(ctx, bd); | |
862 | ||
863 | return ret; | |
864 | } | |
865 | ||
866 | ||
99971952 TB |
867 | /** Computes the number of samples left to decode for the current frame and |
868 | * sets these samples to zero. | |
869 | */ | |
870 | static void zero_remaining(unsigned int b, unsigned int b_max, | |
871 | const unsigned int *div_blocks, int32_t *buf) | |
872 | { | |
873 | unsigned int count = 0; | |
874 | ||
875 | while (b < b_max) | |
876 | count += div_blocks[b]; | |
877 | ||
0bb622ba | 878 | if (count) |
9349e558 | 879 | memset(buf, 0, sizeof(*buf) * count); |
99971952 TB |
880 | } |
881 | ||
882 | ||
883 | /** Decodes blocks independently. | |
884 | */ | |
885 | static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame, | |
886 | unsigned int c, const unsigned int *div_blocks, | |
887 | unsigned int *js_blocks) | |
888 | { | |
99971952 | 889 | unsigned int b; |
1261b07f TB |
890 | ALSBlockData bd; |
891 | ||
892 | memset(&bd, 0, sizeof(ALSBlockData)); | |
893 | ||
894 | bd.ra_block = ra_frame; | |
895 | bd.use_ltp = ctx->use_ltp; | |
896 | bd.ltp_lag = ctx->ltp_lag; | |
897 | bd.ltp_gain = ctx->ltp_gain[0]; | |
898 | bd.quant_cof = ctx->quant_cof; | |
899 | bd.lpc_cof = ctx->lpc_cof; | |
900 | bd.prev_raw_samples = ctx->prev_raw_samples; | |
901 | bd.raw_samples = ctx->raw_samples[c]; | |
902 | ||
99971952 TB |
903 | |
904 | for (b = 0; b < ctx->num_blocks; b++) { | |
1261b07f TB |
905 | bd.shift_lsbs = 0; |
906 | bd.block_length = div_blocks[b]; | |
907 | ||
908 | if (read_decode_block(ctx, &bd)) { | |
99971952 | 909 | // damaged block, write zero for the rest of the frame |
1261b07f | 910 | zero_remaining(b, ctx->num_blocks, div_blocks, bd.raw_samples); |
99971952 TB |
911 | return -1; |
912 | } | |
1261b07f TB |
913 | bd.raw_samples += div_blocks[b]; |
914 | bd.ra_block = 0; | |
99971952 TB |
915 | } |
916 | ||
917 | return 0; | |
918 | } | |
919 | ||
920 | ||
921 | /** Decodes blocks dependently. | |
922 | */ | |
923 | static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame, | |
924 | unsigned int c, const unsigned int *div_blocks, | |
925 | unsigned int *js_blocks) | |
926 | { | |
927 | ALSSpecificConfig *sconf = &ctx->sconf; | |
928 | unsigned int offset = 0; | |
99971952 | 929 | unsigned int b; |
1261b07f TB |
930 | ALSBlockData bd[2]; |
931 | ||
932 | memset(bd, 0, 2 * sizeof(ALSBlockData)); | |
933 | ||
934 | bd[0].ra_block = ra_frame; | |
935 | bd[0].use_ltp = ctx->use_ltp; | |
936 | bd[0].ltp_lag = ctx->ltp_lag; | |
937 | bd[0].ltp_gain = ctx->ltp_gain[0]; | |
938 | bd[0].quant_cof = ctx->quant_cof; | |
939 | bd[0].lpc_cof = ctx->lpc_cof; | |
940 | bd[0].prev_raw_samples = ctx->prev_raw_samples; | |
941 | bd[0].js_blocks = *js_blocks; | |
942 | ||
943 | bd[1].ra_block = ra_frame; | |
944 | bd[1].use_ltp = ctx->use_ltp; | |
945 | bd[1].ltp_lag = ctx->ltp_lag; | |
946 | bd[1].ltp_gain = ctx->ltp_gain[0]; | |
947 | bd[1].quant_cof = ctx->quant_cof; | |
948 | bd[1].lpc_cof = ctx->lpc_cof; | |
949 | bd[1].prev_raw_samples = ctx->prev_raw_samples; | |
950 | bd[1].js_blocks = *(js_blocks + 1); | |
99971952 TB |
951 | |
952 | // decode all blocks | |
953 | for (b = 0; b < ctx->num_blocks; b++) { | |
954 | unsigned int s; | |
1261b07f TB |
955 | |
956 | bd[0].shift_lsbs = 0; | |
957 | bd[1].shift_lsbs = 0; | |
958 | ||
959 | bd[0].block_length = div_blocks[b]; | |
960 | bd[1].block_length = div_blocks[b]; | |
961 | ||
962 | bd[0].raw_samples = ctx->raw_samples[c ] + offset; | |
963 | bd[1].raw_samples = ctx->raw_samples[c + 1] + offset; | |
964 | ||
965 | bd[0].raw_other = bd[1].raw_samples; | |
966 | bd[1].raw_other = bd[0].raw_samples; | |
967 | ||
968 | if(read_decode_block(ctx, &bd[0]) || read_decode_block(ctx, &bd[1])) { | |
99971952 | 969 | // damaged block, write zero for the rest of the frame |
1261b07f TB |
970 | zero_remaining(b, ctx->num_blocks, div_blocks, bd[0].raw_samples); |
971 | zero_remaining(b, ctx->num_blocks, div_blocks, bd[1].raw_samples); | |
99971952 TB |
972 | return -1; |
973 | } | |
974 | ||
975 | // reconstruct joint-stereo blocks | |
1261b07f TB |
976 | if (bd[0].js_blocks) { |
977 | if (bd[1].js_blocks) | |
99971952 TB |
978 | av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair!\n"); |
979 | ||
980 | for (s = 0; s < div_blocks[b]; s++) | |
1261b07f TB |
981 | bd[0].raw_samples[s] = bd[1].raw_samples[s] - bd[0].raw_samples[s]; |
982 | } else if (bd[1].js_blocks) { | |
99971952 | 983 | for (s = 0; s < div_blocks[b]; s++) |
1261b07f | 984 | bd[1].raw_samples[s] = bd[1].raw_samples[s] + bd[0].raw_samples[s]; |
99971952 TB |
985 | } |
986 | ||
987 | offset += div_blocks[b]; | |
1261b07f TB |
988 | bd[0].ra_block = 0; |
989 | bd[1].ra_block = 0; | |
99971952 TB |
990 | } |
991 | ||
992 | // store carryover raw samples, | |
993 | // the others channel raw samples are stored by the calling function. | |
994 | memmove(ctx->raw_samples[c] - sconf->max_order, | |
995 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
996 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
997 | ||
998 | return 0; | |
999 | } | |
1000 | ||
1001 | ||
1002 | /** Reads the frame data. | |
1003 | */ | |
1004 | static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame) | |
1005 | { | |
1006 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1007 | AVCodecContext *avctx = ctx->avctx; | |
1008 | GetBitContext *gb = &ctx->gb; | |
1009 | unsigned int div_blocks[32]; ///< block sizes. | |
1010 | unsigned int c; | |
1011 | unsigned int js_blocks[2]; | |
1012 | ||
1013 | uint32_t bs_info = 0; | |
1014 | ||
1015 | // skip the size of the ra unit if present in the frame | |
1016 | if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame) | |
1017 | skip_bits_long(gb, 32); | |
1018 | ||
1019 | if (sconf->mc_coding && sconf->joint_stereo) { | |
1020 | ctx->js_switch = get_bits1(gb); | |
1021 | align_get_bits(gb); | |
1022 | } | |
1023 | ||
1024 | if (!sconf->mc_coding || ctx->js_switch) { | |
1025 | int independent_bs = !sconf->joint_stereo; | |
1026 | ||
1027 | for (c = 0; c < avctx->channels; c++) { | |
1028 | js_blocks[0] = 0; | |
1029 | js_blocks[1] = 0; | |
1030 | ||
1031 | get_block_sizes(ctx, div_blocks, &bs_info); | |
1032 | ||
1033 | // if joint_stereo and block_switching is set, independent decoding | |
1034 | // is signaled via the first bit of bs_info | |
1035 | if (sconf->joint_stereo && sconf->block_switching) | |
1036 | if (bs_info >> 31) | |
1037 | independent_bs = 2; | |
1038 | ||
1039 | // if this is the last channel, it has to be decoded independently | |
1040 | if (c == avctx->channels - 1) | |
1041 | independent_bs = 1; | |
1042 | ||
1043 | if (independent_bs) { | |
1044 | if (decode_blocks_ind(ctx, ra_frame, c, div_blocks, js_blocks)) | |
1045 | return -1; | |
1046 | ||
1047 | independent_bs--; | |
1048 | } else { | |
1049 | if (decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks)) | |
1050 | return -1; | |
1051 | ||
1052 | c++; | |
1053 | } | |
1054 | ||
1055 | // store carryover raw samples | |
1056 | memmove(ctx->raw_samples[c] - sconf->max_order, | |
1057 | ctx->raw_samples[c] - sconf->max_order + sconf->frame_length, | |
1058 | sizeof(*ctx->raw_samples[c]) * sconf->max_order); | |
1059 | } | |
1060 | } else { // multi-channel coding | |
1061 | get_block_sizes(ctx, div_blocks, &bs_info); | |
1062 | ||
1063 | // TODO: multi channel coding might use a temporary buffer instead as | |
1064 | // the actual channel is not known when read_block-data is called | |
1065 | if (decode_blocks_ind(ctx, ra_frame, 0, div_blocks, js_blocks)) | |
1066 | return -1; | |
1067 | // TODO: read_channel_data | |
1068 | } | |
1069 | ||
1070 | // TODO: read_diff_float_data | |
1071 | ||
1072 | return 0; | |
1073 | } | |
1074 | ||
1075 | ||
1076 | /** Decodes an ALS frame. | |
1077 | */ | |
1078 | static int decode_frame(AVCodecContext *avctx, | |
1079 | void *data, int *data_size, | |
1080 | AVPacket *avpkt) | |
1081 | { | |
1082 | ALSDecContext *ctx = avctx->priv_data; | |
1083 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1084 | const uint8_t *buffer = avpkt->data; | |
1085 | int buffer_size = avpkt->size; | |
1086 | int invalid_frame, size; | |
1087 | unsigned int c, sample, ra_frame, bytes_read, shift; | |
1088 | ||
1089 | init_get_bits(&ctx->gb, buffer, buffer_size * 8); | |
1090 | ||
1091 | // In the case that the distance between random access frames is set to zero | |
1092 | // (sconf->ra_distance == 0) no frame is treated as a random access frame. | |
1093 | // For the first frame, if prediction is used, all samples used from the | |
1094 | // previous frame are assumed to be zero. | |
1095 | ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance); | |
1096 | ||
1097 | // the last frame to decode might have a different length | |
1098 | if (sconf->samples != 0xFFFFFFFF) | |
1099 | ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length, | |
1100 | sconf->frame_length); | |
1101 | else | |
1102 | ctx->cur_frame_length = sconf->frame_length; | |
1103 | ||
1104 | // decode the frame data | |
1105 | if ((invalid_frame = read_frame_data(ctx, ra_frame) < 0)) | |
1106 | av_log(ctx->avctx, AV_LOG_WARNING, | |
1107 | "Reading frame data failed. Skipping RA unit.\n"); | |
1108 | ||
1109 | ctx->frame_id++; | |
1110 | ||
1111 | // check for size of decoded data | |
1112 | size = ctx->cur_frame_length * avctx->channels * | |
1113 | (av_get_bits_per_sample_format(avctx->sample_fmt) >> 3); | |
1114 | ||
1115 | if (size > *data_size) { | |
1116 | av_log(avctx, AV_LOG_ERROR, "Decoded data exceeds buffer size.\n"); | |
1117 | return -1; | |
1118 | } | |
1119 | ||
1120 | *data_size = size; | |
1121 | ||
1122 | // transform decoded frame into output format | |
1123 | #define INTERLEAVE_OUTPUT(bps) \ | |
1124 | { \ | |
1125 | int##bps##_t *dest = (int##bps##_t*) data; \ | |
1126 | shift = bps - ctx->avctx->bits_per_raw_sample; \ | |
1127 | for (sample = 0; sample < ctx->cur_frame_length; sample++) \ | |
1128 | for (c = 0; c < avctx->channels; c++) \ | |
1129 | *dest++ = ctx->raw_samples[c][sample] << shift; \ | |
1130 | } | |
1131 | ||
1132 | if (ctx->avctx->bits_per_raw_sample <= 16) { | |
1133 | INTERLEAVE_OUTPUT(16) | |
1134 | } else { | |
1135 | INTERLEAVE_OUTPUT(32) | |
1136 | } | |
1137 | ||
1138 | bytes_read = invalid_frame ? buffer_size : | |
1139 | (get_bits_count(&ctx->gb) + 7) >> 3; | |
1140 | ||
1141 | return bytes_read; | |
1142 | } | |
1143 | ||
1144 | ||
1145 | /** Uninitializes the ALS decoder. | |
1146 | */ | |
1147 | static av_cold int decode_end(AVCodecContext *avctx) | |
1148 | { | |
1149 | ALSDecContext *ctx = avctx->priv_data; | |
1150 | ||
1151 | av_freep(&ctx->sconf.chan_pos); | |
1152 | ||
1261b07f TB |
1153 | av_freep(&ctx->use_ltp); |
1154 | av_freep(&ctx->ltp_lag); | |
1155 | av_freep(&ctx->ltp_gain); | |
1156 | av_freep(&ctx->ltp_gain_buffer); | |
99971952 TB |
1157 | av_freep(&ctx->quant_cof); |
1158 | av_freep(&ctx->lpc_cof); | |
1159 | av_freep(&ctx->prev_raw_samples); | |
1160 | av_freep(&ctx->raw_samples); | |
1161 | av_freep(&ctx->raw_buffer); | |
1162 | ||
1163 | return 0; | |
1164 | } | |
1165 | ||
1166 | ||
1167 | /** Initializes the ALS decoder. | |
1168 | */ | |
1169 | static av_cold int decode_init(AVCodecContext *avctx) | |
1170 | { | |
1171 | unsigned int c; | |
1172 | unsigned int channel_size; | |
1261b07f | 1173 | int num_buffers; |
99971952 TB |
1174 | ALSDecContext *ctx = avctx->priv_data; |
1175 | ALSSpecificConfig *sconf = &ctx->sconf; | |
1176 | ctx->avctx = avctx; | |
1177 | ||
1178 | if (!avctx->extradata) { | |
1179 | av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n"); | |
1180 | return -1; | |
1181 | } | |
1182 | ||
1183 | if (read_specific_config(ctx)) { | |
1184 | av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n"); | |
1185 | decode_end(avctx); | |
1186 | return -1; | |
1187 | } | |
1188 | ||
1189 | if (check_specific_config(ctx)) { | |
1190 | decode_end(avctx); | |
1191 | return -1; | |
1192 | } | |
1193 | ||
1194 | if (sconf->floating) { | |
1195 | avctx->sample_fmt = SAMPLE_FMT_FLT; | |
1196 | avctx->bits_per_raw_sample = 32; | |
1197 | } else { | |
1198 | avctx->sample_fmt = sconf->resolution > 1 | |
1199 | ? SAMPLE_FMT_S32 : SAMPLE_FMT_S16; | |
1200 | avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8; | |
1201 | } | |
1202 | ||
93d38cf6 TB |
1203 | // set lag value for long-term prediction |
1204 | ctx->ltp_lag_length = 8 + (avctx->sample_rate >= 96000) + | |
1205 | (avctx->sample_rate >= 192000); | |
1206 | ||
1261b07f TB |
1207 | // allocate quantized parcor coefficient buffer |
1208 | num_buffers = sconf->mc_coding ? avctx->channels : 1; | |
1209 | ||
1210 | // allocate and assign lag and gain data buffer for ltp mode | |
1211 | ctx->use_ltp = av_mallocz(sizeof(*ctx->use_ltp) * num_buffers); | |
1212 | ctx->ltp_lag = av_malloc (sizeof(*ctx->ltp_lag) * num_buffers); | |
1213 | ctx->ltp_gain = av_malloc (sizeof(*ctx->ltp_gain) * num_buffers); | |
1214 | ctx->ltp_gain_buffer = av_malloc (sizeof(*ctx->ltp_gain_buffer) * | |
1215 | num_buffers * 5); | |
1216 | ||
1217 | if (!ctx->use_ltp || !ctx->ltp_lag || | |
1218 | !ctx->ltp_gain || !ctx->ltp_gain_buffer) { | |
1219 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1220 | decode_end(avctx); | |
1221 | return AVERROR(ENOMEM); | |
1222 | } | |
1223 | ||
1224 | for (c = 0; c < num_buffers; c++) | |
1225 | ctx->ltp_gain[c] = ctx->ltp_gain_buffer + c * 5; | |
1226 | ||
99971952 TB |
1227 | avctx->frame_size = sconf->frame_length; |
1228 | channel_size = sconf->frame_length + sconf->max_order; | |
1229 | ||
1230 | ctx->prev_raw_samples = av_malloc (sizeof(*ctx->prev_raw_samples) * sconf->max_order); | |
1231 | ctx->raw_buffer = av_mallocz(sizeof(*ctx-> raw_buffer) * avctx->channels * channel_size); | |
1232 | ctx->raw_samples = av_malloc (sizeof(*ctx-> raw_samples) * avctx->channels); | |
1233 | ||
1234 | // allocate previous raw sample buffer | |
1235 | if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) { | |
1236 | av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n"); | |
1237 | decode_end(avctx); | |
1238 | return AVERROR(ENOMEM); | |
1239 | } | |
1240 | ||
1241 | // assign raw samples buffers | |
1242 | ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order; | |
1243 | for (c = 1; c < avctx->channels; c++) | |
1244 | ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size; | |
1245 | ||
1246 | return 0; | |
1247 | } | |
1248 | ||
1249 | ||
1250 | /** Flushes (resets) the frame ID after seeking. | |
1251 | */ | |
1252 | static av_cold void flush(AVCodecContext *avctx) | |
1253 | { | |
1254 | ALSDecContext *ctx = avctx->priv_data; | |
1255 | ||
1256 | ctx->frame_id = 0; | |
1257 | } | |
1258 | ||
1259 | ||
1260 | AVCodec als_decoder = { | |
1261 | "als", | |
1262 | CODEC_TYPE_AUDIO, | |
1263 | CODEC_ID_MP4ALS, | |
1264 | sizeof(ALSDecContext), | |
1265 | decode_init, | |
1266 | NULL, | |
1267 | decode_end, | |
1268 | decode_frame, | |
1269 | .flush = flush, | |
1270 | .capabilities = CODEC_CAP_SUBFRAMES, | |
1271 | .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"), | |
1272 | }; | |
1273 |