You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

2178 lines
78KB

  1. /*
  2. * MPEG-4 ALS decoder
  3. * Copyright (c) 2009 Thilo Borgmann <thilo.borgmann _at_ mail.de>
  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. * @file
  23. * MPEG-4 ALS decoder
  24. * @author Thilo Borgmann <thilo.borgmann _at_ mail.de>
  25. */
  26. #include <inttypes.h>
  27. #include "avcodec.h"
  28. #include "get_bits.h"
  29. #include "unary.h"
  30. #include "mpeg4audio.h"
  31. #include "bgmc.h"
  32. #include "bswapdsp.h"
  33. #include "internal.h"
  34. #include "mlz.h"
  35. #include "libavutil/samplefmt.h"
  36. #include "libavutil/crc.h"
  37. #include "libavutil/softfloat_ieee754.h"
  38. #include "libavutil/intfloat.h"
  39. #include "libavutil/intreadwrite.h"
  40. #include <stdint.h>
  41. /** Rice parameters and corresponding index offsets for decoding the
  42. * indices of scaled PARCOR values. The table chosen is set globally
  43. * by the encoder and stored in ALSSpecificConfig.
  44. */
  45. static const int8_t parcor_rice_table[3][20][2] = {
  46. { {-52, 4}, {-29, 5}, {-31, 4}, { 19, 4}, {-16, 4},
  47. { 12, 3}, { -7, 3}, { 9, 3}, { -5, 3}, { 6, 3},
  48. { -4, 3}, { 3, 3}, { -3, 2}, { 3, 2}, { -2, 2},
  49. { 3, 2}, { -1, 2}, { 2, 2}, { -1, 2}, { 2, 2} },
  50. { {-58, 3}, {-42, 4}, {-46, 4}, { 37, 5}, {-36, 4},
  51. { 29, 4}, {-29, 4}, { 25, 4}, {-23, 4}, { 20, 4},
  52. {-17, 4}, { 16, 4}, {-12, 4}, { 12, 3}, {-10, 4},
  53. { 7, 3}, { -4, 4}, { 3, 3}, { -1, 3}, { 1, 3} },
  54. { {-59, 3}, {-45, 5}, {-50, 4}, { 38, 4}, {-39, 4},
  55. { 32, 4}, {-30, 4}, { 25, 3}, {-23, 3}, { 20, 3},
  56. {-20, 3}, { 16, 3}, {-13, 3}, { 10, 3}, { -7, 3},
  57. { 3, 3}, { 0, 3}, { -1, 3}, { 2, 3}, { -1, 2} }
  58. };
  59. /** Scaled PARCOR values used for the first two PARCOR coefficients.
  60. * To be indexed by the Rice coded indices.
  61. * Generated by: parcor_scaled_values[i] = 32 + ((i * (i+1)) << 7) - (1 << 20)
  62. * Actual values are divided by 32 in order to be stored in 16 bits.
  63. */
  64. static const int16_t parcor_scaled_values[] = {
  65. -1048544 / 32, -1048288 / 32, -1047776 / 32, -1047008 / 32,
  66. -1045984 / 32, -1044704 / 32, -1043168 / 32, -1041376 / 32,
  67. -1039328 / 32, -1037024 / 32, -1034464 / 32, -1031648 / 32,
  68. -1028576 / 32, -1025248 / 32, -1021664 / 32, -1017824 / 32,
  69. -1013728 / 32, -1009376 / 32, -1004768 / 32, -999904 / 32,
  70. -994784 / 32, -989408 / 32, -983776 / 32, -977888 / 32,
  71. -971744 / 32, -965344 / 32, -958688 / 32, -951776 / 32,
  72. -944608 / 32, -937184 / 32, -929504 / 32, -921568 / 32,
  73. -913376 / 32, -904928 / 32, -896224 / 32, -887264 / 32,
  74. -878048 / 32, -868576 / 32, -858848 / 32, -848864 / 32,
  75. -838624 / 32, -828128 / 32, -817376 / 32, -806368 / 32,
  76. -795104 / 32, -783584 / 32, -771808 / 32, -759776 / 32,
  77. -747488 / 32, -734944 / 32, -722144 / 32, -709088 / 32,
  78. -695776 / 32, -682208 / 32, -668384 / 32, -654304 / 32,
  79. -639968 / 32, -625376 / 32, -610528 / 32, -595424 / 32,
  80. -580064 / 32, -564448 / 32, -548576 / 32, -532448 / 32,
  81. -516064 / 32, -499424 / 32, -482528 / 32, -465376 / 32,
  82. -447968 / 32, -430304 / 32, -412384 / 32, -394208 / 32,
  83. -375776 / 32, -357088 / 32, -338144 / 32, -318944 / 32,
  84. -299488 / 32, -279776 / 32, -259808 / 32, -239584 / 32,
  85. -219104 / 32, -198368 / 32, -177376 / 32, -156128 / 32,
  86. -134624 / 32, -112864 / 32, -90848 / 32, -68576 / 32,
  87. -46048 / 32, -23264 / 32, -224 / 32, 23072 / 32,
  88. 46624 / 32, 70432 / 32, 94496 / 32, 118816 / 32,
  89. 143392 / 32, 168224 / 32, 193312 / 32, 218656 / 32,
  90. 244256 / 32, 270112 / 32, 296224 / 32, 322592 / 32,
  91. 349216 / 32, 376096 / 32, 403232 / 32, 430624 / 32,
  92. 458272 / 32, 486176 / 32, 514336 / 32, 542752 / 32,
  93. 571424 / 32, 600352 / 32, 629536 / 32, 658976 / 32,
  94. 688672 / 32, 718624 / 32, 748832 / 32, 779296 / 32,
  95. 810016 / 32, 840992 / 32, 872224 / 32, 903712 / 32,
  96. 935456 / 32, 967456 / 32, 999712 / 32, 1032224 / 32
  97. };
  98. /** Gain values of p(0) for long-term prediction.
  99. * To be indexed by the Rice coded indices.
  100. */
  101. static const uint8_t ltp_gain_values [4][4] = {
  102. { 0, 8, 16, 24},
  103. {32, 40, 48, 56},
  104. {64, 70, 76, 82},
  105. {88, 92, 96, 100}
  106. };
  107. /** Inter-channel weighting factors for multi-channel correlation.
  108. * To be indexed by the Rice coded indices.
  109. */
  110. static const int16_t mcc_weightings[] = {
  111. 204, 192, 179, 166, 153, 140, 128, 115,
  112. 102, 89, 76, 64, 51, 38, 25, 12,
  113. 0, -12, -25, -38, -51, -64, -76, -89,
  114. -102, -115, -128, -140, -153, -166, -179, -192
  115. };
  116. /** Tail codes used in arithmetic coding using block Gilbert-Moore codes.
  117. */
  118. static const uint8_t tail_code[16][6] = {
  119. { 74, 44, 25, 13, 7, 3},
  120. { 68, 42, 24, 13, 7, 3},
  121. { 58, 39, 23, 13, 7, 3},
  122. {126, 70, 37, 19, 10, 5},
  123. {132, 70, 37, 20, 10, 5},
  124. {124, 70, 38, 20, 10, 5},
  125. {120, 69, 37, 20, 11, 5},
  126. {116, 67, 37, 20, 11, 5},
  127. {108, 66, 36, 20, 10, 5},
  128. {102, 62, 36, 20, 10, 5},
  129. { 88, 58, 34, 19, 10, 5},
  130. {162, 89, 49, 25, 13, 7},
  131. {156, 87, 49, 26, 14, 7},
  132. {150, 86, 47, 26, 14, 7},
  133. {142, 84, 47, 26, 14, 7},
  134. {131, 79, 46, 26, 14, 7}
  135. };
  136. enum RA_Flag {
  137. RA_FLAG_NONE,
  138. RA_FLAG_FRAMES,
  139. RA_FLAG_HEADER
  140. };
  141. typedef struct ALSSpecificConfig {
  142. uint32_t samples; ///< number of samples, 0xFFFFFFFF if unknown
  143. int resolution; ///< 000 = 8-bit; 001 = 16-bit; 010 = 24-bit; 011 = 32-bit
  144. int floating; ///< 1 = IEEE 32-bit floating-point, 0 = integer
  145. int msb_first; ///< 1 = original CRC calculated on big-endian system, 0 = little-endian
  146. int frame_length; ///< frame length for each frame (last frame may differ)
  147. int ra_distance; ///< distance between RA frames (in frames, 0...255)
  148. enum RA_Flag ra_flag; ///< indicates where the size of ra units is stored
  149. int adapt_order; ///< adaptive order: 1 = on, 0 = off
  150. int coef_table; ///< table index of Rice code parameters
  151. int long_term_prediction; ///< long term prediction (LTP): 1 = on, 0 = off
  152. int max_order; ///< maximum prediction order (0..1023)
  153. int block_switching; ///< number of block switching levels
  154. int bgmc; ///< "Block Gilbert-Moore Code": 1 = on, 0 = off (Rice coding only)
  155. int sb_part; ///< sub-block partition
  156. int joint_stereo; ///< joint stereo: 1 = on, 0 = off
  157. int mc_coding; ///< extended inter-channel coding (multi channel coding): 1 = on, 0 = off
  158. int chan_config; ///< indicates that a chan_config_info field is present
  159. int chan_sort; ///< channel rearrangement: 1 = on, 0 = off
  160. int rlslms; ///< use "Recursive Least Square-Least Mean Square" predictor: 1 = on, 0 = off
  161. int chan_config_info; ///< mapping of channels to loudspeaker locations. Unused until setting channel configuration is implemented.
  162. int *chan_pos; ///< original channel positions
  163. int crc_enabled; ///< enable Cyclic Redundancy Checksum
  164. } ALSSpecificConfig;
  165. typedef struct ALSChannelData {
  166. int stop_flag;
  167. int master_channel;
  168. int time_diff_flag;
  169. int time_diff_sign;
  170. int time_diff_index;
  171. int weighting[6];
  172. } ALSChannelData;
  173. typedef struct ALSDecContext {
  174. AVCodecContext *avctx;
  175. ALSSpecificConfig sconf;
  176. GetBitContext gb;
  177. BswapDSPContext bdsp;
  178. const AVCRC *crc_table;
  179. uint32_t crc_org; ///< CRC value of the original input data
  180. uint32_t crc; ///< CRC value calculated from decoded data
  181. unsigned int cur_frame_length; ///< length of the current frame to decode
  182. unsigned int frame_id; ///< the frame ID / number of the current frame
  183. unsigned int js_switch; ///< if true, joint-stereo decoding is enforced
  184. unsigned int cs_switch; ///< if true, channel rearrangement is done
  185. unsigned int num_blocks; ///< number of blocks used in the current frame
  186. unsigned int s_max; ///< maximum Rice parameter allowed in entropy coding
  187. uint8_t *bgmc_lut; ///< pointer at lookup tables used for BGMC
  188. int *bgmc_lut_status; ///< pointer at lookup table status flags used for BGMC
  189. int ltp_lag_length; ///< number of bits used for ltp lag value
  190. int *const_block; ///< contains const_block flags for all channels
  191. unsigned int *shift_lsbs; ///< contains shift_lsbs flags for all channels
  192. unsigned int *opt_order; ///< contains opt_order flags for all channels
  193. int *store_prev_samples; ///< contains store_prev_samples flags for all channels
  194. int *use_ltp; ///< contains use_ltp flags for all channels
  195. int *ltp_lag; ///< contains ltp lag values for all channels
  196. int **ltp_gain; ///< gain values for ltp 5-tap filter for a channel
  197. int *ltp_gain_buffer; ///< contains all gain values for ltp 5-tap filter
  198. int32_t **quant_cof; ///< quantized parcor coefficients for a channel
  199. int32_t *quant_cof_buffer; ///< contains all quantized parcor coefficients
  200. int32_t **lpc_cof; ///< coefficients of the direct form prediction filter for a channel
  201. int32_t *lpc_cof_buffer; ///< contains all coefficients of the direct form prediction filter
  202. int32_t *lpc_cof_reversed_buffer; ///< temporary buffer to set up a reversed versio of lpc_cof_buffer
  203. ALSChannelData **chan_data; ///< channel data for multi-channel correlation
  204. ALSChannelData *chan_data_buffer; ///< contains channel data for all channels
  205. int *reverted_channels; ///< stores a flag for each reverted channel
  206. int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block
  207. int32_t **raw_samples; ///< decoded raw samples for each channel
  208. int32_t *raw_buffer; ///< contains all decoded raw samples including carryover samples
  209. uint8_t *crc_buffer; ///< buffer of byte order corrected samples used for CRC check
  210. MLZ* mlz; ///< masked lz decompression structure
  211. SoftFloat_IEEE754 *acf; ///< contains common multiplier for all channels
  212. int *last_acf_mantissa; ///< contains the last acf mantissa data of common multiplier for all channels
  213. int *shift_value; ///< value by which the binary point is to be shifted for all channels
  214. int *last_shift_value; ///< contains last shift value for all channels
  215. int **raw_mantissa; ///< decoded mantissa bits of the difference signal
  216. unsigned char *larray; ///< buffer to store the output of masked lz decompression
  217. int *nbits; ///< contains the number of bits to read for masked lz decompression for all samples
  218. } ALSDecContext;
  219. typedef struct ALSBlockData {
  220. unsigned int block_length; ///< number of samples within the block
  221. unsigned int ra_block; ///< if true, this is a random access block
  222. int *const_block; ///< if true, this is a constant value block
  223. int js_blocks; ///< true if this block contains a difference signal
  224. unsigned int *shift_lsbs; ///< shift of values for this block
  225. unsigned int *opt_order; ///< prediction order of this block
  226. int *store_prev_samples;///< if true, carryover samples have to be stored
  227. int *use_ltp; ///< if true, long-term prediction is used
  228. int *ltp_lag; ///< lag value for long-term prediction
  229. int *ltp_gain; ///< gain values for ltp 5-tap filter
  230. int32_t *quant_cof; ///< quantized parcor coefficients
  231. int32_t *lpc_cof; ///< coefficients of the direct form prediction
  232. int32_t *raw_samples; ///< decoded raw samples / residuals for this block
  233. int32_t *prev_raw_samples; ///< contains unshifted raw samples from the previous block
  234. int32_t *raw_other; ///< decoded raw samples of the other channel of a channel pair
  235. } ALSBlockData;
  236. static av_cold void dprint_specific_config(ALSDecContext *ctx)
  237. {
  238. #ifdef DEBUG
  239. AVCodecContext *avctx = ctx->avctx;
  240. ALSSpecificConfig *sconf = &ctx->sconf;
  241. ff_dlog(avctx, "resolution = %i\n", sconf->resolution);
  242. ff_dlog(avctx, "floating = %i\n", sconf->floating);
  243. ff_dlog(avctx, "frame_length = %i\n", sconf->frame_length);
  244. ff_dlog(avctx, "ra_distance = %i\n", sconf->ra_distance);
  245. ff_dlog(avctx, "ra_flag = %i\n", sconf->ra_flag);
  246. ff_dlog(avctx, "adapt_order = %i\n", sconf->adapt_order);
  247. ff_dlog(avctx, "coef_table = %i\n", sconf->coef_table);
  248. ff_dlog(avctx, "long_term_prediction = %i\n", sconf->long_term_prediction);
  249. ff_dlog(avctx, "max_order = %i\n", sconf->max_order);
  250. ff_dlog(avctx, "block_switching = %i\n", sconf->block_switching);
  251. ff_dlog(avctx, "bgmc = %i\n", sconf->bgmc);
  252. ff_dlog(avctx, "sb_part = %i\n", sconf->sb_part);
  253. ff_dlog(avctx, "joint_stereo = %i\n", sconf->joint_stereo);
  254. ff_dlog(avctx, "mc_coding = %i\n", sconf->mc_coding);
  255. ff_dlog(avctx, "chan_config = %i\n", sconf->chan_config);
  256. ff_dlog(avctx, "chan_sort = %i\n", sconf->chan_sort);
  257. ff_dlog(avctx, "RLSLMS = %i\n", sconf->rlslms);
  258. ff_dlog(avctx, "chan_config_info = %i\n", sconf->chan_config_info);
  259. #endif
  260. }
  261. /** Read an ALSSpecificConfig from a buffer into the output struct.
  262. */
  263. static av_cold int read_specific_config(ALSDecContext *ctx)
  264. {
  265. GetBitContext gb;
  266. uint64_t ht_size;
  267. int i, config_offset;
  268. MPEG4AudioConfig m4ac = {0};
  269. ALSSpecificConfig *sconf = &ctx->sconf;
  270. AVCodecContext *avctx = ctx->avctx;
  271. uint32_t als_id, header_size, trailer_size;
  272. int ret;
  273. if ((ret = init_get_bits8(&gb, avctx->extradata, avctx->extradata_size)) < 0)
  274. return ret;
  275. config_offset = avpriv_mpeg4audio_get_config(&m4ac, avctx->extradata,
  276. avctx->extradata_size * 8, 1);
  277. if (config_offset < 0)
  278. return AVERROR_INVALIDDATA;
  279. skip_bits_long(&gb, config_offset);
  280. if (get_bits_left(&gb) < (30 << 3))
  281. return AVERROR_INVALIDDATA;
  282. // read the fixed items
  283. als_id = get_bits_long(&gb, 32);
  284. avctx->sample_rate = m4ac.sample_rate;
  285. skip_bits_long(&gb, 32); // sample rate already known
  286. sconf->samples = get_bits_long(&gb, 32);
  287. avctx->channels = m4ac.channels;
  288. skip_bits(&gb, 16); // number of channels already known
  289. skip_bits(&gb, 3); // skip file_type
  290. sconf->resolution = get_bits(&gb, 3);
  291. sconf->floating = get_bits1(&gb);
  292. sconf->msb_first = get_bits1(&gb);
  293. sconf->frame_length = get_bits(&gb, 16) + 1;
  294. sconf->ra_distance = get_bits(&gb, 8);
  295. sconf->ra_flag = get_bits(&gb, 2);
  296. sconf->adapt_order = get_bits1(&gb);
  297. sconf->coef_table = get_bits(&gb, 2);
  298. sconf->long_term_prediction = get_bits1(&gb);
  299. sconf->max_order = get_bits(&gb, 10);
  300. sconf->block_switching = get_bits(&gb, 2);
  301. sconf->bgmc = get_bits1(&gb);
  302. sconf->sb_part = get_bits1(&gb);
  303. sconf->joint_stereo = get_bits1(&gb);
  304. sconf->mc_coding = get_bits1(&gb);
  305. sconf->chan_config = get_bits1(&gb);
  306. sconf->chan_sort = get_bits1(&gb);
  307. sconf->crc_enabled = get_bits1(&gb);
  308. sconf->rlslms = get_bits1(&gb);
  309. skip_bits(&gb, 5); // skip 5 reserved bits
  310. skip_bits1(&gb); // skip aux_data_enabled
  311. // check for ALSSpecificConfig struct
  312. if (als_id != MKBETAG('A','L','S','\0'))
  313. return AVERROR_INVALIDDATA;
  314. if (avctx->channels > FF_SANE_NB_CHANNELS) {
  315. avpriv_request_sample(avctx, "Huge number of channels\n");
  316. return AVERROR_PATCHWELCOME;
  317. }
  318. ctx->cur_frame_length = sconf->frame_length;
  319. // read channel config
  320. if (sconf->chan_config)
  321. sconf->chan_config_info = get_bits(&gb, 16);
  322. // TODO: use this to set avctx->channel_layout
  323. // read channel sorting
  324. if (sconf->chan_sort && avctx->channels > 1) {
  325. int chan_pos_bits = av_ceil_log2(avctx->channels);
  326. int bits_needed = avctx->channels * chan_pos_bits + 7;
  327. if (get_bits_left(&gb) < bits_needed)
  328. return AVERROR_INVALIDDATA;
  329. if (!(sconf->chan_pos = av_malloc_array(avctx->channels, sizeof(*sconf->chan_pos))))
  330. return AVERROR(ENOMEM);
  331. ctx->cs_switch = 1;
  332. for (i = 0; i < avctx->channels; i++) {
  333. sconf->chan_pos[i] = -1;
  334. }
  335. for (i = 0; i < avctx->channels; i++) {
  336. int idx;
  337. idx = get_bits(&gb, chan_pos_bits);
  338. if (idx >= avctx->channels || sconf->chan_pos[idx] != -1) {
  339. av_log(avctx, AV_LOG_WARNING, "Invalid channel reordering.\n");
  340. ctx->cs_switch = 0;
  341. break;
  342. }
  343. sconf->chan_pos[idx] = i;
  344. }
  345. align_get_bits(&gb);
  346. }
  347. // read fixed header and trailer sizes,
  348. // if size = 0xFFFFFFFF then there is no data field!
  349. if (get_bits_left(&gb) < 64)
  350. return AVERROR_INVALIDDATA;
  351. header_size = get_bits_long(&gb, 32);
  352. trailer_size = get_bits_long(&gb, 32);
  353. if (header_size == 0xFFFFFFFF)
  354. header_size = 0;
  355. if (trailer_size == 0xFFFFFFFF)
  356. trailer_size = 0;
  357. ht_size = ((int64_t)(header_size) + (int64_t)(trailer_size)) << 3;
  358. // skip the header and trailer data
  359. if (get_bits_left(&gb) < ht_size)
  360. return AVERROR_INVALIDDATA;
  361. if (ht_size > INT32_MAX)
  362. return AVERROR_PATCHWELCOME;
  363. skip_bits_long(&gb, ht_size);
  364. // initialize CRC calculation
  365. if (sconf->crc_enabled) {
  366. if (get_bits_left(&gb) < 32)
  367. return AVERROR_INVALIDDATA;
  368. if (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL)) {
  369. ctx->crc_table = av_crc_get_table(AV_CRC_32_IEEE_LE);
  370. ctx->crc = 0xFFFFFFFF;
  371. ctx->crc_org = ~get_bits_long(&gb, 32);
  372. } else
  373. skip_bits_long(&gb, 32);
  374. }
  375. // no need to read the rest of ALSSpecificConfig (ra_unit_size & aux data)
  376. dprint_specific_config(ctx);
  377. return 0;
  378. }
  379. /** Check the ALSSpecificConfig for unsupported features.
  380. */
  381. static int check_specific_config(ALSDecContext *ctx)
  382. {
  383. ALSSpecificConfig *sconf = &ctx->sconf;
  384. int error = 0;
  385. // report unsupported feature and set error value
  386. #define MISSING_ERR(cond, str, errval) \
  387. { \
  388. if (cond) { \
  389. avpriv_report_missing_feature(ctx->avctx, \
  390. str); \
  391. error = errval; \
  392. } \
  393. }
  394. MISSING_ERR(sconf->rlslms, "Adaptive RLS-LMS prediction", AVERROR_PATCHWELCOME);
  395. return error;
  396. }
  397. /** Parse the bs_info field to extract the block partitioning used in
  398. * block switching mode, refer to ISO/IEC 14496-3, section 11.6.2.
  399. */
  400. static void parse_bs_info(const uint32_t bs_info, unsigned int n,
  401. unsigned int div, unsigned int **div_blocks,
  402. unsigned int *num_blocks)
  403. {
  404. if (n < 31 && ((bs_info << n) & 0x40000000)) {
  405. // if the level is valid and the investigated bit n is set
  406. // then recursively check both children at bits (2n+1) and (2n+2)
  407. n *= 2;
  408. div += 1;
  409. parse_bs_info(bs_info, n + 1, div, div_blocks, num_blocks);
  410. parse_bs_info(bs_info, n + 2, div, div_blocks, num_blocks);
  411. } else {
  412. // else the bit is not set or the last level has been reached
  413. // (bit implicitly not set)
  414. **div_blocks = div;
  415. (*div_blocks)++;
  416. (*num_blocks)++;
  417. }
  418. }
  419. /** Read and decode a Rice codeword.
  420. */
  421. static int32_t decode_rice(GetBitContext *gb, unsigned int k)
  422. {
  423. int max = get_bits_left(gb) - k;
  424. unsigned q = get_unary(gb, 0, max);
  425. int r = k ? get_bits1(gb) : !(q & 1);
  426. if (k > 1) {
  427. q <<= (k - 1);
  428. q += get_bits_long(gb, k - 1);
  429. } else if (!k) {
  430. q >>= 1;
  431. }
  432. return r ? q : ~q;
  433. }
  434. /** Convert PARCOR coefficient k to direct filter coefficient.
  435. */
  436. static void parcor_to_lpc(unsigned int k, const int32_t *par, int32_t *cof)
  437. {
  438. int i, j;
  439. for (i = 0, j = k - 1; i < j; i++, j--) {
  440. unsigned tmp1 = ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20);
  441. cof[j] += ((MUL64(par[k], cof[i]) + (1 << 19)) >> 20);
  442. cof[i] += tmp1;
  443. }
  444. if (i == j)
  445. cof[i] += ((MUL64(par[k], cof[j]) + (1 << 19)) >> 20);
  446. cof[k] = par[k];
  447. }
  448. /** Read block switching field if necessary and set actual block sizes.
  449. * Also assure that the block sizes of the last frame correspond to the
  450. * actual number of samples.
  451. */
  452. static void get_block_sizes(ALSDecContext *ctx, unsigned int *div_blocks,
  453. uint32_t *bs_info)
  454. {
  455. ALSSpecificConfig *sconf = &ctx->sconf;
  456. GetBitContext *gb = &ctx->gb;
  457. unsigned int *ptr_div_blocks = div_blocks;
  458. unsigned int b;
  459. if (sconf->block_switching) {
  460. unsigned int bs_info_len = 1 << (sconf->block_switching + 2);
  461. *bs_info = get_bits_long(gb, bs_info_len);
  462. *bs_info <<= (32 - bs_info_len);
  463. }
  464. ctx->num_blocks = 0;
  465. parse_bs_info(*bs_info, 0, 0, &ptr_div_blocks, &ctx->num_blocks);
  466. // The last frame may have an overdetermined block structure given in
  467. // the bitstream. In that case the defined block structure would need
  468. // more samples than available to be consistent.
  469. // The block structure is actually used but the block sizes are adapted
  470. // to fit the actual number of available samples.
  471. // Example: 5 samples, 2nd level block sizes: 2 2 2 2.
  472. // This results in the actual block sizes: 2 2 1 0.
  473. // This is not specified in 14496-3 but actually done by the reference
  474. // codec RM22 revision 2.
  475. // This appears to happen in case of an odd number of samples in the last
  476. // frame which is actually not allowed by the block length switching part
  477. // of 14496-3.
  478. // The ALS conformance files feature an odd number of samples in the last
  479. // frame.
  480. for (b = 0; b < ctx->num_blocks; b++)
  481. div_blocks[b] = ctx->sconf.frame_length >> div_blocks[b];
  482. if (ctx->cur_frame_length != ctx->sconf.frame_length) {
  483. unsigned int remaining = ctx->cur_frame_length;
  484. for (b = 0; b < ctx->num_blocks; b++) {
  485. if (remaining <= div_blocks[b]) {
  486. div_blocks[b] = remaining;
  487. ctx->num_blocks = b + 1;
  488. break;
  489. }
  490. remaining -= div_blocks[b];
  491. }
  492. }
  493. }
  494. /** Read the block data for a constant block
  495. */
  496. static int read_const_block_data(ALSDecContext *ctx, ALSBlockData *bd)
  497. {
  498. ALSSpecificConfig *sconf = &ctx->sconf;
  499. AVCodecContext *avctx = ctx->avctx;
  500. GetBitContext *gb = &ctx->gb;
  501. if (bd->block_length <= 0)
  502. return AVERROR_INVALIDDATA;
  503. *bd->raw_samples = 0;
  504. *bd->const_block = get_bits1(gb); // 1 = constant value, 0 = zero block (silence)
  505. bd->js_blocks = get_bits1(gb);
  506. // skip 5 reserved bits
  507. skip_bits(gb, 5);
  508. if (*bd->const_block) {
  509. unsigned int const_val_bits = sconf->floating ? 24 : avctx->bits_per_raw_sample;
  510. *bd->raw_samples = get_sbits_long(gb, const_val_bits);
  511. }
  512. // ensure constant block decoding by reusing this field
  513. *bd->const_block = 1;
  514. return 0;
  515. }
  516. /** Decode the block data for a constant block
  517. */
  518. static void decode_const_block_data(ALSDecContext *ctx, ALSBlockData *bd)
  519. {
  520. int smp = bd->block_length - 1;
  521. int32_t val = *bd->raw_samples;
  522. int32_t *dst = bd->raw_samples + 1;
  523. // write raw samples into buffer
  524. for (; smp; smp--)
  525. *dst++ = val;
  526. }
  527. /** Read the block data for a non-constant block
  528. */
  529. static int read_var_block_data(ALSDecContext *ctx, ALSBlockData *bd)
  530. {
  531. ALSSpecificConfig *sconf = &ctx->sconf;
  532. AVCodecContext *avctx = ctx->avctx;
  533. GetBitContext *gb = &ctx->gb;
  534. unsigned int k;
  535. unsigned int s[8];
  536. unsigned int sx[8];
  537. unsigned int sub_blocks, log2_sub_blocks, sb_length;
  538. unsigned int start = 0;
  539. unsigned int opt_order;
  540. int sb;
  541. int32_t *quant_cof = bd->quant_cof;
  542. int32_t *current_res;
  543. // ensure variable block decoding by reusing this field
  544. *bd->const_block = 0;
  545. *bd->opt_order = 1;
  546. bd->js_blocks = get_bits1(gb);
  547. opt_order = *bd->opt_order;
  548. // determine the number of subblocks for entropy decoding
  549. if (!sconf->bgmc && !sconf->sb_part) {
  550. log2_sub_blocks = 0;
  551. } else {
  552. if (sconf->bgmc && sconf->sb_part)
  553. log2_sub_blocks = get_bits(gb, 2);
  554. else
  555. log2_sub_blocks = 2 * get_bits1(gb);
  556. }
  557. sub_blocks = 1 << log2_sub_blocks;
  558. // do not continue in case of a damaged stream since
  559. // block_length must be evenly divisible by sub_blocks
  560. if (bd->block_length & (sub_blocks - 1) || bd->block_length <= 0) {
  561. av_log(avctx, AV_LOG_WARNING,
  562. "Block length is not evenly divisible by the number of subblocks.\n");
  563. return AVERROR_INVALIDDATA;
  564. }
  565. sb_length = bd->block_length >> log2_sub_blocks;
  566. if (sconf->bgmc) {
  567. s[0] = get_bits(gb, 8 + (sconf->resolution > 1));
  568. for (k = 1; k < sub_blocks; k++)
  569. s[k] = s[k - 1] + decode_rice(gb, 2);
  570. for (k = 0; k < sub_blocks; k++) {
  571. sx[k] = s[k] & 0x0F;
  572. s [k] >>= 4;
  573. }
  574. } else {
  575. s[0] = get_bits(gb, 4 + (sconf->resolution > 1));
  576. for (k = 1; k < sub_blocks; k++)
  577. s[k] = s[k - 1] + decode_rice(gb, 0);
  578. }
  579. for (k = 1; k < sub_blocks; k++)
  580. if (s[k] > 32) {
  581. av_log(avctx, AV_LOG_ERROR, "k invalid for rice code.\n");
  582. return AVERROR_INVALIDDATA;
  583. }
  584. if (get_bits1(gb))
  585. *bd->shift_lsbs = get_bits(gb, 4) + 1;
  586. *bd->store_prev_samples = (bd->js_blocks && bd->raw_other) || *bd->shift_lsbs;
  587. if (!sconf->rlslms) {
  588. if (sconf->adapt_order && sconf->max_order) {
  589. int opt_order_length = av_ceil_log2(av_clip((bd->block_length >> 3) - 1,
  590. 2, sconf->max_order + 1));
  591. *bd->opt_order = get_bits(gb, opt_order_length);
  592. if (*bd->opt_order > sconf->max_order) {
  593. *bd->opt_order = sconf->max_order;
  594. av_log(avctx, AV_LOG_ERROR, "Predictor order too large.\n");
  595. return AVERROR_INVALIDDATA;
  596. }
  597. } else {
  598. *bd->opt_order = sconf->max_order;
  599. }
  600. if (*bd->opt_order > bd->block_length) {
  601. *bd->opt_order = bd->block_length;
  602. av_log(avctx, AV_LOG_ERROR, "Predictor order too large.\n");
  603. return AVERROR_INVALIDDATA;
  604. }
  605. opt_order = *bd->opt_order;
  606. if (opt_order) {
  607. int add_base;
  608. if (sconf->coef_table == 3) {
  609. add_base = 0x7F;
  610. // read coefficient 0
  611. quant_cof[0] = 32 * parcor_scaled_values[get_bits(gb, 7)];
  612. // read coefficient 1
  613. if (opt_order > 1)
  614. quant_cof[1] = -32 * parcor_scaled_values[get_bits(gb, 7)];
  615. // read coefficients 2 to opt_order
  616. for (k = 2; k < opt_order; k++)
  617. quant_cof[k] = get_bits(gb, 7);
  618. } else {
  619. int k_max;
  620. add_base = 1;
  621. // read coefficient 0 to 19
  622. k_max = FFMIN(opt_order, 20);
  623. for (k = 0; k < k_max; k++) {
  624. int rice_param = parcor_rice_table[sconf->coef_table][k][1];
  625. int offset = parcor_rice_table[sconf->coef_table][k][0];
  626. quant_cof[k] = decode_rice(gb, rice_param) + offset;
  627. if (quant_cof[k] < -64 || quant_cof[k] > 63) {
  628. av_log(avctx, AV_LOG_ERROR,
  629. "quant_cof %"PRId32" is out of range.\n",
  630. quant_cof[k]);
  631. return AVERROR_INVALIDDATA;
  632. }
  633. }
  634. // read coefficients 20 to 126
  635. k_max = FFMIN(opt_order, 127);
  636. for (; k < k_max; k++)
  637. quant_cof[k] = decode_rice(gb, 2) + (k & 1);
  638. // read coefficients 127 to opt_order
  639. for (; k < opt_order; k++)
  640. quant_cof[k] = decode_rice(gb, 1);
  641. quant_cof[0] = 32 * parcor_scaled_values[quant_cof[0] + 64];
  642. if (opt_order > 1)
  643. quant_cof[1] = -32 * parcor_scaled_values[quant_cof[1] + 64];
  644. }
  645. for (k = 2; k < opt_order; k++)
  646. quant_cof[k] = (quant_cof[k] * (1 << 14)) + (add_base << 13);
  647. }
  648. }
  649. // read LTP gain and lag values
  650. if (sconf->long_term_prediction) {
  651. *bd->use_ltp = get_bits1(gb);
  652. if (*bd->use_ltp) {
  653. int r, c;
  654. bd->ltp_gain[0] = decode_rice(gb, 1) * 8;
  655. bd->ltp_gain[1] = decode_rice(gb, 2) * 8;
  656. r = get_unary(gb, 0, 4);
  657. c = get_bits(gb, 2);
  658. if (r >= 4) {
  659. av_log(avctx, AV_LOG_ERROR, "r overflow\n");
  660. return AVERROR_INVALIDDATA;
  661. }
  662. bd->ltp_gain[2] = ltp_gain_values[r][c];
  663. bd->ltp_gain[3] = decode_rice(gb, 2) * 8;
  664. bd->ltp_gain[4] = decode_rice(gb, 1) * 8;
  665. *bd->ltp_lag = get_bits(gb, ctx->ltp_lag_length);
  666. *bd->ltp_lag += FFMAX(4, opt_order + 1);
  667. }
  668. }
  669. // read first value and residuals in case of a random access block
  670. if (bd->ra_block) {
  671. start = FFMIN(opt_order, 3);
  672. av_assert0(sb_length <= sconf->frame_length);
  673. if (sb_length <= start) {
  674. // opt_order or sb_length may be corrupted, either way this is unsupported and not well defined in the specification
  675. av_log(avctx, AV_LOG_ERROR, "Sub block length smaller or equal start\n");
  676. return AVERROR_PATCHWELCOME;
  677. }
  678. if (opt_order)
  679. bd->raw_samples[0] = decode_rice(gb, avctx->bits_per_raw_sample - 4);
  680. if (opt_order > 1)
  681. bd->raw_samples[1] = decode_rice(gb, FFMIN(s[0] + 3, ctx->s_max));
  682. if (opt_order > 2)
  683. bd->raw_samples[2] = decode_rice(gb, FFMIN(s[0] + 1, ctx->s_max));
  684. }
  685. // read all residuals
  686. if (sconf->bgmc) {
  687. int delta[8];
  688. unsigned int k [8];
  689. unsigned int b = av_clip((av_ceil_log2(bd->block_length) - 3) >> 1, 0, 5);
  690. // read most significant bits
  691. unsigned int high;
  692. unsigned int low;
  693. unsigned int value;
  694. int ret = ff_bgmc_decode_init(gb, &high, &low, &value);
  695. if (ret < 0)
  696. return ret;
  697. current_res = bd->raw_samples + start;
  698. for (sb = 0; sb < sub_blocks; sb++) {
  699. unsigned int sb_len = sb_length - (sb ? 0 : start);
  700. k [sb] = s[sb] > b ? s[sb] - b : 0;
  701. delta[sb] = 5 - s[sb] + k[sb];
  702. if (k[sb] >= 32)
  703. return AVERROR_INVALIDDATA;
  704. ff_bgmc_decode(gb, sb_len, current_res,
  705. delta[sb], sx[sb], &high, &low, &value, ctx->bgmc_lut, ctx->bgmc_lut_status);
  706. current_res += sb_len;
  707. }
  708. ff_bgmc_decode_end(gb);
  709. // read least significant bits and tails
  710. current_res = bd->raw_samples + start;
  711. for (sb = 0; sb < sub_blocks; sb++, start = 0) {
  712. unsigned int cur_tail_code = tail_code[sx[sb]][delta[sb]];
  713. unsigned int cur_k = k[sb];
  714. unsigned int cur_s = s[sb];
  715. for (; start < sb_length; start++) {
  716. int32_t res = *current_res;
  717. if (res == cur_tail_code) {
  718. unsigned int max_msb = (2 + (sx[sb] > 2) + (sx[sb] > 10))
  719. << (5 - delta[sb]);
  720. res = decode_rice(gb, cur_s);
  721. if (res >= 0) {
  722. res += (max_msb ) << cur_k;
  723. } else {
  724. res -= (max_msb - 1) << cur_k;
  725. }
  726. } else {
  727. if (res > cur_tail_code)
  728. res--;
  729. if (res & 1)
  730. res = -res;
  731. res >>= 1;
  732. if (cur_k) {
  733. res *= 1U << cur_k;
  734. res |= get_bits_long(gb, cur_k);
  735. }
  736. }
  737. *current_res++ = res;
  738. }
  739. }
  740. } else {
  741. current_res = bd->raw_samples + start;
  742. for (sb = 0; sb < sub_blocks; sb++, start = 0)
  743. for (; start < sb_length; start++)
  744. *current_res++ = decode_rice(gb, s[sb]);
  745. }
  746. return 0;
  747. }
  748. /** Decode the block data for a non-constant block
  749. */
  750. static int decode_var_block_data(ALSDecContext *ctx, ALSBlockData *bd)
  751. {
  752. ALSSpecificConfig *sconf = &ctx->sconf;
  753. unsigned int block_length = bd->block_length;
  754. unsigned int smp = 0;
  755. unsigned int k;
  756. int opt_order = *bd->opt_order;
  757. int sb;
  758. int64_t y;
  759. int32_t *quant_cof = bd->quant_cof;
  760. int32_t *lpc_cof = bd->lpc_cof;
  761. int32_t *raw_samples = bd->raw_samples;
  762. int32_t *raw_samples_end = bd->raw_samples + bd->block_length;
  763. int32_t *lpc_cof_reversed = ctx->lpc_cof_reversed_buffer;
  764. // reverse long-term prediction
  765. if (*bd->use_ltp) {
  766. int ltp_smp;
  767. for (ltp_smp = FFMAX(*bd->ltp_lag - 2, 0); ltp_smp < block_length; ltp_smp++) {
  768. int center = ltp_smp - *bd->ltp_lag;
  769. int begin = FFMAX(0, center - 2);
  770. int end = center + 3;
  771. int tab = 5 - (end - begin);
  772. int base;
  773. y = 1 << 6;
  774. for (base = begin; base < end; base++, tab++)
  775. y += (uint64_t)MUL64(bd->ltp_gain[tab], raw_samples[base]);
  776. raw_samples[ltp_smp] += y >> 7;
  777. }
  778. }
  779. // reconstruct all samples from residuals
  780. if (bd->ra_block) {
  781. for (smp = 0; smp < opt_order; smp++) {
  782. y = 1 << 19;
  783. for (sb = 0; sb < smp; sb++)
  784. y += (uint64_t)MUL64(lpc_cof[sb], raw_samples[-(sb + 1)]);
  785. *raw_samples++ -= y >> 20;
  786. parcor_to_lpc(smp, quant_cof, lpc_cof);
  787. }
  788. } else {
  789. for (k = 0; k < opt_order; k++)
  790. parcor_to_lpc(k, quant_cof, lpc_cof);
  791. // store previous samples in case that they have to be altered
  792. if (*bd->store_prev_samples)
  793. memcpy(bd->prev_raw_samples, raw_samples - sconf->max_order,
  794. sizeof(*bd->prev_raw_samples) * sconf->max_order);
  795. // reconstruct difference signal for prediction (joint-stereo)
  796. if (bd->js_blocks && bd->raw_other) {
  797. uint32_t *left, *right;
  798. if (bd->raw_other > raw_samples) { // D = R - L
  799. left = raw_samples;
  800. right = bd->raw_other;
  801. } else { // D = R - L
  802. left = bd->raw_other;
  803. right = raw_samples;
  804. }
  805. for (sb = -1; sb >= -sconf->max_order; sb--)
  806. raw_samples[sb] = right[sb] - left[sb];
  807. }
  808. // reconstruct shifted signal
  809. if (*bd->shift_lsbs)
  810. for (sb = -1; sb >= -sconf->max_order; sb--)
  811. raw_samples[sb] >>= *bd->shift_lsbs;
  812. }
  813. // reverse linear prediction coefficients for efficiency
  814. lpc_cof = lpc_cof + opt_order;
  815. for (sb = 0; sb < opt_order; sb++)
  816. lpc_cof_reversed[sb] = lpc_cof[-(sb + 1)];
  817. // reconstruct raw samples
  818. raw_samples = bd->raw_samples + smp;
  819. lpc_cof = lpc_cof_reversed + opt_order;
  820. for (; raw_samples < raw_samples_end; raw_samples++) {
  821. y = 1 << 19;
  822. for (sb = -opt_order; sb < 0; sb++)
  823. y += (uint64_t)MUL64(lpc_cof[sb], raw_samples[sb]);
  824. *raw_samples -= y >> 20;
  825. }
  826. raw_samples = bd->raw_samples;
  827. // restore previous samples in case that they have been altered
  828. if (*bd->store_prev_samples)
  829. memcpy(raw_samples - sconf->max_order, bd->prev_raw_samples,
  830. sizeof(*raw_samples) * sconf->max_order);
  831. return 0;
  832. }
  833. /** Read the block data.
  834. */
  835. static int read_block(ALSDecContext *ctx, ALSBlockData *bd)
  836. {
  837. int ret;
  838. GetBitContext *gb = &ctx->gb;
  839. ALSSpecificConfig *sconf = &ctx->sconf;
  840. *bd->shift_lsbs = 0;
  841. // read block type flag and read the samples accordingly
  842. if (get_bits1(gb)) {
  843. ret = read_var_block_data(ctx, bd);
  844. } else {
  845. ret = read_const_block_data(ctx, bd);
  846. }
  847. if (!sconf->mc_coding || ctx->js_switch)
  848. align_get_bits(gb);
  849. return ret;
  850. }
  851. /** Decode the block data.
  852. */
  853. static int decode_block(ALSDecContext *ctx, ALSBlockData *bd)
  854. {
  855. unsigned int smp;
  856. int ret = 0;
  857. // read block type flag and read the samples accordingly
  858. if (*bd->const_block)
  859. decode_const_block_data(ctx, bd);
  860. else
  861. ret = decode_var_block_data(ctx, bd); // always return 0
  862. if (ret < 0)
  863. return ret;
  864. // TODO: read RLSLMS extension data
  865. if (*bd->shift_lsbs)
  866. for (smp = 0; smp < bd->block_length; smp++)
  867. bd->raw_samples[smp] = (unsigned)bd->raw_samples[smp] << *bd->shift_lsbs;
  868. return 0;
  869. }
  870. /** Read and decode block data successively.
  871. */
  872. static int read_decode_block(ALSDecContext *ctx, ALSBlockData *bd)
  873. {
  874. int ret;
  875. if ((ret = read_block(ctx, bd)) < 0)
  876. return ret;
  877. return decode_block(ctx, bd);
  878. }
  879. /** Compute the number of samples left to decode for the current frame and
  880. * sets these samples to zero.
  881. */
  882. static void zero_remaining(unsigned int b, unsigned int b_max,
  883. const unsigned int *div_blocks, int32_t *buf)
  884. {
  885. unsigned int count = 0;
  886. while (b < b_max)
  887. count += div_blocks[b++];
  888. if (count)
  889. memset(buf, 0, sizeof(*buf) * count);
  890. }
  891. /** Decode blocks independently.
  892. */
  893. static int decode_blocks_ind(ALSDecContext *ctx, unsigned int ra_frame,
  894. unsigned int c, const unsigned int *div_blocks,
  895. unsigned int *js_blocks)
  896. {
  897. int ret;
  898. unsigned int b;
  899. ALSBlockData bd = { 0 };
  900. bd.ra_block = ra_frame;
  901. bd.const_block = ctx->const_block;
  902. bd.shift_lsbs = ctx->shift_lsbs;
  903. bd.opt_order = ctx->opt_order;
  904. bd.store_prev_samples = ctx->store_prev_samples;
  905. bd.use_ltp = ctx->use_ltp;
  906. bd.ltp_lag = ctx->ltp_lag;
  907. bd.ltp_gain = ctx->ltp_gain[0];
  908. bd.quant_cof = ctx->quant_cof[0];
  909. bd.lpc_cof = ctx->lpc_cof[0];
  910. bd.prev_raw_samples = ctx->prev_raw_samples;
  911. bd.raw_samples = ctx->raw_samples[c];
  912. for (b = 0; b < ctx->num_blocks; b++) {
  913. bd.block_length = div_blocks[b];
  914. if ((ret = read_decode_block(ctx, &bd)) < 0) {
  915. // damaged block, write zero for the rest of the frame
  916. zero_remaining(b, ctx->num_blocks, div_blocks, bd.raw_samples);
  917. return ret;
  918. }
  919. bd.raw_samples += div_blocks[b];
  920. bd.ra_block = 0;
  921. }
  922. return 0;
  923. }
  924. /** Decode blocks dependently.
  925. */
  926. static int decode_blocks(ALSDecContext *ctx, unsigned int ra_frame,
  927. unsigned int c, const unsigned int *div_blocks,
  928. unsigned int *js_blocks)
  929. {
  930. ALSSpecificConfig *sconf = &ctx->sconf;
  931. unsigned int offset = 0;
  932. unsigned int b;
  933. int ret;
  934. ALSBlockData bd[2] = { { 0 } };
  935. bd[0].ra_block = ra_frame;
  936. bd[0].const_block = ctx->const_block;
  937. bd[0].shift_lsbs = ctx->shift_lsbs;
  938. bd[0].opt_order = ctx->opt_order;
  939. bd[0].store_prev_samples = ctx->store_prev_samples;
  940. bd[0].use_ltp = ctx->use_ltp;
  941. bd[0].ltp_lag = ctx->ltp_lag;
  942. bd[0].ltp_gain = ctx->ltp_gain[0];
  943. bd[0].quant_cof = ctx->quant_cof[0];
  944. bd[0].lpc_cof = ctx->lpc_cof[0];
  945. bd[0].prev_raw_samples = ctx->prev_raw_samples;
  946. bd[0].js_blocks = *js_blocks;
  947. bd[1].ra_block = ra_frame;
  948. bd[1].const_block = ctx->const_block;
  949. bd[1].shift_lsbs = ctx->shift_lsbs;
  950. bd[1].opt_order = ctx->opt_order;
  951. bd[1].store_prev_samples = ctx->store_prev_samples;
  952. bd[1].use_ltp = ctx->use_ltp;
  953. bd[1].ltp_lag = ctx->ltp_lag;
  954. bd[1].ltp_gain = ctx->ltp_gain[0];
  955. bd[1].quant_cof = ctx->quant_cof[0];
  956. bd[1].lpc_cof = ctx->lpc_cof[0];
  957. bd[1].prev_raw_samples = ctx->prev_raw_samples;
  958. bd[1].js_blocks = *(js_blocks + 1);
  959. // decode all blocks
  960. for (b = 0; b < ctx->num_blocks; b++) {
  961. unsigned int s;
  962. bd[0].block_length = div_blocks[b];
  963. bd[1].block_length = div_blocks[b];
  964. bd[0].raw_samples = ctx->raw_samples[c ] + offset;
  965. bd[1].raw_samples = ctx->raw_samples[c + 1] + offset;
  966. bd[0].raw_other = bd[1].raw_samples;
  967. bd[1].raw_other = bd[0].raw_samples;
  968. if ((ret = read_decode_block(ctx, &bd[0])) < 0 ||
  969. (ret = read_decode_block(ctx, &bd[1])) < 0)
  970. goto fail;
  971. // reconstruct joint-stereo blocks
  972. if (bd[0].js_blocks) {
  973. if (bd[1].js_blocks)
  974. av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel pair.\n");
  975. for (s = 0; s < div_blocks[b]; s++)
  976. bd[0].raw_samples[s] = bd[1].raw_samples[s] - (unsigned)bd[0].raw_samples[s];
  977. } else if (bd[1].js_blocks) {
  978. for (s = 0; s < div_blocks[b]; s++)
  979. bd[1].raw_samples[s] = bd[1].raw_samples[s] + (unsigned)bd[0].raw_samples[s];
  980. }
  981. offset += div_blocks[b];
  982. bd[0].ra_block = 0;
  983. bd[1].ra_block = 0;
  984. }
  985. // store carryover raw samples,
  986. // the others channel raw samples are stored by the calling function.
  987. memmove(ctx->raw_samples[c] - sconf->max_order,
  988. ctx->raw_samples[c] - sconf->max_order + sconf->frame_length,
  989. sizeof(*ctx->raw_samples[c]) * sconf->max_order);
  990. return 0;
  991. fail:
  992. // damaged block, write zero for the rest of the frame
  993. zero_remaining(b, ctx->num_blocks, div_blocks, bd[0].raw_samples);
  994. zero_remaining(b, ctx->num_blocks, div_blocks, bd[1].raw_samples);
  995. return ret;
  996. }
  997. static inline int als_weighting(GetBitContext *gb, int k, int off)
  998. {
  999. int idx = av_clip(decode_rice(gb, k) + off,
  1000. 0, FF_ARRAY_ELEMS(mcc_weightings) - 1);
  1001. return mcc_weightings[idx];
  1002. }
  1003. /** Read the channel data.
  1004. */
  1005. static int read_channel_data(ALSDecContext *ctx, ALSChannelData *cd, int c)
  1006. {
  1007. GetBitContext *gb = &ctx->gb;
  1008. ALSChannelData *current = cd;
  1009. unsigned int channels = ctx->avctx->channels;
  1010. int entries = 0;
  1011. while (entries < channels && !(current->stop_flag = get_bits1(gb))) {
  1012. current->master_channel = get_bits_long(gb, av_ceil_log2(channels));
  1013. if (current->master_channel >= channels) {
  1014. av_log(ctx->avctx, AV_LOG_ERROR, "Invalid master channel.\n");
  1015. return AVERROR_INVALIDDATA;
  1016. }
  1017. if (current->master_channel != c) {
  1018. current->time_diff_flag = get_bits1(gb);
  1019. current->weighting[0] = als_weighting(gb, 1, 16);
  1020. current->weighting[1] = als_weighting(gb, 2, 14);
  1021. current->weighting[2] = als_weighting(gb, 1, 16);
  1022. if (current->time_diff_flag) {
  1023. current->weighting[3] = als_weighting(gb, 1, 16);
  1024. current->weighting[4] = als_weighting(gb, 1, 16);
  1025. current->weighting[5] = als_weighting(gb, 1, 16);
  1026. current->time_diff_sign = get_bits1(gb);
  1027. current->time_diff_index = get_bits(gb, ctx->ltp_lag_length - 3) + 3;
  1028. }
  1029. }
  1030. current++;
  1031. entries++;
  1032. }
  1033. if (entries == channels) {
  1034. av_log(ctx->avctx, AV_LOG_ERROR, "Damaged channel data.\n");
  1035. return AVERROR_INVALIDDATA;
  1036. }
  1037. align_get_bits(gb);
  1038. return 0;
  1039. }
  1040. /** Recursively reverts the inter-channel correlation for a block.
  1041. */
  1042. static int revert_channel_correlation(ALSDecContext *ctx, ALSBlockData *bd,
  1043. ALSChannelData **cd, int *reverted,
  1044. unsigned int offset, int c)
  1045. {
  1046. ALSChannelData *ch = cd[c];
  1047. unsigned int dep = 0;
  1048. unsigned int channels = ctx->avctx->channels;
  1049. unsigned int channel_size = ctx->sconf.frame_length + ctx->sconf.max_order;
  1050. if (reverted[c])
  1051. return 0;
  1052. reverted[c] = 1;
  1053. while (dep < channels && !ch[dep].stop_flag) {
  1054. revert_channel_correlation(ctx, bd, cd, reverted, offset,
  1055. ch[dep].master_channel);
  1056. dep++;
  1057. }
  1058. if (dep == channels) {
  1059. av_log(ctx->avctx, AV_LOG_WARNING, "Invalid channel correlation.\n");
  1060. return AVERROR_INVALIDDATA;
  1061. }
  1062. bd->const_block = ctx->const_block + c;
  1063. bd->shift_lsbs = ctx->shift_lsbs + c;
  1064. bd->opt_order = ctx->opt_order + c;
  1065. bd->store_prev_samples = ctx->store_prev_samples + c;
  1066. bd->use_ltp = ctx->use_ltp + c;
  1067. bd->ltp_lag = ctx->ltp_lag + c;
  1068. bd->ltp_gain = ctx->ltp_gain[c];
  1069. bd->lpc_cof = ctx->lpc_cof[c];
  1070. bd->quant_cof = ctx->quant_cof[c];
  1071. bd->raw_samples = ctx->raw_samples[c] + offset;
  1072. for (dep = 0; !ch[dep].stop_flag; dep++) {
  1073. ptrdiff_t smp;
  1074. ptrdiff_t begin = 1;
  1075. ptrdiff_t end = bd->block_length - 1;
  1076. int64_t y;
  1077. int32_t *master = ctx->raw_samples[ch[dep].master_channel] + offset;
  1078. if (ch[dep].master_channel == c)
  1079. continue;
  1080. if (ch[dep].time_diff_flag) {
  1081. int t = ch[dep].time_diff_index;
  1082. if (ch[dep].time_diff_sign) {
  1083. t = -t;
  1084. if (begin < t) {
  1085. av_log(ctx->avctx, AV_LOG_ERROR, "begin %"PTRDIFF_SPECIFIER" smaller than time diff index %d.\n", begin, t);
  1086. return AVERROR_INVALIDDATA;
  1087. }
  1088. begin -= t;
  1089. } else {
  1090. if (end < t) {
  1091. av_log(ctx->avctx, AV_LOG_ERROR, "end %"PTRDIFF_SPECIFIER" smaller than time diff index %d.\n", end, t);
  1092. return AVERROR_INVALIDDATA;
  1093. }
  1094. end -= t;
  1095. }
  1096. if (FFMIN(begin - 1, begin - 1 + t) < ctx->raw_buffer - master ||
  1097. FFMAX(end + 1, end + 1 + t) > ctx->raw_buffer + channels * channel_size - master) {
  1098. av_log(ctx->avctx, AV_LOG_ERROR,
  1099. "sample pointer range [%p, %p] not contained in raw_buffer [%p, %p].\n",
  1100. master + FFMIN(begin - 1, begin - 1 + t), master + FFMAX(end + 1, end + 1 + t),
  1101. ctx->raw_buffer, ctx->raw_buffer + channels * channel_size);
  1102. return AVERROR_INVALIDDATA;
  1103. }
  1104. for (smp = begin; smp < end; smp++) {
  1105. y = (1 << 6) +
  1106. MUL64(ch[dep].weighting[0], master[smp - 1 ]) +
  1107. MUL64(ch[dep].weighting[1], master[smp ]) +
  1108. MUL64(ch[dep].weighting[2], master[smp + 1 ]) +
  1109. MUL64(ch[dep].weighting[3], master[smp - 1 + t]) +
  1110. MUL64(ch[dep].weighting[4], master[smp + t]) +
  1111. MUL64(ch[dep].weighting[5], master[smp + 1 + t]);
  1112. bd->raw_samples[smp] += y >> 7;
  1113. }
  1114. } else {
  1115. if (begin - 1 < ctx->raw_buffer - master ||
  1116. end + 1 > ctx->raw_buffer + channels * channel_size - master) {
  1117. av_log(ctx->avctx, AV_LOG_ERROR,
  1118. "sample pointer range [%p, %p] not contained in raw_buffer [%p, %p].\n",
  1119. master + begin - 1, master + end + 1,
  1120. ctx->raw_buffer, ctx->raw_buffer + channels * channel_size);
  1121. return AVERROR_INVALIDDATA;
  1122. }
  1123. for (smp = begin; smp < end; smp++) {
  1124. y = (1 << 6) +
  1125. MUL64(ch[dep].weighting[0], master[smp - 1]) +
  1126. MUL64(ch[dep].weighting[1], master[smp ]) +
  1127. MUL64(ch[dep].weighting[2], master[smp + 1]);
  1128. bd->raw_samples[smp] += y >> 7;
  1129. }
  1130. }
  1131. }
  1132. return 0;
  1133. }
  1134. /** multiply two softfloats and handle the rounding off
  1135. */
  1136. static SoftFloat_IEEE754 multiply(SoftFloat_IEEE754 a, SoftFloat_IEEE754 b) {
  1137. uint64_t mantissa_temp;
  1138. uint64_t mask_64;
  1139. int cutoff_bit_count;
  1140. unsigned char last_2_bits;
  1141. unsigned int mantissa;
  1142. int32_t sign;
  1143. uint32_t return_val = 0;
  1144. int bit_count = 48;
  1145. sign = a.sign ^ b.sign;
  1146. // Multiply mantissa bits in a 64-bit register
  1147. mantissa_temp = (uint64_t)a.mant * (uint64_t)b.mant;
  1148. mask_64 = (uint64_t)0x1 << 47;
  1149. if (!mantissa_temp)
  1150. return FLOAT_0;
  1151. // Count the valid bit count
  1152. while (!(mantissa_temp & mask_64) && mask_64) {
  1153. bit_count--;
  1154. mask_64 >>= 1;
  1155. }
  1156. // Round off
  1157. cutoff_bit_count = bit_count - 24;
  1158. if (cutoff_bit_count > 0) {
  1159. last_2_bits = (unsigned char)(((unsigned int)mantissa_temp >> (cutoff_bit_count - 1)) & 0x3 );
  1160. if ((last_2_bits == 0x3) || ((last_2_bits == 0x1) && ((unsigned int)mantissa_temp & ((0x1UL << (cutoff_bit_count - 1)) - 1)))) {
  1161. // Need to round up
  1162. mantissa_temp += (uint64_t)0x1 << cutoff_bit_count;
  1163. }
  1164. }
  1165. if (cutoff_bit_count >= 0) {
  1166. mantissa = (unsigned int)(mantissa_temp >> cutoff_bit_count);
  1167. } else {
  1168. mantissa = (unsigned int)(mantissa_temp <<-cutoff_bit_count);
  1169. }
  1170. // Need one more shift?
  1171. if (mantissa & 0x01000000ul) {
  1172. bit_count++;
  1173. mantissa >>= 1;
  1174. }
  1175. if (!sign) {
  1176. return_val = 0x80000000U;
  1177. }
  1178. return_val |= ((unsigned)av_clip(a.exp + b.exp + bit_count - 47, -126, 127) << 23) & 0x7F800000;
  1179. return_val |= mantissa;
  1180. return av_bits2sf_ieee754(return_val);
  1181. }
  1182. /** Read and decode the floating point sample data
  1183. */
  1184. static int read_diff_float_data(ALSDecContext *ctx, unsigned int ra_frame) {
  1185. AVCodecContext *avctx = ctx->avctx;
  1186. GetBitContext *gb = &ctx->gb;
  1187. SoftFloat_IEEE754 *acf = ctx->acf;
  1188. int *shift_value = ctx->shift_value;
  1189. int *last_shift_value = ctx->last_shift_value;
  1190. int *last_acf_mantissa = ctx->last_acf_mantissa;
  1191. int **raw_mantissa = ctx->raw_mantissa;
  1192. int *nbits = ctx->nbits;
  1193. unsigned char *larray = ctx->larray;
  1194. int frame_length = ctx->cur_frame_length;
  1195. SoftFloat_IEEE754 scale = av_int2sf_ieee754(0x1u, 23);
  1196. unsigned int partA_flag;
  1197. unsigned int highest_byte;
  1198. unsigned int shift_amp;
  1199. uint32_t tmp_32;
  1200. int use_acf;
  1201. int nchars;
  1202. int i;
  1203. int c;
  1204. long k;
  1205. long nbits_aligned;
  1206. unsigned long acc;
  1207. unsigned long j;
  1208. uint32_t sign;
  1209. uint32_t e;
  1210. uint32_t mantissa;
  1211. skip_bits_long(gb, 32); //num_bytes_diff_float
  1212. use_acf = get_bits1(gb);
  1213. if (ra_frame) {
  1214. memset(last_acf_mantissa, 0, avctx->channels * sizeof(*last_acf_mantissa));
  1215. memset(last_shift_value, 0, avctx->channels * sizeof(*last_shift_value) );
  1216. ff_mlz_flush_dict(ctx->mlz);
  1217. }
  1218. for (c = 0; c < avctx->channels; ++c) {
  1219. if (use_acf) {
  1220. //acf_flag
  1221. if (get_bits1(gb)) {
  1222. tmp_32 = get_bits(gb, 23);
  1223. last_acf_mantissa[c] = tmp_32;
  1224. } else {
  1225. tmp_32 = last_acf_mantissa[c];
  1226. }
  1227. acf[c] = av_bits2sf_ieee754(tmp_32);
  1228. } else {
  1229. acf[c] = FLOAT_1;
  1230. }
  1231. highest_byte = get_bits(gb, 2);
  1232. partA_flag = get_bits1(gb);
  1233. shift_amp = get_bits1(gb);
  1234. if (shift_amp) {
  1235. shift_value[c] = get_bits(gb, 8);
  1236. last_shift_value[c] = shift_value[c];
  1237. } else {
  1238. shift_value[c] = last_shift_value[c];
  1239. }
  1240. if (partA_flag) {
  1241. if (!get_bits1(gb)) { //uncompressed
  1242. for (i = 0; i < frame_length; ++i) {
  1243. if (ctx->raw_samples[c][i] == 0) {
  1244. ctx->raw_mantissa[c][i] = get_bits_long(gb, 32);
  1245. }
  1246. }
  1247. } else { //compressed
  1248. nchars = 0;
  1249. for (i = 0; i < frame_length; ++i) {
  1250. if (ctx->raw_samples[c][i] == 0) {
  1251. nchars += 4;
  1252. }
  1253. }
  1254. tmp_32 = ff_mlz_decompression(ctx->mlz, gb, nchars, larray);
  1255. if(tmp_32 != nchars) {
  1256. av_log(ctx->avctx, AV_LOG_ERROR, "Error in MLZ decompression (%"PRId32", %d).\n", tmp_32, nchars);
  1257. return AVERROR_INVALIDDATA;
  1258. }
  1259. for (i = 0; i < frame_length; ++i) {
  1260. ctx->raw_mantissa[c][i] = AV_RB32(larray);
  1261. }
  1262. }
  1263. }
  1264. //decode part B
  1265. if (highest_byte) {
  1266. for (i = 0; i < frame_length; ++i) {
  1267. if (ctx->raw_samples[c][i] != 0) {
  1268. //The following logic is taken from Tabel 14.45 and 14.46 from the ISO spec
  1269. if (av_cmp_sf_ieee754(acf[c], FLOAT_1)) {
  1270. nbits[i] = 23 - av_log2(abs(ctx->raw_samples[c][i]));
  1271. } else {
  1272. nbits[i] = 23;
  1273. }
  1274. nbits[i] = FFMIN(nbits[i], highest_byte*8);
  1275. }
  1276. }
  1277. if (!get_bits1(gb)) { //uncompressed
  1278. for (i = 0; i < frame_length; ++i) {
  1279. if (ctx->raw_samples[c][i] != 0) {
  1280. raw_mantissa[c][i] = get_bitsz(gb, nbits[i]);
  1281. }
  1282. }
  1283. } else { //compressed
  1284. nchars = 0;
  1285. for (i = 0; i < frame_length; ++i) {
  1286. if (ctx->raw_samples[c][i]) {
  1287. nchars += (int) nbits[i] / 8;
  1288. if (nbits[i] & 7) {
  1289. ++nchars;
  1290. }
  1291. }
  1292. }
  1293. tmp_32 = ff_mlz_decompression(ctx->mlz, gb, nchars, larray);
  1294. if(tmp_32 != nchars) {
  1295. av_log(ctx->avctx, AV_LOG_ERROR, "Error in MLZ decompression (%"PRId32", %d).\n", tmp_32, nchars);
  1296. return AVERROR_INVALIDDATA;
  1297. }
  1298. j = 0;
  1299. for (i = 0; i < frame_length; ++i) {
  1300. if (ctx->raw_samples[c][i]) {
  1301. if (nbits[i] & 7) {
  1302. nbits_aligned = 8 * ((unsigned int)(nbits[i] / 8) + 1);
  1303. } else {
  1304. nbits_aligned = nbits[i];
  1305. }
  1306. acc = 0;
  1307. for (k = 0; k < nbits_aligned/8; ++k) {
  1308. acc = (acc << 8) + larray[j++];
  1309. }
  1310. acc >>= (nbits_aligned - nbits[i]);
  1311. raw_mantissa[c][i] = acc;
  1312. }
  1313. }
  1314. }
  1315. }
  1316. for (i = 0; i < frame_length; ++i) {
  1317. SoftFloat_IEEE754 pcm_sf = av_int2sf_ieee754(ctx->raw_samples[c][i], 0);
  1318. pcm_sf = av_div_sf_ieee754(pcm_sf, scale);
  1319. if (ctx->raw_samples[c][i] != 0) {
  1320. if (!av_cmp_sf_ieee754(acf[c], FLOAT_1)) {
  1321. pcm_sf = multiply(acf[c], pcm_sf);
  1322. }
  1323. sign = pcm_sf.sign;
  1324. e = pcm_sf.exp;
  1325. mantissa = (pcm_sf.mant | 0x800000) + raw_mantissa[c][i];
  1326. while(mantissa >= 0x1000000) {
  1327. e++;
  1328. mantissa >>= 1;
  1329. }
  1330. if (mantissa) e += (shift_value[c] - 127);
  1331. mantissa &= 0x007fffffUL;
  1332. tmp_32 = (sign << 31) | ((e + EXP_BIAS) << 23) | (mantissa);
  1333. ctx->raw_samples[c][i] = tmp_32;
  1334. } else {
  1335. ctx->raw_samples[c][i] = raw_mantissa[c][i] & 0x007fffffUL;
  1336. }
  1337. }
  1338. align_get_bits(gb);
  1339. }
  1340. return 0;
  1341. }
  1342. /** Read the frame data.
  1343. */
  1344. static int read_frame_data(ALSDecContext *ctx, unsigned int ra_frame)
  1345. {
  1346. ALSSpecificConfig *sconf = &ctx->sconf;
  1347. AVCodecContext *avctx = ctx->avctx;
  1348. GetBitContext *gb = &ctx->gb;
  1349. unsigned int div_blocks[32]; ///< block sizes.
  1350. unsigned int c;
  1351. unsigned int js_blocks[2];
  1352. uint32_t bs_info = 0;
  1353. int ret;
  1354. // skip the size of the ra unit if present in the frame
  1355. if (sconf->ra_flag == RA_FLAG_FRAMES && ra_frame)
  1356. skip_bits_long(gb, 32);
  1357. if (sconf->mc_coding && sconf->joint_stereo) {
  1358. ctx->js_switch = get_bits1(gb);
  1359. align_get_bits(gb);
  1360. }
  1361. if (!sconf->mc_coding || ctx->js_switch) {
  1362. int independent_bs = !sconf->joint_stereo;
  1363. for (c = 0; c < avctx->channels; c++) {
  1364. js_blocks[0] = 0;
  1365. js_blocks[1] = 0;
  1366. get_block_sizes(ctx, div_blocks, &bs_info);
  1367. // if joint_stereo and block_switching is set, independent decoding
  1368. // is signaled via the first bit of bs_info
  1369. if (sconf->joint_stereo && sconf->block_switching)
  1370. if (bs_info >> 31)
  1371. independent_bs = 2;
  1372. // if this is the last channel, it has to be decoded independently
  1373. if (c == avctx->channels - 1 || (c & 1))
  1374. independent_bs = 1;
  1375. if (independent_bs) {
  1376. ret = decode_blocks_ind(ctx, ra_frame, c,
  1377. div_blocks, js_blocks);
  1378. if (ret < 0)
  1379. return ret;
  1380. independent_bs--;
  1381. } else {
  1382. ret = decode_blocks(ctx, ra_frame, c, div_blocks, js_blocks);
  1383. if (ret < 0)
  1384. return ret;
  1385. c++;
  1386. }
  1387. // store carryover raw samples
  1388. memmove(ctx->raw_samples[c] - sconf->max_order,
  1389. ctx->raw_samples[c] - sconf->max_order + sconf->frame_length,
  1390. sizeof(*ctx->raw_samples[c]) * sconf->max_order);
  1391. }
  1392. } else { // multi-channel coding
  1393. ALSBlockData bd = { 0 };
  1394. int b, ret;
  1395. int *reverted_channels = ctx->reverted_channels;
  1396. unsigned int offset = 0;
  1397. for (c = 0; c < avctx->channels; c++)
  1398. if (ctx->chan_data[c] < ctx->chan_data_buffer) {
  1399. av_log(ctx->avctx, AV_LOG_ERROR, "Invalid channel data.\n");
  1400. return AVERROR_INVALIDDATA;
  1401. }
  1402. memset(reverted_channels, 0, sizeof(*reverted_channels) * avctx->channels);
  1403. bd.ra_block = ra_frame;
  1404. bd.prev_raw_samples = ctx->prev_raw_samples;
  1405. get_block_sizes(ctx, div_blocks, &bs_info);
  1406. for (b = 0; b < ctx->num_blocks; b++) {
  1407. bd.block_length = div_blocks[b];
  1408. if (bd.block_length <= 0) {
  1409. av_log(ctx->avctx, AV_LOG_WARNING,
  1410. "Invalid block length %u in channel data!\n",
  1411. bd.block_length);
  1412. continue;
  1413. }
  1414. for (c = 0; c < avctx->channels; c++) {
  1415. bd.const_block = ctx->const_block + c;
  1416. bd.shift_lsbs = ctx->shift_lsbs + c;
  1417. bd.opt_order = ctx->opt_order + c;
  1418. bd.store_prev_samples = ctx->store_prev_samples + c;
  1419. bd.use_ltp = ctx->use_ltp + c;
  1420. bd.ltp_lag = ctx->ltp_lag + c;
  1421. bd.ltp_gain = ctx->ltp_gain[c];
  1422. bd.lpc_cof = ctx->lpc_cof[c];
  1423. bd.quant_cof = ctx->quant_cof[c];
  1424. bd.raw_samples = ctx->raw_samples[c] + offset;
  1425. bd.raw_other = NULL;
  1426. if ((ret = read_block(ctx, &bd)) < 0)
  1427. return ret;
  1428. if ((ret = read_channel_data(ctx, ctx->chan_data[c], c)) < 0)
  1429. return ret;
  1430. }
  1431. for (c = 0; c < avctx->channels; c++) {
  1432. ret = revert_channel_correlation(ctx, &bd, ctx->chan_data,
  1433. reverted_channels, offset, c);
  1434. if (ret < 0)
  1435. return ret;
  1436. }
  1437. for (c = 0; c < avctx->channels; c++) {
  1438. bd.const_block = ctx->const_block + c;
  1439. bd.shift_lsbs = ctx->shift_lsbs + c;
  1440. bd.opt_order = ctx->opt_order + c;
  1441. bd.store_prev_samples = ctx->store_prev_samples + c;
  1442. bd.use_ltp = ctx->use_ltp + c;
  1443. bd.ltp_lag = ctx->ltp_lag + c;
  1444. bd.ltp_gain = ctx->ltp_gain[c];
  1445. bd.lpc_cof = ctx->lpc_cof[c];
  1446. bd.quant_cof = ctx->quant_cof[c];
  1447. bd.raw_samples = ctx->raw_samples[c] + offset;
  1448. if ((ret = decode_block(ctx, &bd)) < 0)
  1449. return ret;
  1450. }
  1451. memset(reverted_channels, 0, avctx->channels * sizeof(*reverted_channels));
  1452. offset += div_blocks[b];
  1453. bd.ra_block = 0;
  1454. }
  1455. // store carryover raw samples
  1456. for (c = 0; c < avctx->channels; c++)
  1457. memmove(ctx->raw_samples[c] - sconf->max_order,
  1458. ctx->raw_samples[c] - sconf->max_order + sconf->frame_length,
  1459. sizeof(*ctx->raw_samples[c]) * sconf->max_order);
  1460. }
  1461. if (sconf->floating) {
  1462. read_diff_float_data(ctx, ra_frame);
  1463. }
  1464. if (get_bits_left(gb) < 0) {
  1465. av_log(ctx->avctx, AV_LOG_ERROR, "Overread %d\n", -get_bits_left(gb));
  1466. return AVERROR_INVALIDDATA;
  1467. }
  1468. return 0;
  1469. }
  1470. /** Decode an ALS frame.
  1471. */
  1472. static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr,
  1473. AVPacket *avpkt)
  1474. {
  1475. ALSDecContext *ctx = avctx->priv_data;
  1476. AVFrame *frame = data;
  1477. ALSSpecificConfig *sconf = &ctx->sconf;
  1478. const uint8_t *buffer = avpkt->data;
  1479. int buffer_size = avpkt->size;
  1480. int invalid_frame, ret;
  1481. unsigned int c, sample, ra_frame, bytes_read, shift;
  1482. if ((ret = init_get_bits8(&ctx->gb, buffer, buffer_size)) < 0)
  1483. return ret;
  1484. // In the case that the distance between random access frames is set to zero
  1485. // (sconf->ra_distance == 0) no frame is treated as a random access frame.
  1486. // For the first frame, if prediction is used, all samples used from the
  1487. // previous frame are assumed to be zero.
  1488. ra_frame = sconf->ra_distance && !(ctx->frame_id % sconf->ra_distance);
  1489. // the last frame to decode might have a different length
  1490. if (sconf->samples != 0xFFFFFFFF)
  1491. ctx->cur_frame_length = FFMIN(sconf->samples - ctx->frame_id * (uint64_t) sconf->frame_length,
  1492. sconf->frame_length);
  1493. else
  1494. ctx->cur_frame_length = sconf->frame_length;
  1495. // decode the frame data
  1496. if ((invalid_frame = read_frame_data(ctx, ra_frame)) < 0)
  1497. av_log(ctx->avctx, AV_LOG_WARNING,
  1498. "Reading frame data failed. Skipping RA unit.\n");
  1499. ctx->frame_id++;
  1500. /* get output buffer */
  1501. frame->nb_samples = ctx->cur_frame_length;
  1502. if ((ret = ff_get_buffer(avctx, frame, 0)) < 0)
  1503. return ret;
  1504. // transform decoded frame into output format
  1505. #define INTERLEAVE_OUTPUT(bps) \
  1506. { \
  1507. int##bps##_t *dest = (int##bps##_t*)frame->data[0]; \
  1508. int channels = avctx->channels; \
  1509. int32_t **raw_samples = ctx->raw_samples; \
  1510. shift = bps - ctx->avctx->bits_per_raw_sample; \
  1511. if (!ctx->cs_switch) { \
  1512. for (sample = 0; sample < ctx->cur_frame_length; sample++) \
  1513. for (c = 0; c < channels; c++) \
  1514. *dest++ = raw_samples[c][sample] * (1U << shift); \
  1515. } else { \
  1516. for (sample = 0; sample < ctx->cur_frame_length; sample++) \
  1517. for (c = 0; c < channels; c++) \
  1518. *dest++ = raw_samples[sconf->chan_pos[c]][sample] * (1U << shift);\
  1519. } \
  1520. }
  1521. if (ctx->avctx->bits_per_raw_sample <= 16) {
  1522. INTERLEAVE_OUTPUT(16)
  1523. } else {
  1524. INTERLEAVE_OUTPUT(32)
  1525. }
  1526. // update CRC
  1527. if (sconf->crc_enabled && (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL))) {
  1528. int swap = HAVE_BIGENDIAN != sconf->msb_first;
  1529. if (ctx->avctx->bits_per_raw_sample == 24) {
  1530. int32_t *src = (int32_t *)frame->data[0];
  1531. for (sample = 0;
  1532. sample < ctx->cur_frame_length * avctx->channels;
  1533. sample++) {
  1534. int32_t v;
  1535. if (swap)
  1536. v = av_bswap32(src[sample]);
  1537. else
  1538. v = src[sample];
  1539. if (!HAVE_BIGENDIAN)
  1540. v >>= 8;
  1541. ctx->crc = av_crc(ctx->crc_table, ctx->crc, (uint8_t*)(&v), 3);
  1542. }
  1543. } else {
  1544. uint8_t *crc_source;
  1545. if (swap) {
  1546. if (ctx->avctx->bits_per_raw_sample <= 16) {
  1547. int16_t *src = (int16_t*) frame->data[0];
  1548. int16_t *dest = (int16_t*) ctx->crc_buffer;
  1549. for (sample = 0;
  1550. sample < ctx->cur_frame_length * avctx->channels;
  1551. sample++)
  1552. *dest++ = av_bswap16(src[sample]);
  1553. } else {
  1554. ctx->bdsp.bswap_buf((uint32_t *) ctx->crc_buffer,
  1555. (uint32_t *) frame->data[0],
  1556. ctx->cur_frame_length * avctx->channels);
  1557. }
  1558. crc_source = ctx->crc_buffer;
  1559. } else {
  1560. crc_source = frame->data[0];
  1561. }
  1562. ctx->crc = av_crc(ctx->crc_table, ctx->crc, crc_source,
  1563. ctx->cur_frame_length * avctx->channels *
  1564. av_get_bytes_per_sample(avctx->sample_fmt));
  1565. }
  1566. // check CRC sums if this is the last frame
  1567. if (ctx->cur_frame_length != sconf->frame_length &&
  1568. ctx->crc_org != ctx->crc) {
  1569. av_log(avctx, AV_LOG_ERROR, "CRC error.\n");
  1570. if (avctx->err_recognition & AV_EF_EXPLODE)
  1571. return AVERROR_INVALIDDATA;
  1572. }
  1573. }
  1574. *got_frame_ptr = 1;
  1575. bytes_read = invalid_frame ? buffer_size :
  1576. (get_bits_count(&ctx->gb) + 7) >> 3;
  1577. return bytes_read;
  1578. }
  1579. /** Uninitialize the ALS decoder.
  1580. */
  1581. static av_cold int decode_end(AVCodecContext *avctx)
  1582. {
  1583. ALSDecContext *ctx = avctx->priv_data;
  1584. int i;
  1585. av_freep(&ctx->sconf.chan_pos);
  1586. ff_bgmc_end(&ctx->bgmc_lut, &ctx->bgmc_lut_status);
  1587. av_freep(&ctx->const_block);
  1588. av_freep(&ctx->shift_lsbs);
  1589. av_freep(&ctx->opt_order);
  1590. av_freep(&ctx->store_prev_samples);
  1591. av_freep(&ctx->use_ltp);
  1592. av_freep(&ctx->ltp_lag);
  1593. av_freep(&ctx->ltp_gain);
  1594. av_freep(&ctx->ltp_gain_buffer);
  1595. av_freep(&ctx->quant_cof);
  1596. av_freep(&ctx->lpc_cof);
  1597. av_freep(&ctx->quant_cof_buffer);
  1598. av_freep(&ctx->lpc_cof_buffer);
  1599. av_freep(&ctx->lpc_cof_reversed_buffer);
  1600. av_freep(&ctx->prev_raw_samples);
  1601. av_freep(&ctx->raw_samples);
  1602. av_freep(&ctx->raw_buffer);
  1603. av_freep(&ctx->chan_data);
  1604. av_freep(&ctx->chan_data_buffer);
  1605. av_freep(&ctx->reverted_channels);
  1606. av_freep(&ctx->crc_buffer);
  1607. if (ctx->mlz) {
  1608. av_freep(&ctx->mlz->dict);
  1609. av_freep(&ctx->mlz);
  1610. }
  1611. av_freep(&ctx->acf);
  1612. av_freep(&ctx->last_acf_mantissa);
  1613. av_freep(&ctx->shift_value);
  1614. av_freep(&ctx->last_shift_value);
  1615. if (ctx->raw_mantissa) {
  1616. for (i = 0; i < avctx->channels; i++) {
  1617. av_freep(&ctx->raw_mantissa[i]);
  1618. }
  1619. av_freep(&ctx->raw_mantissa);
  1620. }
  1621. av_freep(&ctx->larray);
  1622. av_freep(&ctx->nbits);
  1623. return 0;
  1624. }
  1625. /** Initialize the ALS decoder.
  1626. */
  1627. static av_cold int decode_init(AVCodecContext *avctx)
  1628. {
  1629. unsigned int c;
  1630. unsigned int channel_size;
  1631. int num_buffers, ret;
  1632. ALSDecContext *ctx = avctx->priv_data;
  1633. ALSSpecificConfig *sconf = &ctx->sconf;
  1634. ctx->avctx = avctx;
  1635. if (!avctx->extradata) {
  1636. av_log(avctx, AV_LOG_ERROR, "Missing required ALS extradata.\n");
  1637. return AVERROR_INVALIDDATA;
  1638. }
  1639. if ((ret = read_specific_config(ctx)) < 0) {
  1640. av_log(avctx, AV_LOG_ERROR, "Reading ALSSpecificConfig failed.\n");
  1641. goto fail;
  1642. }
  1643. if ((ret = check_specific_config(ctx)) < 0) {
  1644. goto fail;
  1645. }
  1646. if (sconf->bgmc) {
  1647. ret = ff_bgmc_init(avctx, &ctx->bgmc_lut, &ctx->bgmc_lut_status);
  1648. if (ret < 0)
  1649. goto fail;
  1650. }
  1651. if (sconf->floating) {
  1652. avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
  1653. avctx->bits_per_raw_sample = 32;
  1654. } else {
  1655. avctx->sample_fmt = sconf->resolution > 1
  1656. ? AV_SAMPLE_FMT_S32 : AV_SAMPLE_FMT_S16;
  1657. avctx->bits_per_raw_sample = (sconf->resolution + 1) * 8;
  1658. if (avctx->bits_per_raw_sample > 32) {
  1659. av_log(avctx, AV_LOG_ERROR, "Bits per raw sample %d larger than 32.\n",
  1660. avctx->bits_per_raw_sample);
  1661. ret = AVERROR_INVALIDDATA;
  1662. goto fail;
  1663. }
  1664. }
  1665. // set maximum Rice parameter for progressive decoding based on resolution
  1666. // This is not specified in 14496-3 but actually done by the reference
  1667. // codec RM22 revision 2.
  1668. ctx->s_max = sconf->resolution > 1 ? 31 : 15;
  1669. // set lag value for long-term prediction
  1670. ctx->ltp_lag_length = 8 + (avctx->sample_rate >= 96000) +
  1671. (avctx->sample_rate >= 192000);
  1672. // allocate quantized parcor coefficient buffer
  1673. num_buffers = sconf->mc_coding ? avctx->channels : 1;
  1674. if (num_buffers * (uint64_t)num_buffers > INT_MAX) // protect chan_data_buffer allocation
  1675. return AVERROR_INVALIDDATA;
  1676. ctx->quant_cof = av_malloc_array(num_buffers, sizeof(*ctx->quant_cof));
  1677. ctx->lpc_cof = av_malloc_array(num_buffers, sizeof(*ctx->lpc_cof));
  1678. ctx->quant_cof_buffer = av_malloc_array(num_buffers * sconf->max_order,
  1679. sizeof(*ctx->quant_cof_buffer));
  1680. ctx->lpc_cof_buffer = av_malloc_array(num_buffers * sconf->max_order,
  1681. sizeof(*ctx->lpc_cof_buffer));
  1682. ctx->lpc_cof_reversed_buffer = av_malloc_array(sconf->max_order,
  1683. sizeof(*ctx->lpc_cof_buffer));
  1684. if (!ctx->quant_cof || !ctx->lpc_cof ||
  1685. !ctx->quant_cof_buffer || !ctx->lpc_cof_buffer ||
  1686. !ctx->lpc_cof_reversed_buffer) {
  1687. av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
  1688. ret = AVERROR(ENOMEM);
  1689. goto fail;
  1690. }
  1691. // assign quantized parcor coefficient buffers
  1692. for (c = 0; c < num_buffers; c++) {
  1693. ctx->quant_cof[c] = ctx->quant_cof_buffer + c * sconf->max_order;
  1694. ctx->lpc_cof[c] = ctx->lpc_cof_buffer + c * sconf->max_order;
  1695. }
  1696. // allocate and assign lag and gain data buffer for ltp mode
  1697. ctx->const_block = av_malloc_array(num_buffers, sizeof(*ctx->const_block));
  1698. ctx->shift_lsbs = av_malloc_array(num_buffers, sizeof(*ctx->shift_lsbs));
  1699. ctx->opt_order = av_malloc_array(num_buffers, sizeof(*ctx->opt_order));
  1700. ctx->store_prev_samples = av_malloc_array(num_buffers, sizeof(*ctx->store_prev_samples));
  1701. ctx->use_ltp = av_mallocz_array(num_buffers, sizeof(*ctx->use_ltp));
  1702. ctx->ltp_lag = av_malloc_array(num_buffers, sizeof(*ctx->ltp_lag));
  1703. ctx->ltp_gain = av_malloc_array(num_buffers, sizeof(*ctx->ltp_gain));
  1704. ctx->ltp_gain_buffer = av_malloc_array(num_buffers * 5, sizeof(*ctx->ltp_gain_buffer));
  1705. if (!ctx->const_block || !ctx->shift_lsbs ||
  1706. !ctx->opt_order || !ctx->store_prev_samples ||
  1707. !ctx->use_ltp || !ctx->ltp_lag ||
  1708. !ctx->ltp_gain || !ctx->ltp_gain_buffer) {
  1709. av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
  1710. ret = AVERROR(ENOMEM);
  1711. goto fail;
  1712. }
  1713. for (c = 0; c < num_buffers; c++)
  1714. ctx->ltp_gain[c] = ctx->ltp_gain_buffer + c * 5;
  1715. // allocate and assign channel data buffer for mcc mode
  1716. if (sconf->mc_coding) {
  1717. ctx->chan_data_buffer = av_mallocz_array(num_buffers * num_buffers,
  1718. sizeof(*ctx->chan_data_buffer));
  1719. ctx->chan_data = av_mallocz_array(num_buffers,
  1720. sizeof(*ctx->chan_data));
  1721. ctx->reverted_channels = av_malloc_array(num_buffers,
  1722. sizeof(*ctx->reverted_channels));
  1723. if (!ctx->chan_data_buffer || !ctx->chan_data || !ctx->reverted_channels) {
  1724. av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
  1725. ret = AVERROR(ENOMEM);
  1726. goto fail;
  1727. }
  1728. for (c = 0; c < num_buffers; c++)
  1729. ctx->chan_data[c] = ctx->chan_data_buffer + c * num_buffers;
  1730. } else {
  1731. ctx->chan_data = NULL;
  1732. ctx->chan_data_buffer = NULL;
  1733. ctx->reverted_channels = NULL;
  1734. }
  1735. channel_size = sconf->frame_length + sconf->max_order;
  1736. ctx->prev_raw_samples = av_malloc_array(sconf->max_order, sizeof(*ctx->prev_raw_samples));
  1737. ctx->raw_buffer = av_mallocz_array(avctx->channels * channel_size, sizeof(*ctx->raw_buffer));
  1738. ctx->raw_samples = av_malloc_array(avctx->channels, sizeof(*ctx->raw_samples));
  1739. if (sconf->floating) {
  1740. ctx->acf = av_malloc_array(avctx->channels, sizeof(*ctx->acf));
  1741. ctx->shift_value = av_malloc_array(avctx->channels, sizeof(*ctx->shift_value));
  1742. ctx->last_shift_value = av_malloc_array(avctx->channels, sizeof(*ctx->last_shift_value));
  1743. ctx->last_acf_mantissa = av_malloc_array(avctx->channels, sizeof(*ctx->last_acf_mantissa));
  1744. ctx->raw_mantissa = av_mallocz_array(avctx->channels, sizeof(*ctx->raw_mantissa));
  1745. ctx->larray = av_malloc_array(ctx->cur_frame_length * 4, sizeof(*ctx->larray));
  1746. ctx->nbits = av_malloc_array(ctx->cur_frame_length, sizeof(*ctx->nbits));
  1747. ctx->mlz = av_mallocz(sizeof(*ctx->mlz));
  1748. if (!ctx->mlz || !ctx->acf || !ctx->shift_value || !ctx->last_shift_value
  1749. || !ctx->last_acf_mantissa || !ctx->raw_mantissa) {
  1750. av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
  1751. ret = AVERROR(ENOMEM);
  1752. goto fail;
  1753. }
  1754. ff_mlz_init_dict(avctx, ctx->mlz);
  1755. ff_mlz_flush_dict(ctx->mlz);
  1756. for (c = 0; c < avctx->channels; ++c) {
  1757. ctx->raw_mantissa[c] = av_mallocz_array(ctx->cur_frame_length, sizeof(**ctx->raw_mantissa));
  1758. }
  1759. }
  1760. // allocate previous raw sample buffer
  1761. if (!ctx->prev_raw_samples || !ctx->raw_buffer|| !ctx->raw_samples) {
  1762. av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
  1763. ret = AVERROR(ENOMEM);
  1764. goto fail;
  1765. }
  1766. // assign raw samples buffers
  1767. ctx->raw_samples[0] = ctx->raw_buffer + sconf->max_order;
  1768. for (c = 1; c < avctx->channels; c++)
  1769. ctx->raw_samples[c] = ctx->raw_samples[c - 1] + channel_size;
  1770. // allocate crc buffer
  1771. if (HAVE_BIGENDIAN != sconf->msb_first && sconf->crc_enabled &&
  1772. (avctx->err_recognition & (AV_EF_CRCCHECK|AV_EF_CAREFUL))) {
  1773. ctx->crc_buffer = av_malloc_array(ctx->cur_frame_length *
  1774. avctx->channels *
  1775. av_get_bytes_per_sample(avctx->sample_fmt),
  1776. sizeof(*ctx->crc_buffer));
  1777. if (!ctx->crc_buffer) {
  1778. av_log(avctx, AV_LOG_ERROR, "Allocating buffer memory failed.\n");
  1779. ret = AVERROR(ENOMEM);
  1780. goto fail;
  1781. }
  1782. }
  1783. ff_bswapdsp_init(&ctx->bdsp);
  1784. return 0;
  1785. fail:
  1786. return ret;
  1787. }
  1788. /** Flush (reset) the frame ID after seeking.
  1789. */
  1790. static av_cold void flush(AVCodecContext *avctx)
  1791. {
  1792. ALSDecContext *ctx = avctx->priv_data;
  1793. ctx->frame_id = 0;
  1794. }
  1795. AVCodec ff_als_decoder = {
  1796. .name = "als",
  1797. .long_name = NULL_IF_CONFIG_SMALL("MPEG-4 Audio Lossless Coding (ALS)"),
  1798. .type = AVMEDIA_TYPE_AUDIO,
  1799. .id = AV_CODEC_ID_MP4ALS,
  1800. .priv_data_size = sizeof(ALSDecContext),
  1801. .init = decode_init,
  1802. .close = decode_end,
  1803. .decode = decode_frame,
  1804. .flush = flush,
  1805. .capabilities = AV_CODEC_CAP_SUBFRAMES | AV_CODEC_CAP_DR1,
  1806. .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,
  1807. };