| 1 | /* |
| 2 | * ALAC (Apple Lossless Audio Codec) decoder |
| 3 | * Copyright (c) 2005 David Hammerton |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | /** |
| 23 | * @file |
| 24 | * ALAC (Apple Lossless Audio Codec) decoder |
| 25 | * @author 2005 David Hammerton |
| 26 | * @see http://crazney.net/programs/itunes/alac.html |
| 27 | * |
| 28 | * Note: This decoder expects a 36-byte QuickTime atom to be |
| 29 | * passed through the extradata[_size] fields. This atom is tacked onto |
| 30 | * the end of an 'alac' stsd atom and has the following format: |
| 31 | * |
| 32 | * 32bit atom size |
| 33 | * 32bit tag ("alac") |
| 34 | * 32bit tag version (0) |
| 35 | * 32bit samples per frame (used when not set explicitly in the frames) |
| 36 | * 8bit compatible version (0) |
| 37 | * 8bit sample size |
| 38 | * 8bit history mult (40) |
| 39 | * 8bit initial history (10) |
| 40 | * 8bit rice param limit (14) |
| 41 | * 8bit channels |
| 42 | * 16bit maxRun (255) |
| 43 | * 32bit max coded frame size (0 means unknown) |
| 44 | * 32bit average bitrate (0 means unknown) |
| 45 | * 32bit samplerate |
| 46 | */ |
| 47 | |
| 48 | #include <inttypes.h> |
| 49 | |
| 50 | #include "libavutil/channel_layout.h" |
| 51 | #include "libavutil/opt.h" |
| 52 | #include "avcodec.h" |
| 53 | #include "get_bits.h" |
| 54 | #include "bytestream.h" |
| 55 | #include "internal.h" |
| 56 | #include "thread.h" |
| 57 | #include "unary.h" |
| 58 | #include "mathops.h" |
| 59 | #include "alac_data.h" |
| 60 | |
| 61 | #define ALAC_EXTRADATA_SIZE 36 |
| 62 | |
| 63 | typedef struct { |
| 64 | AVClass *class; |
| 65 | AVCodecContext *avctx; |
| 66 | GetBitContext gb; |
| 67 | int channels; |
| 68 | |
| 69 | int32_t *predict_error_buffer[2]; |
| 70 | int32_t *output_samples_buffer[2]; |
| 71 | int32_t *extra_bits_buffer[2]; |
| 72 | |
| 73 | uint32_t max_samples_per_frame; |
| 74 | uint8_t sample_size; |
| 75 | uint8_t rice_history_mult; |
| 76 | uint8_t rice_initial_history; |
| 77 | uint8_t rice_limit; |
| 78 | |
| 79 | int extra_bits; /**< number of extra bits beyond 16-bit */ |
| 80 | int nb_samples; /**< number of samples in the current frame */ |
| 81 | |
| 82 | int direct_output; |
| 83 | int extra_bit_bug; |
| 84 | } ALACContext; |
| 85 | |
| 86 | static inline unsigned int decode_scalar(GetBitContext *gb, int k, int bps) |
| 87 | { |
| 88 | unsigned int x = get_unary_0_9(gb); |
| 89 | |
| 90 | if (x > 8) { /* RICE THRESHOLD */ |
| 91 | /* use alternative encoding */ |
| 92 | x = get_bits_long(gb, bps); |
| 93 | } else if (k != 1) { |
| 94 | int extrabits = show_bits(gb, k); |
| 95 | |
| 96 | /* multiply x by 2^k - 1, as part of their strange algorithm */ |
| 97 | x = (x << k) - x; |
| 98 | |
| 99 | if (extrabits > 1) { |
| 100 | x += extrabits - 1; |
| 101 | skip_bits(gb, k); |
| 102 | } else |
| 103 | skip_bits(gb, k - 1); |
| 104 | } |
| 105 | return x; |
| 106 | } |
| 107 | |
| 108 | static int rice_decompress(ALACContext *alac, int32_t *output_buffer, |
| 109 | int nb_samples, int bps, int rice_history_mult) |
| 110 | { |
| 111 | int i; |
| 112 | unsigned int history = alac->rice_initial_history; |
| 113 | int sign_modifier = 0; |
| 114 | |
| 115 | for (i = 0; i < nb_samples; i++) { |
| 116 | int k; |
| 117 | unsigned int x; |
| 118 | |
| 119 | if(get_bits_left(&alac->gb) <= 0) |
| 120 | return -1; |
| 121 | |
| 122 | /* calculate rice param and decode next value */ |
| 123 | k = av_log2((history >> 9) + 3); |
| 124 | k = FFMIN(k, alac->rice_limit); |
| 125 | x = decode_scalar(&alac->gb, k, bps); |
| 126 | x += sign_modifier; |
| 127 | sign_modifier = 0; |
| 128 | output_buffer[i] = (x >> 1) ^ -(x & 1); |
| 129 | |
| 130 | /* update the history */ |
| 131 | if (x > 0xffff) |
| 132 | history = 0xffff; |
| 133 | else |
| 134 | history += x * rice_history_mult - |
| 135 | ((history * rice_history_mult) >> 9); |
| 136 | |
| 137 | /* special case: there may be compressed blocks of 0 */ |
| 138 | if ((history < 128) && (i + 1 < nb_samples)) { |
| 139 | int block_size; |
| 140 | |
| 141 | /* calculate rice param and decode block size */ |
| 142 | k = 7 - av_log2(history) + ((history + 16) >> 6); |
| 143 | k = FFMIN(k, alac->rice_limit); |
| 144 | block_size = decode_scalar(&alac->gb, k, 16); |
| 145 | |
| 146 | if (block_size > 0) { |
| 147 | if (block_size >= nb_samples - i) { |
| 148 | av_log(alac->avctx, AV_LOG_ERROR, |
| 149 | "invalid zero block size of %d %d %d\n", block_size, |
| 150 | nb_samples, i); |
| 151 | block_size = nb_samples - i - 1; |
| 152 | } |
| 153 | memset(&output_buffer[i + 1], 0, |
| 154 | block_size * sizeof(*output_buffer)); |
| 155 | i += block_size; |
| 156 | } |
| 157 | if (block_size <= 0xffff) |
| 158 | sign_modifier = 1; |
| 159 | history = 0; |
| 160 | } |
| 161 | } |
| 162 | return 0; |
| 163 | } |
| 164 | |
| 165 | static inline int sign_only(int v) |
| 166 | { |
| 167 | return v ? FFSIGN(v) : 0; |
| 168 | } |
| 169 | |
| 170 | static void lpc_prediction(int32_t *error_buffer, int32_t *buffer_out, |
| 171 | int nb_samples, int bps, int16_t *lpc_coefs, |
| 172 | int lpc_order, int lpc_quant) |
| 173 | { |
| 174 | int i; |
| 175 | int32_t *pred = buffer_out; |
| 176 | |
| 177 | /* first sample always copies */ |
| 178 | *buffer_out = *error_buffer; |
| 179 | |
| 180 | if (nb_samples <= 1) |
| 181 | return; |
| 182 | |
| 183 | if (!lpc_order) { |
| 184 | memcpy(&buffer_out[1], &error_buffer[1], |
| 185 | (nb_samples - 1) * sizeof(*buffer_out)); |
| 186 | return; |
| 187 | } |
| 188 | |
| 189 | if (lpc_order == 31) { |
| 190 | /* simple 1st-order prediction */ |
| 191 | for (i = 1; i < nb_samples; i++) { |
| 192 | buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i], |
| 193 | bps); |
| 194 | } |
| 195 | return; |
| 196 | } |
| 197 | |
| 198 | /* read warm-up samples */ |
| 199 | for (i = 1; i <= lpc_order && i < nb_samples; i++) |
| 200 | buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i], bps); |
| 201 | |
| 202 | /* NOTE: 4 and 8 are very common cases that could be optimized. */ |
| 203 | |
| 204 | for (; i < nb_samples; i++) { |
| 205 | int j; |
| 206 | int val = 0; |
| 207 | int error_val = error_buffer[i]; |
| 208 | int error_sign; |
| 209 | int d = *pred++; |
| 210 | |
| 211 | /* LPC prediction */ |
| 212 | for (j = 0; j < lpc_order; j++) |
| 213 | val += (pred[j] - d) * lpc_coefs[j]; |
| 214 | val = (val + (1 << (lpc_quant - 1))) >> lpc_quant; |
| 215 | val += d + error_val; |
| 216 | buffer_out[i] = sign_extend(val, bps); |
| 217 | |
| 218 | /* adapt LPC coefficients */ |
| 219 | error_sign = sign_only(error_val); |
| 220 | if (error_sign) { |
| 221 | for (j = 0; j < lpc_order && error_val * error_sign > 0; j++) { |
| 222 | int sign; |
| 223 | val = d - pred[j]; |
| 224 | sign = sign_only(val) * error_sign; |
| 225 | lpc_coefs[j] -= sign; |
| 226 | val *= sign; |
| 227 | error_val -= (val >> lpc_quant) * (j + 1); |
| 228 | } |
| 229 | } |
| 230 | } |
| 231 | } |
| 232 | |
| 233 | static void decorrelate_stereo(int32_t *buffer[2], int nb_samples, |
| 234 | int decorr_shift, int decorr_left_weight) |
| 235 | { |
| 236 | int i; |
| 237 | |
| 238 | for (i = 0; i < nb_samples; i++) { |
| 239 | int32_t a, b; |
| 240 | |
| 241 | a = buffer[0][i]; |
| 242 | b = buffer[1][i]; |
| 243 | |
| 244 | a -= (b * decorr_left_weight) >> decorr_shift; |
| 245 | b += a; |
| 246 | |
| 247 | buffer[0][i] = b; |
| 248 | buffer[1][i] = a; |
| 249 | } |
| 250 | } |
| 251 | |
| 252 | static void append_extra_bits(int32_t *buffer[2], int32_t *extra_bits_buffer[2], |
| 253 | int extra_bits, int channels, int nb_samples) |
| 254 | { |
| 255 | int i, ch; |
| 256 | |
| 257 | for (ch = 0; ch < channels; ch++) |
| 258 | for (i = 0; i < nb_samples; i++) |
| 259 | buffer[ch][i] = (buffer[ch][i] << extra_bits) | extra_bits_buffer[ch][i]; |
| 260 | } |
| 261 | |
| 262 | static int decode_element(AVCodecContext *avctx, AVFrame *frame, int ch_index, |
| 263 | int channels) |
| 264 | { |
| 265 | ALACContext *alac = avctx->priv_data; |
| 266 | int has_size, bps, is_compressed, decorr_shift, decorr_left_weight, ret; |
| 267 | uint32_t output_samples; |
| 268 | int i, ch; |
| 269 | |
| 270 | skip_bits(&alac->gb, 4); /* element instance tag */ |
| 271 | skip_bits(&alac->gb, 12); /* unused header bits */ |
| 272 | |
| 273 | /* the number of output samples is stored in the frame */ |
| 274 | has_size = get_bits1(&alac->gb); |
| 275 | |
| 276 | alac->extra_bits = get_bits(&alac->gb, 2) << 3; |
| 277 | bps = alac->sample_size - alac->extra_bits + channels - 1; |
| 278 | if (bps > 32U) { |
| 279 | av_log(avctx, AV_LOG_ERROR, "bps is unsupported: %d\n", bps); |
| 280 | return AVERROR_PATCHWELCOME; |
| 281 | } |
| 282 | |
| 283 | /* whether the frame is compressed */ |
| 284 | is_compressed = !get_bits1(&alac->gb); |
| 285 | |
| 286 | if (has_size) |
| 287 | output_samples = get_bits_long(&alac->gb, 32); |
| 288 | else |
| 289 | output_samples = alac->max_samples_per_frame; |
| 290 | if (!output_samples || output_samples > alac->max_samples_per_frame) { |
| 291 | av_log(avctx, AV_LOG_ERROR, "invalid samples per frame: %"PRIu32"\n", |
| 292 | output_samples); |
| 293 | return AVERROR_INVALIDDATA; |
| 294 | } |
| 295 | if (!alac->nb_samples) { |
| 296 | ThreadFrame tframe = { .f = frame }; |
| 297 | /* get output buffer */ |
| 298 | frame->nb_samples = output_samples; |
| 299 | if ((ret = ff_thread_get_buffer(avctx, &tframe, 0)) < 0) |
| 300 | return ret; |
| 301 | } else if (output_samples != alac->nb_samples) { |
| 302 | av_log(avctx, AV_LOG_ERROR, "sample count mismatch: %"PRIu32" != %d\n", |
| 303 | output_samples, alac->nb_samples); |
| 304 | return AVERROR_INVALIDDATA; |
| 305 | } |
| 306 | alac->nb_samples = output_samples; |
| 307 | if (alac->direct_output) { |
| 308 | for (ch = 0; ch < channels; ch++) |
| 309 | alac->output_samples_buffer[ch] = (int32_t *)frame->extended_data[ch_index + ch]; |
| 310 | } |
| 311 | |
| 312 | if (is_compressed) { |
| 313 | int16_t lpc_coefs[2][32]; |
| 314 | int lpc_order[2]; |
| 315 | int prediction_type[2]; |
| 316 | int lpc_quant[2]; |
| 317 | int rice_history_mult[2]; |
| 318 | |
| 319 | decorr_shift = get_bits(&alac->gb, 8); |
| 320 | decorr_left_weight = get_bits(&alac->gb, 8); |
| 321 | |
| 322 | for (ch = 0; ch < channels; ch++) { |
| 323 | prediction_type[ch] = get_bits(&alac->gb, 4); |
| 324 | lpc_quant[ch] = get_bits(&alac->gb, 4); |
| 325 | rice_history_mult[ch] = get_bits(&alac->gb, 3); |
| 326 | lpc_order[ch] = get_bits(&alac->gb, 5); |
| 327 | |
| 328 | if (lpc_order[ch] >= alac->max_samples_per_frame) |
| 329 | return AVERROR_INVALIDDATA; |
| 330 | |
| 331 | /* read the predictor table */ |
| 332 | for (i = lpc_order[ch] - 1; i >= 0; i--) |
| 333 | lpc_coefs[ch][i] = get_sbits(&alac->gb, 16); |
| 334 | } |
| 335 | |
| 336 | if (alac->extra_bits) { |
| 337 | for (i = 0; i < alac->nb_samples; i++) { |
| 338 | if(get_bits_left(&alac->gb) <= 0) |
| 339 | return -1; |
| 340 | for (ch = 0; ch < channels; ch++) |
| 341 | alac->extra_bits_buffer[ch][i] = get_bits(&alac->gb, alac->extra_bits); |
| 342 | } |
| 343 | } |
| 344 | for (ch = 0; ch < channels; ch++) { |
| 345 | int ret=rice_decompress(alac, alac->predict_error_buffer[ch], |
| 346 | alac->nb_samples, bps, |
| 347 | rice_history_mult[ch] * alac->rice_history_mult / 4); |
| 348 | if(ret<0) |
| 349 | return ret; |
| 350 | |
| 351 | /* adaptive FIR filter */ |
| 352 | if (prediction_type[ch] == 15) { |
| 353 | /* Prediction type 15 runs the adaptive FIR twice. |
| 354 | * The first pass uses the special-case coef_num = 31, while |
| 355 | * the second pass uses the coefs from the bitstream. |
| 356 | * |
| 357 | * However, this prediction type is not currently used by the |
| 358 | * reference encoder. |
| 359 | */ |
| 360 | lpc_prediction(alac->predict_error_buffer[ch], |
| 361 | alac->predict_error_buffer[ch], |
| 362 | alac->nb_samples, bps, NULL, 31, 0); |
| 363 | } else if (prediction_type[ch] > 0) { |
| 364 | av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n", |
| 365 | prediction_type[ch]); |
| 366 | } |
| 367 | lpc_prediction(alac->predict_error_buffer[ch], |
| 368 | alac->output_samples_buffer[ch], alac->nb_samples, |
| 369 | bps, lpc_coefs[ch], lpc_order[ch], lpc_quant[ch]); |
| 370 | } |
| 371 | } else { |
| 372 | /* not compressed, easy case */ |
| 373 | for (i = 0; i < alac->nb_samples; i++) { |
| 374 | if(get_bits_left(&alac->gb) <= 0) |
| 375 | return -1; |
| 376 | for (ch = 0; ch < channels; ch++) { |
| 377 | alac->output_samples_buffer[ch][i] = |
| 378 | get_sbits_long(&alac->gb, alac->sample_size); |
| 379 | } |
| 380 | } |
| 381 | alac->extra_bits = 0; |
| 382 | decorr_shift = 0; |
| 383 | decorr_left_weight = 0; |
| 384 | } |
| 385 | |
| 386 | if (alac->extra_bits && alac->extra_bit_bug) { |
| 387 | append_extra_bits(alac->output_samples_buffer, alac->extra_bits_buffer, |
| 388 | alac->extra_bits, channels, alac->nb_samples); |
| 389 | } |
| 390 | |
| 391 | if (channels == 2 && decorr_left_weight) { |
| 392 | decorrelate_stereo(alac->output_samples_buffer, alac->nb_samples, |
| 393 | decorr_shift, decorr_left_weight); |
| 394 | } |
| 395 | |
| 396 | if (alac->extra_bits && !alac->extra_bit_bug) { |
| 397 | append_extra_bits(alac->output_samples_buffer, alac->extra_bits_buffer, |
| 398 | alac->extra_bits, channels, alac->nb_samples); |
| 399 | } |
| 400 | |
| 401 | if(av_sample_fmt_is_planar(avctx->sample_fmt)) { |
| 402 | switch(alac->sample_size) { |
| 403 | case 16: { |
| 404 | for (ch = 0; ch < channels; ch++) { |
| 405 | int16_t *outbuffer = (int16_t *)frame->extended_data[ch_index + ch]; |
| 406 | for (i = 0; i < alac->nb_samples; i++) |
| 407 | *outbuffer++ = alac->output_samples_buffer[ch][i]; |
| 408 | }} |
| 409 | break; |
| 410 | case 24: { |
| 411 | for (ch = 0; ch < channels; ch++) { |
| 412 | for (i = 0; i < alac->nb_samples; i++) |
| 413 | alac->output_samples_buffer[ch][i] <<= 8; |
| 414 | }} |
| 415 | break; |
| 416 | } |
| 417 | }else{ |
| 418 | switch(alac->sample_size) { |
| 419 | case 16: { |
| 420 | int16_t *outbuffer = ((int16_t *)frame->extended_data[0]) + ch_index; |
| 421 | for (i = 0; i < alac->nb_samples; i++) { |
| 422 | for (ch = 0; ch < channels; ch++) |
| 423 | *outbuffer++ = alac->output_samples_buffer[ch][i]; |
| 424 | outbuffer += alac->channels - channels; |
| 425 | } |
| 426 | } |
| 427 | break; |
| 428 | case 24: { |
| 429 | int32_t *outbuffer = ((int32_t *)frame->extended_data[0]) + ch_index; |
| 430 | for (i = 0; i < alac->nb_samples; i++) { |
| 431 | for (ch = 0; ch < channels; ch++) |
| 432 | *outbuffer++ = alac->output_samples_buffer[ch][i] << 8; |
| 433 | outbuffer += alac->channels - channels; |
| 434 | } |
| 435 | } |
| 436 | break; |
| 437 | case 32: { |
| 438 | int32_t *outbuffer = ((int32_t *)frame->extended_data[0]) + ch_index; |
| 439 | for (i = 0; i < alac->nb_samples; i++) { |
| 440 | for (ch = 0; ch < channels; ch++) |
| 441 | *outbuffer++ = alac->output_samples_buffer[ch][i]; |
| 442 | outbuffer += alac->channels - channels; |
| 443 | } |
| 444 | } |
| 445 | break; |
| 446 | } |
| 447 | } |
| 448 | |
| 449 | return 0; |
| 450 | } |
| 451 | |
| 452 | static int alac_decode_frame(AVCodecContext *avctx, void *data, |
| 453 | int *got_frame_ptr, AVPacket *avpkt) |
| 454 | { |
| 455 | ALACContext *alac = avctx->priv_data; |
| 456 | AVFrame *frame = data; |
| 457 | enum AlacRawDataBlockType element; |
| 458 | int channels; |
| 459 | int ch, ret, got_end; |
| 460 | |
| 461 | if ((ret = init_get_bits8(&alac->gb, avpkt->data, avpkt->size)) < 0) |
| 462 | return ret; |
| 463 | |
| 464 | got_end = 0; |
| 465 | alac->nb_samples = 0; |
| 466 | ch = 0; |
| 467 | while (get_bits_left(&alac->gb) >= 3) { |
| 468 | element = get_bits(&alac->gb, 3); |
| 469 | if (element == TYPE_END) { |
| 470 | got_end = 1; |
| 471 | break; |
| 472 | } |
| 473 | if (element > TYPE_CPE && element != TYPE_LFE) { |
| 474 | av_log(avctx, AV_LOG_ERROR, "syntax element unsupported: %d\n", element); |
| 475 | return AVERROR_PATCHWELCOME; |
| 476 | } |
| 477 | |
| 478 | channels = (element == TYPE_CPE) ? 2 : 1; |
| 479 | if (ch + channels > alac->channels || |
| 480 | ff_alac_channel_layout_offsets[alac->channels - 1][ch] + channels > alac->channels) { |
| 481 | av_log(avctx, AV_LOG_ERROR, "invalid element channel count\n"); |
| 482 | return AVERROR_INVALIDDATA; |
| 483 | } |
| 484 | |
| 485 | ret = decode_element(avctx, frame, |
| 486 | ff_alac_channel_layout_offsets[alac->channels - 1][ch], |
| 487 | channels); |
| 488 | if (ret < 0 && get_bits_left(&alac->gb)) |
| 489 | return ret; |
| 490 | |
| 491 | ch += channels; |
| 492 | } |
| 493 | if (!got_end) { |
| 494 | av_log(avctx, AV_LOG_ERROR, "no end tag found. incomplete packet.\n"); |
| 495 | return AVERROR_INVALIDDATA; |
| 496 | } |
| 497 | |
| 498 | if (avpkt->size * 8 - get_bits_count(&alac->gb) > 8) { |
| 499 | av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n", |
| 500 | avpkt->size * 8 - get_bits_count(&alac->gb)); |
| 501 | } |
| 502 | |
| 503 | if (alac->channels == ch) |
| 504 | *got_frame_ptr = 1; |
| 505 | else |
| 506 | av_log(avctx, AV_LOG_WARNING, "Failed to decode all channels\n"); |
| 507 | |
| 508 | return avpkt->size; |
| 509 | } |
| 510 | |
| 511 | static av_cold int alac_decode_close(AVCodecContext *avctx) |
| 512 | { |
| 513 | ALACContext *alac = avctx->priv_data; |
| 514 | |
| 515 | int ch; |
| 516 | for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) { |
| 517 | av_freep(&alac->predict_error_buffer[ch]); |
| 518 | if (!alac->direct_output) |
| 519 | av_freep(&alac->output_samples_buffer[ch]); |
| 520 | av_freep(&alac->extra_bits_buffer[ch]); |
| 521 | } |
| 522 | |
| 523 | return 0; |
| 524 | } |
| 525 | |
| 526 | static int allocate_buffers(ALACContext *alac) |
| 527 | { |
| 528 | int ch; |
| 529 | int buf_size = alac->max_samples_per_frame * sizeof(int32_t); |
| 530 | |
| 531 | for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) { |
| 532 | FF_ALLOC_OR_GOTO(alac->avctx, alac->predict_error_buffer[ch], |
| 533 | buf_size, buf_alloc_fail); |
| 534 | |
| 535 | alac->direct_output = alac->sample_size > 16 && av_sample_fmt_is_planar(alac->avctx->sample_fmt); |
| 536 | if (!alac->direct_output) { |
| 537 | FF_ALLOC_OR_GOTO(alac->avctx, alac->output_samples_buffer[ch], |
| 538 | buf_size, buf_alloc_fail); |
| 539 | } |
| 540 | |
| 541 | FF_ALLOC_OR_GOTO(alac->avctx, alac->extra_bits_buffer[ch], |
| 542 | buf_size, buf_alloc_fail); |
| 543 | } |
| 544 | return 0; |
| 545 | buf_alloc_fail: |
| 546 | alac_decode_close(alac->avctx); |
| 547 | return AVERROR(ENOMEM); |
| 548 | } |
| 549 | |
| 550 | static int alac_set_info(ALACContext *alac) |
| 551 | { |
| 552 | GetByteContext gb; |
| 553 | |
| 554 | bytestream2_init(&gb, alac->avctx->extradata, |
| 555 | alac->avctx->extradata_size); |
| 556 | |
| 557 | bytestream2_skipu(&gb, 12); // size:4, alac:4, version:4 |
| 558 | |
| 559 | alac->max_samples_per_frame = bytestream2_get_be32u(&gb); |
| 560 | if (!alac->max_samples_per_frame || |
| 561 | alac->max_samples_per_frame > INT_MAX / sizeof(int32_t)) { |
| 562 | av_log(alac->avctx, AV_LOG_ERROR, |
| 563 | "max samples per frame invalid: %"PRIu32"\n", |
| 564 | alac->max_samples_per_frame); |
| 565 | return AVERROR_INVALIDDATA; |
| 566 | } |
| 567 | bytestream2_skipu(&gb, 1); // compatible version |
| 568 | alac->sample_size = bytestream2_get_byteu(&gb); |
| 569 | alac->rice_history_mult = bytestream2_get_byteu(&gb); |
| 570 | alac->rice_initial_history = bytestream2_get_byteu(&gb); |
| 571 | alac->rice_limit = bytestream2_get_byteu(&gb); |
| 572 | alac->channels = bytestream2_get_byteu(&gb); |
| 573 | bytestream2_get_be16u(&gb); // maxRun |
| 574 | bytestream2_get_be32u(&gb); // max coded frame size |
| 575 | bytestream2_get_be32u(&gb); // average bitrate |
| 576 | bytestream2_get_be32u(&gb); // samplerate |
| 577 | |
| 578 | return 0; |
| 579 | } |
| 580 | |
| 581 | static av_cold int alac_decode_init(AVCodecContext * avctx) |
| 582 | { |
| 583 | int ret; |
| 584 | int req_packed; |
| 585 | ALACContext *alac = avctx->priv_data; |
| 586 | alac->avctx = avctx; |
| 587 | |
| 588 | /* initialize from the extradata */ |
| 589 | if (alac->avctx->extradata_size < ALAC_EXTRADATA_SIZE) { |
| 590 | av_log(avctx, AV_LOG_ERROR, "extradata is too small\n"); |
| 591 | return AVERROR_INVALIDDATA; |
| 592 | } |
| 593 | if (alac_set_info(alac)) { |
| 594 | av_log(avctx, AV_LOG_ERROR, "set_info failed\n"); |
| 595 | return -1; |
| 596 | } |
| 597 | |
| 598 | req_packed = LIBAVCODEC_VERSION_MAJOR < 55 && !av_sample_fmt_is_planar(avctx->request_sample_fmt); |
| 599 | switch (alac->sample_size) { |
| 600 | case 16: avctx->sample_fmt = req_packed ? AV_SAMPLE_FMT_S16 : AV_SAMPLE_FMT_S16P; |
| 601 | break; |
| 602 | case 24: |
| 603 | case 32: avctx->sample_fmt = req_packed ? AV_SAMPLE_FMT_S32 : AV_SAMPLE_FMT_S32P; |
| 604 | break; |
| 605 | default: avpriv_request_sample(avctx, "Sample depth %d", alac->sample_size); |
| 606 | return AVERROR_PATCHWELCOME; |
| 607 | } |
| 608 | avctx->bits_per_raw_sample = alac->sample_size; |
| 609 | |
| 610 | if (alac->channels < 1) { |
| 611 | av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n"); |
| 612 | alac->channels = avctx->channels; |
| 613 | } else { |
| 614 | if (alac->channels > ALAC_MAX_CHANNELS) |
| 615 | alac->channels = avctx->channels; |
| 616 | else |
| 617 | avctx->channels = alac->channels; |
| 618 | } |
| 619 | if (avctx->channels > ALAC_MAX_CHANNELS || avctx->channels <= 0 ) { |
| 620 | av_log(avctx, AV_LOG_ERROR, "Unsupported channel count: %d\n", |
| 621 | avctx->channels); |
| 622 | return AVERROR_PATCHWELCOME; |
| 623 | } |
| 624 | avctx->channel_layout = ff_alac_channel_layouts[alac->channels - 1]; |
| 625 | |
| 626 | if ((ret = allocate_buffers(alac)) < 0) { |
| 627 | av_log(avctx, AV_LOG_ERROR, "Error allocating buffers\n"); |
| 628 | return ret; |
| 629 | } |
| 630 | |
| 631 | return 0; |
| 632 | } |
| 633 | |
| 634 | static int init_thread_copy(AVCodecContext *avctx) |
| 635 | { |
| 636 | ALACContext *alac = avctx->priv_data; |
| 637 | alac->avctx = avctx; |
| 638 | return allocate_buffers(alac); |
| 639 | } |
| 640 | |
| 641 | static const AVOption options[] = { |
| 642 | { "extra_bits_bug", "Force non-standard decoding process", |
| 643 | offsetof(ALACContext, extra_bit_bug), AV_OPT_TYPE_INT, { .i64 = 0 }, |
| 644 | 0, 1, AV_OPT_FLAG_AUDIO_PARAM | AV_OPT_FLAG_DECODING_PARAM }, |
| 645 | { NULL }, |
| 646 | }; |
| 647 | |
| 648 | static const AVClass alac_class = { |
| 649 | .class_name = "alac", |
| 650 | .item_name = av_default_item_name, |
| 651 | .option = options, |
| 652 | .version = LIBAVUTIL_VERSION_INT, |
| 653 | }; |
| 654 | |
| 655 | AVCodec ff_alac_decoder = { |
| 656 | .name = "alac", |
| 657 | .long_name = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"), |
| 658 | .type = AVMEDIA_TYPE_AUDIO, |
| 659 | .id = AV_CODEC_ID_ALAC, |
| 660 | .priv_data_size = sizeof(ALACContext), |
| 661 | .init = alac_decode_init, |
| 662 | .close = alac_decode_close, |
| 663 | .decode = alac_decode_frame, |
| 664 | .init_thread_copy = ONLY_IF_THREADS_ENABLED(init_thread_copy), |
| 665 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS, |
| 666 | .priv_class = &alac_class |
| 667 | }; |