libavcodec/alac.c

   1 /*
   2  * ALAC (Apple Lossless Audio Codec) decoder
   3  * Copyright (c) 2005 David Hammerton
   4  *
   5  * This file is part of Libav.
   6  *
   7  * Libav 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  * Libav 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 Libav; 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      (14)
  40  *  8bit  rice param limit     (10)
  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 "libavutil/channel_layout.h"
  49 #include "avcodec.h"
  50 #include "get_bits.h"
  51 #include "bytestream.h"
  52 #include "internal.h"
  53 #include "unary.h"
  54 #include "mathops.h"
  55 #include "alac_data.h"
  56
  57 #define ALAC_EXTRADATA_SIZE 36
  58
  59 typedef struct {
  60     AVCodecContext *avctx;
  61     GetBitContext gb;
  62     int channels;
  63
  64     int32_t *predict_error_buffer[2];
  65     int32_t *output_samples_buffer[2];
  66     int32_t *extra_bits_buffer[2];
  67
  68     uint32_t max_samples_per_frame;
  69     uint8_t  sample_size;
  70     uint8_t  rice_history_mult;
  71     uint8_t  rice_initial_history;
  72     uint8_t  rice_limit;
  73
  74     int extra_bits;     /**< number of extra bits beyond 16-bit */
  75     int nb_samples;     /**< number of samples in the current frame */
  76 } ALACContext;
  77
  78 static inline unsigned int decode_scalar(GetBitContext *gb, int k, int bps)
  79 {
  80     unsigned int x = get_unary_0_9(gb);
  81
  82     if (x > 8) { /* RICE THRESHOLD */
  83         /* use alternative encoding */
  84         x = get_bits_long(gb, bps);
  85     } else if (k != 1) {
  86         int extrabits = show_bits(gb, k);
  87
  88         /* multiply x by 2^k - 1, as part of their strange algorithm */
  89         x = (x << k) - x;
  90
  91         if (extrabits > 1) {
  92             x += extrabits - 1;
  93             skip_bits(gb, k);
  94         } else
  95             skip_bits(gb, k - 1);
  96     }
  97     return x;
  98 }
  99
 100 static void rice_decompress(ALACContext *alac, int32_t *output_buffer,
 101                             int nb_samples, int bps, int rice_history_mult)
 102 {
 103     int i;
 104     unsigned int history = alac->rice_initial_history;
 105     int sign_modifier = 0;
 106
 107     for (i = 0; i < nb_samples; i++) {
 108         int k;
 109         unsigned int x;
 110
 111         /* calculate rice param and decode next value */
 112         k = av_log2((history >> 9) + 3);
 113         k = FFMIN(k, alac->rice_limit);
 114         x = decode_scalar(&alac->gb, k, bps);
 115         x += sign_modifier;
 116         sign_modifier = 0;
 117         output_buffer[i] = (x >> 1) ^ -(x & 1);
 118
 119         /* update the history */
 120         if (x > 0xffff)
 121             history = 0xffff;
 122         else
 123             history +=         x * rice_history_mult -
 124                        ((history * rice_history_mult) >> 9);
 125
 126         /* special case: there may be compressed blocks of 0 */
 127         if ((history < 128) && (i + 1 < nb_samples)) {
 128             int block_size;
 129
 130             /* calculate rice param and decode block size */
 131             k = 7 - av_log2(history) + ((history + 16) >> 6);
 132             k = FFMIN(k, alac->rice_limit);
 133             block_size = decode_scalar(&alac->gb, k, 16);
 134
 135             if (block_size > 0) {
 136                 if (block_size >= nb_samples - i) {
 137                     av_log(alac->avctx, AV_LOG_ERROR,
 138                            "invalid zero block size of %d %d %d\n", block_size,
 139                            nb_samples, i);
 140                     block_size = nb_samples - i - 1;
 141                 }
 142                 memset(&output_buffer[i + 1], 0,
 143                        block_size * sizeof(*output_buffer));
 144                 i += block_size;
 145             }
 146             if (block_size <= 0xffff)
 147                 sign_modifier = 1;
 148             history = 0;
 149         }
 150     }
 151 }
 152
 153 static inline int sign_only(int v)
 154 {
 155     return v ? FFSIGN(v) : 0;
 156 }
 157
 158 static void lpc_prediction(int32_t *error_buffer, int32_t *buffer_out,
 159                            int nb_samples, int bps, int16_t *lpc_coefs,
 160                            int lpc_order, int lpc_quant)
 161 {
 162     int i;
 163     int32_t *pred = buffer_out;
 164
 165     /* first sample always copies */
 166     *buffer_out = *error_buffer;
 167
 168     if (nb_samples <= 1)
 169         return;
 170
 171     if (!lpc_order) {
 172         memcpy(&buffer_out[1], &error_buffer[1],
 173                (nb_samples - 1) * sizeof(*buffer_out));
 174         return;
 175     }
 176
 177     if (lpc_order == 31) {
 178         /* simple 1st-order prediction */
 179         for (i = 1; i < nb_samples; i++) {
 180             buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i],
 181                                         bps);
 182         }
 183         return;
 184     }
 185
 186     /* read warm-up samples */
 187     for (i = 1; i <= lpc_order; i++)
 188         buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i], bps);
 189
 190     /* NOTE: 4 and 8 are very common cases that could be optimized. */
 191
 192     for (; i < nb_samples; i++) {
 193         int j;
 194         int val = 0;
 195         int error_val = error_buffer[i];
 196         int error_sign;
 197         int d = *pred++;
 198
 199         /* LPC prediction */
 200         for (j = 0; j < lpc_order; j++)
 201             val += (pred[j] - d) * lpc_coefs[j];
 202         val = (val + (1 << (lpc_quant - 1))) >> lpc_quant;
 203         val += d + error_val;
 204         buffer_out[i] = sign_extend(val, bps);
 205
 206         /* adapt LPC coefficients */
 207         error_sign = sign_only(error_val);
 208         if (error_sign) {
 209             for (j = 0; j < lpc_order && error_val * error_sign > 0; j++) {
 210                 int sign;
 211                 val  = d - pred[j];
 212                 sign = sign_only(val) * error_sign;
 213                 lpc_coefs[j] -= sign;
 214                 val *= sign;
 215                 error_val -= (val >> lpc_quant) * (j + 1);
 216             }
 217         }
 218     }
 219 }
 220
 221 static void decorrelate_stereo(int32_t *buffer[2], int nb_samples,
 222                                int decorr_shift, int decorr_left_weight)
 223 {
 224     int i;
 225
 226     for (i = 0; i < nb_samples; i++) {
 227         int32_t a, b;
 228
 229         a = buffer[0][i];
 230         b = buffer[1][i];
 231
 232         a -= (b * decorr_left_weight) >> decorr_shift;
 233         b += a;
 234
 235         buffer[0][i] = b;
 236         buffer[1][i] = a;
 237     }
 238 }
 239
 240 static void append_extra_bits(int32_t *buffer[2], int32_t *extra_bits_buffer[2],
 241                               int extra_bits, int channels, int nb_samples)
 242 {
 243     int i, ch;
 244
 245     for (ch = 0; ch < channels; ch++)
 246         for (i = 0; i < nb_samples; i++)
 247             buffer[ch][i] = (buffer[ch][i] << extra_bits) | extra_bits_buffer[ch][i];
 248 }
 249
 250 static int decode_element(AVCodecContext *avctx, AVFrame *frame, int ch_index,
 251                           int channels)
 252 {
 253     ALACContext *alac = avctx->priv_data;
 254     int has_size, bps, is_compressed, decorr_shift, decorr_left_weight, ret;
 255     uint32_t output_samples;
 256     int i, ch;
 257
 258     skip_bits(&alac->gb, 4);  /* element instance tag */
 259     skip_bits(&alac->gb, 12); /* unused header bits */
 260
 261     /* the number of output samples is stored in the frame */
 262     has_size = get_bits1(&alac->gb);
 263
 264     alac->extra_bits = get_bits(&alac->gb, 2) << 3;
 265     bps = alac->sample_size - alac->extra_bits + channels - 1;
 266     if (bps > 32) {
 267         av_log(avctx, AV_LOG_ERROR, "bps is unsupported: %d\n", bps);
 268         return AVERROR_PATCHWELCOME;
 269     }
 270
 271     /* whether the frame is compressed */
 272     is_compressed = !get_bits1(&alac->gb);
 273
 274     if (has_size)
 275         output_samples = get_bits_long(&alac->gb, 32);
 276     else
 277         output_samples = alac->max_samples_per_frame;
 278     if (!output_samples || output_samples > alac->max_samples_per_frame) {
 279         av_log(avctx, AV_LOG_ERROR, "invalid samples per frame: %d\n",
 280                output_samples);
 281         return AVERROR_INVALIDDATA;
 282     }
 283     if (!alac->nb_samples) {
 284         /* get output buffer */
 285         frame->nb_samples = output_samples;
 286         if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
 287             av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
 288             return ret;
 289         }
 290     } else if (output_samples != alac->nb_samples) {
 291         av_log(avctx, AV_LOG_ERROR, "sample count mismatch: %u != %d\n",
 292                output_samples, alac->nb_samples);
 293         return AVERROR_INVALIDDATA;
 294     }
 295     alac->nb_samples = output_samples;
 296     if (alac->sample_size > 16) {
 297         for (ch = 0; ch < channels; ch++)
 298             alac->output_samples_buffer[ch] = (int32_t *)frame->extended_data[ch_index + ch];
 299     }
 300
 301     if (is_compressed) {
 302         int16_t lpc_coefs[2][32];
 303         int lpc_order[2];
 304         int prediction_type[2];
 305         int lpc_quant[2];
 306         int rice_history_mult[2];
 307
 308         decorr_shift       = get_bits(&alac->gb, 8);
 309         decorr_left_weight = get_bits(&alac->gb, 8);
 310
 311         for (ch = 0; ch < channels; ch++) {
 312             prediction_type[ch]   = get_bits(&alac->gb, 4);
 313             lpc_quant[ch]         = get_bits(&alac->gb, 4);
 314             rice_history_mult[ch] = get_bits(&alac->gb, 3);
 315             lpc_order[ch]         = get_bits(&alac->gb, 5);
 316
 317             /* read the predictor table */
 318             for (i = lpc_order[ch] - 1; i >= 0; i--)
 319                 lpc_coefs[ch][i] = get_sbits(&alac->gb, 16);
 320         }
 321
 322         if (alac->extra_bits) {
 323             for (i = 0; i < alac->nb_samples; i++) {
 324                 for (ch = 0; ch < channels; ch++)
 325                     alac->extra_bits_buffer[ch][i] = get_bits(&alac->gb, alac->extra_bits);
 326             }
 327         }
 328         for (ch = 0; ch < channels; ch++) {
 329             rice_decompress(alac, alac->predict_error_buffer[ch],
 330                             alac->nb_samples, bps,
 331                             rice_history_mult[ch] * alac->rice_history_mult / 4);
 332
 333             /* adaptive FIR filter */
 334             if (prediction_type[ch] == 15) {
 335                 /* Prediction type 15 runs the adaptive FIR twice.
 336                  * The first pass uses the special-case coef_num = 31, while
 337                  * the second pass uses the coefs from the bitstream.
 338                  *
 339                  * However, this prediction type is not currently used by the
 340                  * reference encoder.
 341                  */
 342                 lpc_prediction(alac->predict_error_buffer[ch],
 343                                alac->predict_error_buffer[ch],
 344                                alac->nb_samples, bps, NULL, 31, 0);
 345             } else if (prediction_type[ch] > 0) {
 346                 av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n",
 347                        prediction_type[ch]);
 348             }
 349             lpc_prediction(alac->predict_error_buffer[ch],
 350                            alac->output_samples_buffer[ch], alac->nb_samples,
 351                            bps, lpc_coefs[ch], lpc_order[ch], lpc_quant[ch]);
 352         }
 353     } else {
 354         /* not compressed, easy case */
 355         for (i = 0; i < alac->nb_samples; i++) {
 356             for (ch = 0; ch < channels; ch++) {
 357                 alac->output_samples_buffer[ch][i] =
 358                          get_sbits_long(&alac->gb, alac->sample_size);
 359             }
 360         }
 361         alac->extra_bits   = 0;
 362         decorr_shift       = 0;
 363         decorr_left_weight = 0;
 364     }
 365
 366     if (channels == 2 && decorr_left_weight) {
 367         decorrelate_stereo(alac->output_samples_buffer, alac->nb_samples,
 368                            decorr_shift, decorr_left_weight);
 369     }
 370
 371     if (alac->extra_bits) {
 372         append_extra_bits(alac->output_samples_buffer, alac->extra_bits_buffer,
 373                           alac->extra_bits, channels, alac->nb_samples);
 374     }
 375
 376     switch(alac->sample_size) {
 377     case 16: {
 378         for (ch = 0; ch < channels; ch++) {
 379             int16_t *outbuffer = (int16_t *)frame->extended_data[ch_index + ch];
 380             for (i = 0; i < alac->nb_samples; i++)
 381                 *outbuffer++ = alac->output_samples_buffer[ch][i];
 382         }}
 383         break;
 384     case 24: {
 385         for (ch = 0; ch < channels; ch++) {
 386             for (i = 0; i < alac->nb_samples; i++)
 387                 alac->output_samples_buffer[ch][i] <<= 8;
 388         }}
 389         break;
 390     }
 391
 392     return 0;
 393 }
 394
 395 static int alac_decode_frame(AVCodecContext *avctx, void *data,
 396                              int *got_frame_ptr, AVPacket *avpkt)
 397 {
 398     ALACContext *alac = avctx->priv_data;
 399     AVFrame *frame    = data;
 400     enum AlacRawDataBlockType element;
 401     int channels;
 402     int ch, ret, got_end;
 403
 404     init_get_bits(&alac->gb, avpkt->data, avpkt->size * 8);
 405
 406     got_end = 0;
 407     alac->nb_samples = 0;
 408     ch = 0;
 409     while (get_bits_left(&alac->gb) >= 3) {
 410         element = get_bits(&alac->gb, 3);
 411         if (element == TYPE_END) {
 412             got_end = 1;
 413             break;
 414         }
 415         if (element > TYPE_CPE && element != TYPE_LFE) {
 416             av_log(avctx, AV_LOG_ERROR, "syntax element unsupported: %d", element);
 417             return AVERROR_PATCHWELCOME;
 418         }
 419
 420         channels = (element == TYPE_CPE) ? 2 : 1;
 421         if (ch + channels > alac->channels) {
 422             av_log(avctx, AV_LOG_ERROR, "invalid element channel count\n");
 423             return AVERROR_INVALIDDATA;
 424         }
 425
 426         ret = decode_element(avctx, frame,
 427                              ff_alac_channel_layout_offsets[alac->channels - 1][ch],
 428                              channels);
 429         if (ret < 0 && get_bits_left(&alac->gb))
 430             return ret;
 431
 432         ch += channels;
 433     }
 434     if (!got_end) {
 435         av_log(avctx, AV_LOG_ERROR, "no end tag found. incomplete packet.\n");
 436         return AVERROR_INVALIDDATA;
 437     }
 438
 439     if (avpkt->size * 8 - get_bits_count(&alac->gb) > 8) {
 440         av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n",
 441                avpkt->size * 8 - get_bits_count(&alac->gb));
 442     }
 443
 444     *got_frame_ptr = 1;
 445
 446     return avpkt->size;
 447 }
 448
 449 static av_cold int alac_decode_close(AVCodecContext *avctx)
 450 {
 451     ALACContext *alac = avctx->priv_data;
 452
 453     int ch;
 454     for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
 455         av_freep(&alac->predict_error_buffer[ch]);
 456         if (alac->sample_size == 16)
 457             av_freep(&alac->output_samples_buffer[ch]);
 458         av_freep(&alac->extra_bits_buffer[ch]);
 459     }
 460
 461     return 0;
 462 }
 463
 464 static int allocate_buffers(ALACContext *alac)
 465 {
 466     int ch;
 467     int buf_size = alac->max_samples_per_frame * sizeof(int32_t);
 468
 469     for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
 470         FF_ALLOC_OR_GOTO(alac->avctx, alac->predict_error_buffer[ch],
 471                          buf_size, buf_alloc_fail);
 472
 473         if (alac->sample_size == 16) {
 474             FF_ALLOC_OR_GOTO(alac->avctx, alac->output_samples_buffer[ch],
 475                              buf_size, buf_alloc_fail);
 476         }
 477
 478         FF_ALLOC_OR_GOTO(alac->avctx, alac->extra_bits_buffer[ch],
 479                          buf_size, buf_alloc_fail);
 480     }
 481     return 0;
 482 buf_alloc_fail:
 483     alac_decode_close(alac->avctx);
 484     return AVERROR(ENOMEM);
 485 }
 486
 487 static int alac_set_info(ALACContext *alac)
 488 {
 489     GetByteContext gb;
 490
 491     bytestream2_init(&gb, alac->avctx->extradata,
 492                      alac->avctx->extradata_size);
 493
 494     bytestream2_skipu(&gb, 12); // size:4, alac:4, version:4
 495
 496     alac->max_samples_per_frame = bytestream2_get_be32u(&gb);
 497     if (!alac->max_samples_per_frame || alac->max_samples_per_frame > INT_MAX) {
 498         av_log(alac->avctx, AV_LOG_ERROR, "max samples per frame invalid: %u\n",
 499                alac->max_samples_per_frame);
 500         return AVERROR_INVALIDDATA;
 501     }
 502     bytestream2_skipu(&gb, 1);  // compatible version
 503     alac->sample_size          = bytestream2_get_byteu(&gb);
 504     alac->rice_history_mult    = bytestream2_get_byteu(&gb);
 505     alac->rice_initial_history = bytestream2_get_byteu(&gb);
 506     alac->rice_limit           = bytestream2_get_byteu(&gb);
 507     alac->channels             = bytestream2_get_byteu(&gb);
 508     bytestream2_get_be16u(&gb); // maxRun
 509     bytestream2_get_be32u(&gb); // max coded frame size
 510     bytestream2_get_be32u(&gb); // average bitrate
 511     bytestream2_get_be32u(&gb); // samplerate
 512
 513     return 0;
 514 }
 515
 516 static av_cold int alac_decode_init(AVCodecContext * avctx)
 517 {
 518     int ret;
 519     ALACContext *alac = avctx->priv_data;
 520     alac->avctx = avctx;
 521
 522     /* initialize from the extradata */
 523     if (alac->avctx->extradata_size < ALAC_EXTRADATA_SIZE) {
 524         av_log(avctx, AV_LOG_ERROR, "alac: extradata is too small\n");
 525         return AVERROR_INVALIDDATA;
 526     }
 527     if (alac_set_info(alac)) {
 528         av_log(avctx, AV_LOG_ERROR, "alac: set_info failed\n");
 529         return -1;
 530     }
 531
 532     switch (alac->sample_size) {
 533     case 16: avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
 534              break;
 535     case 24:
 536     case 32: avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
 537              break;
 538     default: av_log_ask_for_sample(avctx, "Sample depth %d is not supported.\n",
 539                                    alac->sample_size);
 540              return AVERROR_PATCHWELCOME;
 541     }
 542     avctx->bits_per_raw_sample = alac->sample_size;
 543
 544     if (alac->channels < 1) {
 545         av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n");
 546         alac->channels = avctx->channels;
 547     } else {
 548         if (alac->channels > ALAC_MAX_CHANNELS)
 549             alac->channels = avctx->channels;
 550         else
 551             avctx->channels = alac->channels;
 552     }
 553     if (avctx->channels > ALAC_MAX_CHANNELS) {
 554         av_log(avctx, AV_LOG_ERROR, "Unsupported channel count: %d\n",
 555                avctx->channels);
 556         return AVERROR_PATCHWELCOME;
 557     }
 558     avctx->channel_layout = ff_alac_channel_layouts[alac->channels - 1];
 559
 560     if ((ret = allocate_buffers(alac)) < 0) {
 561         av_log(avctx, AV_LOG_ERROR, "Error allocating buffers\n");
 562         return ret;
 563     }
 564
 565     return 0;
 566 }
 567
 568 AVCodec ff_alac_decoder = {
 569     .name           = "alac",
 570     .type           = AVMEDIA_TYPE_AUDIO,
 571     .id             = AV_CODEC_ID_ALAC,
 572     .priv_data_size = sizeof(ALACContext),
 573     .init           = alac_decode_init,
 574     .close          = alac_decode_close,
 575     .decode         = alac_decode_frame,
 576     .capabilities   = CODEC_CAP_DR1,
 577     .long_name      = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
 578 };