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  * 32 bits  atom size
  33  * 32 bits  tag                  ("alac")
  34  * 32 bits  tag version          (0)
  35  * 32 bits  samples per frame    (used when not set explicitly in the frames)
  36  *  8 bits  compatible version   (0)
  37  *  8 bits  sample size
  38  *  8 bits  history mult         (40)
  39  *  8 bits  initial history      (14)
  40  *  8 bits  rice param limit     (10)
  41  *  8 bits  channels
  42  * 16 bits  maxRun               (255)
  43  * 32 bits  max coded frame size (0 means unknown)
  44  * 32 bits  average bitrate      (0 means unknown)
  45  * 32 bits  samplerate
  46  */
  47
  48 #include <inttypes.h>
  49
  50 #include "libavutil/channel_layout.h"
  51
  52 #include "avcodec.h"
  53 #include "bitstream.h"
  54 #include "bytestream.h"
  55 #include "internal.h"
  56 #include "mathops.h"
  57 #include "unary.h"
  58 #include "alac_data.h"
  59
  60 #define ALAC_EXTRADATA_SIZE 36
  61
  62 typedef struct ALACContext {
  63     AVCodecContext *avctx;
  64     BitstreamContext bc;
  65     int channels;
  66
  67     int32_t *predict_error_buffer[2];
  68     int32_t *output_samples_buffer[2];
  69     int32_t *extra_bits_buffer[2];
  70
  71     uint32_t max_samples_per_frame;
  72     uint8_t  sample_size;
  73     uint8_t  rice_history_mult;
  74     uint8_t  rice_initial_history;
  75     uint8_t  rice_limit;
  76
  77     int extra_bits;     /**< number of extra bits beyond 16-bit */
  78     int nb_samples;     /**< number of samples in the current frame */
  79 } ALACContext;
  80
  81 static inline unsigned int decode_scalar(BitstreamContext *bc, int k, int bps)
  82 {
  83     unsigned int x = get_unary_0_9(bc);
  84
  85     if (x > 8) { /* RICE THRESHOLD */
  86         /* use alternative encoding */
  87         x = bitstream_read(bc, bps);
  88     } else if (k != 1) {
  89         int extrabits = bitstream_peek(bc, k);
  90
  91         /* multiply x by 2^k - 1, as part of their strange algorithm */
  92         x = (x << k) - x;
  93
  94         if (extrabits > 1) {
  95             x += extrabits - 1;
  96             bitstream_skip(bc, k);
  97         } else
  98             bitstream_skip(bc, k - 1);
  99     }
 100     return x;
 101 }
 102
 103 static void rice_decompress(ALACContext *alac, int32_t *output_buffer,
 104                             int nb_samples, int bps, int rice_history_mult)
 105 {
 106     int i;
 107     unsigned int history = alac->rice_initial_history;
 108     int sign_modifier = 0;
 109
 110     for (i = 0; i < nb_samples; i++) {
 111         int k;
 112         unsigned int x;
 113
 114         /* calculate rice param and decode next value */
 115         k = av_log2((history >> 9) + 3);
 116         k = FFMIN(k, alac->rice_limit);
 117         x = decode_scalar(&alac->bc, k, bps);
 118         x += sign_modifier;
 119         sign_modifier = 0;
 120         output_buffer[i] = (x >> 1) ^ -(x & 1);
 121
 122         /* update the history */
 123         if (x > 0xffff)
 124             history = 0xffff;
 125         else
 126             history +=         x * rice_history_mult -
 127                        ((history * rice_history_mult) >> 9);
 128
 129         /* special case: there may be compressed blocks of 0 */
 130         if ((history < 128) && (i + 1 < nb_samples)) {
 131             int block_size;
 132
 133             /* calculate rice param and decode block size */
 134             k = 7 - av_log2(history) + ((history + 16) >> 6);
 135             k = FFMIN(k, alac->rice_limit);
 136             block_size = decode_scalar(&alac->bc, k, 16);
 137
 138             if (block_size > 0) {
 139                 if (block_size >= nb_samples - i) {
 140                     av_log(alac->avctx, AV_LOG_ERROR,
 141                            "invalid zero block size of %d %d %d\n", block_size,
 142                            nb_samples, i);
 143                     block_size = nb_samples - i - 1;
 144                 }
 145                 memset(&output_buffer[i + 1], 0,
 146                        block_size * sizeof(*output_buffer));
 147                 i += block_size;
 148             }
 149             if (block_size <= 0xffff)
 150                 sign_modifier = 1;
 151             history = 0;
 152         }
 153     }
 154 }
 155
 156 static inline int sign_only(int v)
 157 {
 158     return v ? FFSIGN(v) : 0;
 159 }
 160
 161 static void lpc_prediction(int32_t *error_buffer, int32_t *buffer_out,
 162                            int nb_samples, int bps, int16_t *lpc_coefs,
 163                            int lpc_order, int lpc_quant)
 164 {
 165     int i;
 166     int32_t *pred = buffer_out;
 167
 168     /* first sample always copies */
 169     *buffer_out = *error_buffer;
 170
 171     if (nb_samples <= 1)
 172         return;
 173
 174     if (!lpc_order) {
 175         memcpy(&buffer_out[1], &error_buffer[1],
 176                (nb_samples - 1) * sizeof(*buffer_out));
 177         return;
 178     }
 179
 180     if (lpc_order == 31) {
 181         /* simple 1st-order prediction */
 182         for (i = 1; i < nb_samples; i++) {
 183             buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i],
 184                                         bps);
 185         }
 186         return;
 187     }
 188
 189     /* read warm-up samples */
 190     for (i = 1; i <= lpc_order; i++)
 191         buffer_out[i] = sign_extend(buffer_out[i - 1] + error_buffer[i], bps);
 192
 193     /* NOTE: 4 and 8 are very common cases that could be optimized. */
 194
 195     for (; i < nb_samples; i++) {
 196         int j;
 197         int val = 0;
 198         int error_val = error_buffer[i];
 199         int error_sign;
 200         int d = *pred++;
 201
 202         /* LPC prediction */
 203         for (j = 0; j < lpc_order; j++)
 204             val += (pred[j] - d) * lpc_coefs[j];
 205         val = (val + (1 << (lpc_quant - 1))) >> lpc_quant;
 206         val += d + error_val;
 207         buffer_out[i] = sign_extend(val, bps);
 208
 209         /* adapt LPC coefficients */
 210         error_sign = sign_only(error_val);
 211         if (error_sign) {
 212             for (j = 0; j < lpc_order && error_val * error_sign > 0; j++) {
 213                 int sign;
 214                 val  = d - pred[j];
 215                 sign = sign_only(val) * error_sign;
 216                 lpc_coefs[j] -= sign;
 217                 val *= sign;
 218                 error_val -= (val >> lpc_quant) * (j + 1);
 219             }
 220         }
 221     }
 222 }
 223
 224 static void decorrelate_stereo(int32_t *buffer[2], int nb_samples,
 225                                int decorr_shift, int decorr_left_weight)
 226 {
 227     int i;
 228
 229     for (i = 0; i < nb_samples; i++) {
 230         int32_t a, b;
 231
 232         a = buffer[0][i];
 233         b = buffer[1][i];
 234
 235         a -= (b * decorr_left_weight) >> decorr_shift;
 236         b += a;
 237
 238         buffer[0][i] = b;
 239         buffer[1][i] = a;
 240     }
 241 }
 242
 243 static void append_extra_bits(int32_t *buffer[2], int32_t *extra_bits_buffer[2],
 244                               int extra_bits, int channels, int nb_samples)
 245 {
 246     int i, ch;
 247
 248     for (ch = 0; ch < channels; ch++)
 249         for (i = 0; i < nb_samples; i++)
 250             buffer[ch][i] = (buffer[ch][i] << extra_bits) | extra_bits_buffer[ch][i];
 251 }
 252
 253 static int decode_element(AVCodecContext *avctx, AVFrame *frame, int ch_index,
 254                           int channels)
 255 {
 256     ALACContext *alac = avctx->priv_data;
 257     int has_size, bps, is_compressed, decorr_shift, decorr_left_weight, ret;
 258     uint32_t output_samples;
 259     int i, ch;
 260
 261     bitstream_skip(&alac->bc, 4);  /* element instance tag */
 262     bitstream_skip(&alac->bc, 12); /* unused header bits */
 263
 264     /* the number of output samples is stored in the frame */
 265     has_size = bitstream_read_bit(&alac->bc);
 266
 267     alac->extra_bits = bitstream_read(&alac->bc, 2) << 3;
 268     bps = alac->sample_size - alac->extra_bits + channels - 1;
 269     if (bps > 32) {
 270         avpriv_report_missing_feature(avctx, "bps %d", bps);
 271         return AVERROR_PATCHWELCOME;
 272     }
 273
 274     /* whether the frame is compressed */
 275     is_compressed = !bitstream_read_bit(&alac->bc);
 276
 277     if (has_size)
 278         output_samples = bitstream_read(&alac->bc, 32);
 279     else
 280         output_samples = alac->max_samples_per_frame;
 281     if (!output_samples || output_samples > alac->max_samples_per_frame) {
 282         av_log(avctx, AV_LOG_ERROR, "invalid samples per frame: %"PRIu32"\n",
 283                output_samples);
 284         return AVERROR_INVALIDDATA;
 285     }
 286     if (!alac->nb_samples) {
 287         /* get output buffer */
 288         frame->nb_samples = output_samples;
 289         if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
 290             av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
 291             return ret;
 292         }
 293     } else if (output_samples != alac->nb_samples) {
 294         av_log(avctx, AV_LOG_ERROR, "sample count mismatch: %"PRIu32" != %d\n",
 295                output_samples, alac->nb_samples);
 296         return AVERROR_INVALIDDATA;
 297     }
 298     alac->nb_samples = output_samples;
 299     if (alac->sample_size > 16) {
 300         for (ch = 0; ch < channels; ch++)
 301             alac->output_samples_buffer[ch] = (int32_t *)frame->extended_data[ch_index + ch];
 302     }
 303
 304     if (is_compressed) {
 305         int16_t lpc_coefs[2][32];
 306         int lpc_order[2];
 307         int prediction_type[2];
 308         int lpc_quant[2];
 309         int rice_history_mult[2];
 310
 311         if (!alac->rice_limit) {
 312             avpriv_request_sample(alac->avctx,
 313                                   "Compression with rice limit 0");
 314             return AVERROR(ENOSYS);
 315         }
 316
 317         decorr_shift       = bitstream_read(&alac->bc, 8);
 318         decorr_left_weight = bitstream_read(&alac->bc, 8);
 319
 320         for (ch = 0; ch < channels; ch++) {
 321             prediction_type[ch]   = bitstream_read(&alac->bc, 4);
 322             lpc_quant[ch]         = bitstream_read(&alac->bc, 4);
 323             rice_history_mult[ch] = bitstream_read(&alac->bc, 3);
 324             lpc_order[ch]         = bitstream_read(&alac->bc, 5);
 325
 326             if (lpc_order[ch] >= alac->max_samples_per_frame)
 327                 return AVERROR_INVALIDDATA;
 328
 329             /* read the predictor table */
 330             for (i = lpc_order[ch] - 1; i >= 0; i--)
 331                 lpc_coefs[ch][i] = bitstream_read_signed(&alac->bc, 16);
 332         }
 333
 334         if (alac->extra_bits) {
 335             for (i = 0; i < alac->nb_samples; i++) {
 336                 for (ch = 0; ch < channels; ch++)
 337                     alac->extra_bits_buffer[ch][i] = bitstream_read(&alac->bc, alac->extra_bits);
 338             }
 339         }
 340         for (ch = 0; ch < channels; ch++) {
 341             rice_decompress(alac, alac->predict_error_buffer[ch],
 342                             alac->nb_samples, bps,
 343                             rice_history_mult[ch] * alac->rice_history_mult / 4);
 344
 345             /* adaptive FIR filter */
 346             if (prediction_type[ch] == 15) {
 347                 /* Prediction type 15 runs the adaptive FIR twice.
 348                  * The first pass uses the special-case coef_num = 31, while
 349                  * the second pass uses the coefs from the bitstream.
 350                  *
 351                  * However, this prediction type is not currently used by the
 352                  * reference encoder.
 353                  */
 354                 lpc_prediction(alac->predict_error_buffer[ch],
 355                                alac->predict_error_buffer[ch],
 356                                alac->nb_samples, bps, NULL, 31, 0);
 357             } else if (prediction_type[ch] > 0) {
 358                 av_log(avctx, AV_LOG_WARNING, "unknown prediction type: %i\n",
 359                        prediction_type[ch]);
 360             }
 361             lpc_prediction(alac->predict_error_buffer[ch],
 362                            alac->output_samples_buffer[ch], alac->nb_samples,
 363                            bps, lpc_coefs[ch], lpc_order[ch], lpc_quant[ch]);
 364         }
 365     } else {
 366         /* not compressed, easy case */
 367         for (i = 0; i < alac->nb_samples; i++) {
 368             for (ch = 0; ch < channels; ch++) {
 369                 alac->output_samples_buffer[ch][i] =
 370                          bitstream_read_signed(&alac->bc, alac->sample_size);
 371             }
 372         }
 373         alac->extra_bits   = 0;
 374         decorr_shift       = 0;
 375         decorr_left_weight = 0;
 376     }
 377
 378     if (channels == 2 && decorr_left_weight) {
 379         decorrelate_stereo(alac->output_samples_buffer, alac->nb_samples,
 380                            decorr_shift, decorr_left_weight);
 381     }
 382
 383     if (alac->extra_bits) {
 384         append_extra_bits(alac->output_samples_buffer, alac->extra_bits_buffer,
 385                           alac->extra_bits, channels, alac->nb_samples);
 386     }
 387
 388     switch(alac->sample_size) {
 389     case 16: {
 390         for (ch = 0; ch < channels; ch++) {
 391             int16_t *outbuffer = (int16_t *)frame->extended_data[ch_index + ch];
 392             for (i = 0; i < alac->nb_samples; i++)
 393                 *outbuffer++ = alac->output_samples_buffer[ch][i];
 394         }}
 395         break;
 396     case 24: {
 397         for (ch = 0; ch < channels; ch++) {
 398             for (i = 0; i < alac->nb_samples; i++)
 399                 alac->output_samples_buffer[ch][i] <<= 8;
 400         }}
 401         break;
 402     }
 403
 404     return 0;
 405 }
 406
 407 static int alac_decode_frame(AVCodecContext *avctx, void *data,
 408                              int *got_frame_ptr, AVPacket *avpkt)
 409 {
 410     ALACContext *alac = avctx->priv_data;
 411     AVFrame *frame    = data;
 412     enum AlacRawDataBlockType element;
 413     int channels;
 414     int ch, ret, got_end;
 415
 416     bitstream_init8(&alac->bc, avpkt->data, avpkt->size);
 417
 418     got_end = 0;
 419     alac->nb_samples = 0;
 420     ch = 0;
 421     while (bitstream_bits_left(&alac->bc) >= 3) {
 422         element = bitstream_read(&alac->bc, 3);
 423         if (element == TYPE_END) {
 424             got_end = 1;
 425             break;
 426         }
 427         if (element > TYPE_CPE && element != TYPE_LFE) {
 428             avpriv_report_missing_feature(avctx, "Syntax element %d", element);
 429             return AVERROR_PATCHWELCOME;
 430         }
 431
 432         channels = (element == TYPE_CPE) ? 2 : 1;
 433         if (ch + channels > alac->channels ||
 434             ff_alac_channel_layout_offsets[alac->channels - 1][ch] + channels > alac->channels) {
 435             av_log(avctx, AV_LOG_ERROR, "invalid element channel count\n");
 436             return AVERROR_INVALIDDATA;
 437         }
 438
 439         ret = decode_element(avctx, frame,
 440                              ff_alac_channel_layout_offsets[alac->channels - 1][ch],
 441                              channels);
 442         if (ret < 0 && bitstream_bits_left(&alac->bc))
 443             return ret;
 444
 445         ch += channels;
 446     }
 447     if (!got_end) {
 448         av_log(avctx, AV_LOG_ERROR, "no end tag found. incomplete packet.\n");
 449         return AVERROR_INVALIDDATA;
 450     }
 451     if (!alac->nb_samples) {
 452         av_log(avctx, AV_LOG_ERROR, "No decodable data in the packet\n");
 453         return AVERROR_INVALIDDATA;
 454     }
 455
 456     if (avpkt->size * 8 - bitstream_tell(&alac->bc) > 8) {
 457         av_log(avctx, AV_LOG_ERROR, "Error : %d bits left\n",
 458                avpkt->size * 8 - bitstream_tell(&alac->bc));
 459     }
 460
 461     *got_frame_ptr = 1;
 462
 463     return avpkt->size;
 464 }
 465
 466 static av_cold int alac_decode_close(AVCodecContext *avctx)
 467 {
 468     ALACContext *alac = avctx->priv_data;
 469
 470     int ch;
 471     for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
 472         av_freep(&alac->predict_error_buffer[ch]);
 473         if (alac->sample_size == 16)
 474             av_freep(&alac->output_samples_buffer[ch]);
 475         av_freep(&alac->extra_bits_buffer[ch]);
 476     }
 477
 478     return 0;
 479 }
 480
 481 static int allocate_buffers(ALACContext *alac)
 482 {
 483     int ch;
 484     int buf_size = alac->max_samples_per_frame * sizeof(int32_t);
 485
 486     for (ch = 0; ch < FFMIN(alac->channels, 2); ch++) {
 487         FF_ALLOC_OR_GOTO(alac->avctx, alac->predict_error_buffer[ch],
 488                          buf_size, buf_alloc_fail);
 489
 490         if (alac->sample_size == 16) {
 491             FF_ALLOC_OR_GOTO(alac->avctx, alac->output_samples_buffer[ch],
 492                              buf_size, buf_alloc_fail);
 493         }
 494
 495         FF_ALLOC_OR_GOTO(alac->avctx, alac->extra_bits_buffer[ch],
 496                          buf_size, buf_alloc_fail);
 497     }
 498     return 0;
 499 buf_alloc_fail:
 500     alac_decode_close(alac->avctx);
 501     return AVERROR(ENOMEM);
 502 }
 503
 504 static int alac_set_info(ALACContext *alac)
 505 {
 506     GetByteContext gb;
 507
 508     bytestream2_init(&gb, alac->avctx->extradata,
 509                      alac->avctx->extradata_size);
 510
 511     bytestream2_skipu(&gb, 12); // size:4, alac:4, version:4
 512
 513     alac->max_samples_per_frame = bytestream2_get_be32u(&gb);
 514     if (!alac->max_samples_per_frame ||
 515         alac->max_samples_per_frame > INT_MAX / sizeof(int32_t)) {
 516         av_log(alac->avctx, AV_LOG_ERROR,
 517                "max samples per frame invalid: %"PRIu32"\n",
 518                alac->max_samples_per_frame);
 519         return AVERROR_INVALIDDATA;
 520     }
 521     bytestream2_skipu(&gb, 1);  // compatible version
 522     alac->sample_size          = bytestream2_get_byteu(&gb);
 523     alac->rice_history_mult    = bytestream2_get_byteu(&gb);
 524     alac->rice_initial_history = bytestream2_get_byteu(&gb);
 525     alac->rice_limit           = bytestream2_get_byteu(&gb);
 526     alac->channels             = bytestream2_get_byteu(&gb);
 527     bytestream2_get_be16u(&gb); // maxRun
 528     bytestream2_get_be32u(&gb); // max coded frame size
 529     bytestream2_get_be32u(&gb); // average bitrate
 530     bytestream2_get_be32u(&gb); // samplerate
 531
 532     return 0;
 533 }
 534
 535 static av_cold int alac_decode_init(AVCodecContext * avctx)
 536 {
 537     int ret;
 538     ALACContext *alac = avctx->priv_data;
 539     alac->avctx = avctx;
 540
 541     /* initialize from the extradata */
 542     if (alac->avctx->extradata_size < ALAC_EXTRADATA_SIZE) {
 543         av_log(avctx, AV_LOG_ERROR, "alac: extradata is too small\n");
 544         return AVERROR_INVALIDDATA;
 545     }
 546     if (alac_set_info(alac)) {
 547         av_log(avctx, AV_LOG_ERROR, "alac: set_info failed\n");
 548         return -1;
 549     }
 550
 551     switch (alac->sample_size) {
 552     case 16: avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
 553              break;
 554     case 24:
 555     case 32: avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
 556              break;
 557     default: avpriv_request_sample(avctx, "Sample depth %d", alac->sample_size);
 558              return AVERROR_PATCHWELCOME;
 559     }
 560     avctx->bits_per_raw_sample = alac->sample_size;
 561
 562     if (alac->channels < 1) {
 563         av_log(avctx, AV_LOG_WARNING, "Invalid channel count\n");
 564         alac->channels = avctx->channels;
 565     } else {
 566         if (alac->channels > ALAC_MAX_CHANNELS)
 567             alac->channels = avctx->channels;
 568         else
 569             avctx->channels = alac->channels;
 570     }
 571     if (avctx->channels > ALAC_MAX_CHANNELS) {
 572         avpriv_report_missing_feature(avctx, "Channel count %d",
 573                                       avctx->channels);
 574         return AVERROR_PATCHWELCOME;
 575     }
 576     avctx->channel_layout = ff_alac_channel_layouts[alac->channels - 1];
 577
 578     if ((ret = allocate_buffers(alac)) < 0) {
 579         av_log(avctx, AV_LOG_ERROR, "Error allocating buffers\n");
 580         return ret;
 581     }
 582
 583     return 0;
 584 }
 585
 586 AVCodec ff_alac_decoder = {
 587     .name           = "alac",
 588     .long_name      = NULL_IF_CONFIG_SMALL("ALAC (Apple Lossless Audio Codec)"),
 589     .type           = AVMEDIA_TYPE_AUDIO,
 590     .id             = AV_CODEC_ID_ALAC,
 591     .priv_data_size = sizeof(ALACContext),
 592     .init           = alac_decode_init,
 593     .close          = alac_decode_close,
 594     .decode         = alac_decode_frame,
 595     .capabilities   = AV_CODEC_CAP_DR1,
 596 };