git.sesse.net Git - ffmpeg/blob - libavcodec/shorten.c

   1 /*
   2  * Shorten decoder
   3  * Copyright (c) 2005 Jeff Muizelaar
   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  * Shorten decoder
  25  * @author Jeff Muizelaar
  26  *
  27  */
  28
  29 #define DEBUG
  30 #include <limits.h>
  31 #include "avcodec.h"
  32 #include "get_bits.h"
  33 #include "golomb.h"
  34
  35 #define MAX_CHANNELS 8
  36 #define MAX_BLOCKSIZE 65535
  37
  38 #define OUT_BUFFER_SIZE 16384
  39
  40 #define ULONGSIZE 2
  41
  42 #define WAVE_FORMAT_PCM 0x0001
  43
  44 #define DEFAULT_BLOCK_SIZE 256
  45
  46 #define TYPESIZE 4
  47 #define CHANSIZE 0
  48 #define LPCQSIZE 2
  49 #define ENERGYSIZE 3
  50 #define BITSHIFTSIZE 2
  51
  52 #define TYPE_S16HL 3
  53 #define TYPE_S16LH 5
  54
  55 #define NWRAP 3
  56 #define NSKIPSIZE 1
  57
  58 #define LPCQUANT 5
  59 #define V2LPCQOFFSET (1 << LPCQUANT)
  60
  61 #define FNSIZE 2
  62 #define FN_DIFF0        0
  63 #define FN_DIFF1        1
  64 #define FN_DIFF2        2
  65 #define FN_DIFF3        3
  66 #define FN_QUIT         4
  67 #define FN_BLOCKSIZE    5
  68 #define FN_BITSHIFT     6
  69 #define FN_QLPC         7
  70 #define FN_ZERO         8
  71 #define FN_VERBATIM     9
  72
  73 #define VERBATIM_CKSIZE_SIZE 5
  74 #define VERBATIM_BYTE_SIZE 8
  75 #define CANONICAL_HEADER_SIZE 44
  76
  77 typedef struct ShortenContext {
  78     AVCodecContext *avctx;
  79     GetBitContext gb;
  80
  81     int min_framesize, max_framesize;
  82     int channels;
  83
  84     int32_t *decoded[MAX_CHANNELS];
  85     int32_t *offset[MAX_CHANNELS];
  86     int *coeffs;
  87     uint8_t *bitstream;
  88     int bitstream_size;
  89     int bitstream_index;
  90     unsigned int allocated_bitstream_size;
  91     int header_size;
  92     uint8_t header[OUT_BUFFER_SIZE];
  93     int version;
  94     int cur_chan;
  95     int bitshift;
  96     int nmean;
  97     int internal_ftype;
  98     int nwrap;
  99     int blocksize;
 100     int bitindex;
 101     int32_t lpcqoffset;
 102 } ShortenContext;
 103
 104 static av_cold int shorten_decode_init(AVCodecContext * avctx)
 105 {
 106     ShortenContext *s = avctx->priv_data;
 107     s->avctx = avctx;
 108     avctx->sample_fmt = SAMPLE_FMT_S16;
 109
 110     return 0;
 111 }
 112
 113 static int allocate_buffers(ShortenContext *s)
 114 {
 115     int i, chan;
 116     int *coeffs;
 117
 118     for (chan=0; chan<s->channels; chan++) {
 119         if(FFMAX(1, s->nmean) >= UINT_MAX/sizeof(int32_t)){
 120             av_log(s->avctx, AV_LOG_ERROR, "nmean too large\n");
 121             return -1;
 122         }
 123         if(s->blocksize + s->nwrap >= UINT_MAX/sizeof(int32_t) || s->blocksize + s->nwrap <= (unsigned)s->nwrap){
 124             av_log(s->avctx, AV_LOG_ERROR, "s->blocksize + s->nwrap too large\n");
 125             return -1;
 126         }
 127
 128         s->offset[chan] = av_realloc(s->offset[chan], sizeof(int32_t)*FFMAX(1, s->nmean));
 129
 130         s->decoded[chan] = av_realloc(s->decoded[chan], sizeof(int32_t)*(s->blocksize + s->nwrap));
 131         for (i=0; i<s->nwrap; i++)
 132             s->decoded[chan][i] = 0;
 133         s->decoded[chan] += s->nwrap;
 134     }
 135
 136     coeffs = av_realloc(s->coeffs, s->nwrap * sizeof(*s->coeffs));
 137     if (!coeffs)
 138         return AVERROR(ENOMEM);
 139     s->coeffs = coeffs;
 140
 141     return 0;
 142 }
 143
 144
 145 static inline unsigned int get_uint(ShortenContext *s, int k)
 146 {
 147     if (s->version != 0)
 148         k = get_ur_golomb_shorten(&s->gb, ULONGSIZE);
 149     return get_ur_golomb_shorten(&s->gb, k);
 150 }
 151
 152
 153 static void fix_bitshift(ShortenContext *s, int32_t *buffer)
 154 {
 155     int i;
 156
 157     if (s->bitshift != 0)
 158         for (i = 0; i < s->blocksize; i++)
 159             buffer[s->nwrap + i] <<= s->bitshift;
 160 }
 161
 162
 163 static void init_offset(ShortenContext *s)
 164 {
 165     int32_t mean = 0;
 166     int  chan, i;
 167     int nblock = FFMAX(1, s->nmean);
 168     /* initialise offset */
 169     switch (s->internal_ftype)
 170     {
 171         case TYPE_S16HL:
 172         case TYPE_S16LH:
 173             mean = 0;
 174             break;
 175         default:
 176             av_log(s->avctx, AV_LOG_ERROR, "unknown audio type");
 177             abort();
 178     }
 179
 180     for (chan = 0; chan < s->channels; chan++)
 181         for (i = 0; i < nblock; i++)
 182             s->offset[chan][i] = mean;
 183 }
 184
 185 static inline int get_le32(GetBitContext *gb)
 186 {
 187     return av_bswap32(get_bits_long(gb, 32));
 188 }
 189
 190 static inline short get_le16(GetBitContext *gb)
 191 {
 192     return av_bswap16(get_bits_long(gb, 16));
 193 }
 194
 195 static int decode_wave_header(AVCodecContext *avctx, uint8_t *header, int header_size)
 196 {
 197     GetBitContext hb;
 198     int len;
 199     int chunk_size;
 200     short wave_format;
 201
 202     init_get_bits(&hb, header, header_size*8);
 203     if (get_le32(&hb) != MKTAG('R','I','F','F')) {
 204         av_log(avctx, AV_LOG_ERROR, "missing RIFF tag\n");
 205         return -1;
 206     }
 207
 208     chunk_size = get_le32(&hb);
 209
 210     if (get_le32(&hb) != MKTAG('W','A','V','E')) {
 211         av_log(avctx, AV_LOG_ERROR, "missing WAVE tag\n");
 212         return -1;
 213     }
 214
 215     while (get_le32(&hb) != MKTAG('f','m','t',' ')) {
 216         len = get_le32(&hb);
 217         skip_bits(&hb, 8*len);
 218     }
 219     len = get_le32(&hb);
 220
 221     if (len < 16) {
 222         av_log(avctx, AV_LOG_ERROR, "fmt chunk was too short\n");
 223         return -1;
 224     }
 225
 226     wave_format = get_le16(&hb);
 227
 228     switch (wave_format) {
 229         case WAVE_FORMAT_PCM:
 230             break;
 231         default:
 232             av_log(avctx, AV_LOG_ERROR, "unsupported wave format\n");
 233             return -1;
 234     }
 235
 236     avctx->channels = get_le16(&hb);
 237     avctx->sample_rate = get_le32(&hb);
 238     avctx->bit_rate = get_le32(&hb) * 8;
 239     avctx->block_align = get_le16(&hb);
 240     avctx->bits_per_coded_sample = get_le16(&hb);
 241
 242     if (avctx->bits_per_coded_sample != 16) {
 243         av_log(avctx, AV_LOG_ERROR, "unsupported number of bits per sample\n");
 244         return -1;
 245     }
 246
 247     len -= 16;
 248     if (len > 0)
 249         av_log(avctx, AV_LOG_INFO, "%d header bytes unparsed\n", len);
 250
 251     return 0;
 252 }
 253
 254 static int16_t * interleave_buffer(int16_t *samples, int nchan, int blocksize, int32_t **buffer) {
 255     int i, chan;
 256     for (i=0; i<blocksize; i++)
 257         for (chan=0; chan < nchan; chan++)
 258             *samples++ = FFMIN(buffer[chan][i], 32768);
 259     return samples;
 260 }
 261
 262 static void decode_subframe_lpc(ShortenContext *s, int channel, int residual_size, int pred_order)
 263 {
 264     int sum, i, j;
 265     int *coeffs = s->coeffs;
 266
 267     for (i=0; i<pred_order; i++)
 268         coeffs[i] = get_sr_golomb_shorten(&s->gb, LPCQUANT);
 269
 270     for (i=0; i < s->blocksize; i++) {
 271         sum = s->lpcqoffset;
 272         for (j=0; j<pred_order; j++)
 273             sum += coeffs[j] * s->decoded[channel][i-j-1];
 274         s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + (sum >> LPCQUANT);
 275     }
 276 }
 277
 278
 279 static int shorten_decode_frame(AVCodecContext *avctx,
 280         void *data, int *data_size,
 281         AVPacket *avpkt)
 282 {
 283     const uint8_t *buf = avpkt->data;
 284     int buf_size = avpkt->size;
 285     ShortenContext *s = avctx->priv_data;
 286     int i, input_buf_size = 0;
 287     int16_t *samples = data;
 288     if(s->max_framesize == 0){
 289         s->max_framesize= 1024; // should hopefully be enough for the first header
 290         s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
 291     }
 292
 293     if(1 && s->max_framesize){//FIXME truncated
 294         buf_size= FFMIN(buf_size, s->max_framesize - s->bitstream_size);
 295         input_buf_size= buf_size;
 296
 297         if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
 298             //                printf("memmove\n");
 299             memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
 300             s->bitstream_index=0;
 301         }
 302         memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
 303         buf= &s->bitstream[s->bitstream_index];
 304         buf_size += s->bitstream_size;
 305         s->bitstream_size= buf_size;
 306
 307         if(buf_size < s->max_framesize){
 308             //dprintf(avctx, "wanna more data ... %d\n", buf_size);
 309             *data_size = 0;
 310             return input_buf_size;
 311         }
 312     }
 313     init_get_bits(&s->gb, buf, buf_size*8);
 314     skip_bits(&s->gb, s->bitindex);
 315     if (!s->blocksize)
 316     {
 317         int maxnlpc = 0;
 318         /* shorten signature */
 319         if (get_bits_long(&s->gb, 32) != AV_RB32("ajkg")) {
 320             av_log(s->avctx, AV_LOG_ERROR, "missing shorten magic 'ajkg'\n");
 321             return -1;
 322         }
 323
 324         s->lpcqoffset = 0;
 325         s->blocksize = DEFAULT_BLOCK_SIZE;
 326         s->channels = 1;
 327         s->nmean = -1;
 328         s->version = get_bits(&s->gb, 8);
 329         s->internal_ftype = get_uint(s, TYPESIZE);
 330
 331         s->channels = get_uint(s, CHANSIZE);
 332         if (s->channels > MAX_CHANNELS) {
 333             av_log(s->avctx, AV_LOG_ERROR, "too many channels: %d\n", s->channels);
 334             return -1;
 335         }
 336
 337         /* get blocksize if version > 0 */
 338         if (s->version > 0) {
 339             int skip_bytes;
 340             s->blocksize = get_uint(s, av_log2(DEFAULT_BLOCK_SIZE));
 341             maxnlpc = get_uint(s, LPCQSIZE);
 342             s->nmean = get_uint(s, 0);
 343
 344             skip_bytes = get_uint(s, NSKIPSIZE);
 345             for (i=0; i<skip_bytes; i++) {
 346                 skip_bits(&s->gb, 8);
 347             }
 348         }
 349         s->nwrap = FFMAX(NWRAP, maxnlpc);
 350
 351         if (allocate_buffers(s))
 352             return -1;
 353
 354         init_offset(s);
 355
 356         if (s->version > 1)
 357             s->lpcqoffset = V2LPCQOFFSET;
 358
 359         if (get_ur_golomb_shorten(&s->gb, FNSIZE) != FN_VERBATIM) {
 360             av_log(s->avctx, AV_LOG_ERROR, "missing verbatim section at beginning of stream\n");
 361             return -1;
 362         }
 363
 364         s->header_size = get_ur_golomb_shorten(&s->gb, VERBATIM_CKSIZE_SIZE);
 365         if (s->header_size >= OUT_BUFFER_SIZE || s->header_size < CANONICAL_HEADER_SIZE) {
 366             av_log(s->avctx, AV_LOG_ERROR, "header is wrong size: %d\n", s->header_size);
 367             return -1;
 368         }
 369
 370         for (i=0; i<s->header_size; i++)
 371             s->header[i] = (char)get_ur_golomb_shorten(&s->gb, VERBATIM_BYTE_SIZE);
 372
 373         if (decode_wave_header(avctx, s->header, s->header_size) < 0)
 374             return -1;
 375
 376         s->cur_chan = 0;
 377         s->bitshift = 0;
 378     }
 379     else
 380     {
 381         int cmd;
 382         int len;
 383         cmd = get_ur_golomb_shorten(&s->gb, FNSIZE);
 384         switch (cmd) {
 385             case FN_ZERO:
 386             case FN_DIFF0:
 387             case FN_DIFF1:
 388             case FN_DIFF2:
 389             case FN_DIFF3:
 390             case FN_QLPC:
 391                 {
 392                     int residual_size = 0;
 393                     int channel = s->cur_chan;
 394                     int32_t coffset;
 395                     if (cmd != FN_ZERO) {
 396                         residual_size = get_ur_golomb_shorten(&s->gb, ENERGYSIZE);
 397                         /* this is a hack as version 0 differed in defintion of get_sr_golomb_shorten */
 398                         if (s->version == 0)
 399                             residual_size--;
 400                     }
 401
 402                     if (s->nmean == 0)
 403                         coffset = s->offset[channel][0];
 404                     else {
 405                         int32_t sum = (s->version < 2) ? 0 : s->nmean / 2;
 406                         for (i=0; i<s->nmean; i++)
 407                             sum += s->offset[channel][i];
 408                         coffset = sum / s->nmean;
 409                         if (s->version >= 2)
 410                             coffset >>= FFMIN(1, s->bitshift);
 411                     }
 412                     switch (cmd) {
 413                         case FN_ZERO:
 414                             for (i=0; i<s->blocksize; i++)
 415                                 s->decoded[channel][i] = 0;
 416                             break;
 417                         case FN_DIFF0:
 418                             for (i=0; i<s->blocksize; i++)
 419                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + coffset;
 420                             break;
 421                         case FN_DIFF1:
 422                             for (i=0; i<s->blocksize; i++)
 423                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + s->decoded[channel][i - 1];
 424                             break;
 425                         case FN_DIFF2:
 426                             for (i=0; i<s->blocksize; i++)
 427                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + 2*s->decoded[channel][i-1]
 428                                                                                                       -   s->decoded[channel][i-2];
 429                             break;
 430                         case FN_DIFF3:
 431                             for (i=0; i<s->blocksize; i++)
 432                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + 3*s->decoded[channel][i-1]
 433                                                                                                       - 3*s->decoded[channel][i-2]
 434                                                                                                       +   s->decoded[channel][i-3];
 435                             break;
 436                         case FN_QLPC:
 437                             {
 438                                 int pred_order = get_ur_golomb_shorten(&s->gb, LPCQSIZE);
 439                                 if (pred_order > s->nwrap) {
 440                                     av_log(avctx, AV_LOG_ERROR,
 441                                            "invalid pred_order %d\n",
 442                                            pred_order);
 443                                     return -1;
 444                                 }
 445                                 for (i=0; i<pred_order; i++)
 446                                     s->decoded[channel][i - pred_order] -= coffset;
 447                                 decode_subframe_lpc(s, channel, residual_size, pred_order);
 448                                 if (coffset != 0)
 449                                     for (i=0; i < s->blocksize; i++)
 450                                         s->decoded[channel][i] += coffset;
 451                             }
 452                     }
 453                     if (s->nmean > 0) {
 454                         int32_t sum = (s->version < 2) ? 0 : s->blocksize / 2;
 455                         for (i=0; i<s->blocksize; i++)
 456                             sum += s->decoded[channel][i];
 457
 458                         for (i=1; i<s->nmean; i++)
 459                             s->offset[channel][i-1] = s->offset[channel][i];
 460
 461                         if (s->version < 2)
 462                             s->offset[channel][s->nmean - 1] = sum / s->blocksize;
 463                         else
 464                             s->offset[channel][s->nmean - 1] = (sum / s->blocksize) << s->bitshift;
 465                     }
 466                     for (i=-s->nwrap; i<0; i++)
 467                         s->decoded[channel][i] = s->decoded[channel][i + s->blocksize];
 468
 469                     fix_bitshift(s, s->decoded[channel]);
 470
 471                     s->cur_chan++;
 472                     if (s->cur_chan == s->channels) {
 473                         samples = interleave_buffer(samples, s->channels, s->blocksize, s->decoded);
 474                         s->cur_chan = 0;
 475                         goto frame_done;
 476                     }
 477                     break;
 478                 }
 479                 break;
 480             case FN_VERBATIM:
 481                 len = get_ur_golomb_shorten(&s->gb, VERBATIM_CKSIZE_SIZE);
 482                 while (len--) {
 483                     get_ur_golomb_shorten(&s->gb, VERBATIM_BYTE_SIZE);
 484                 }
 485                 break;
 486             case FN_BITSHIFT:
 487                 s->bitshift = get_ur_golomb_shorten(&s->gb, BITSHIFTSIZE);
 488                 break;
 489             case FN_BLOCKSIZE:
 490                 s->blocksize = get_uint(s, av_log2(s->blocksize));
 491                 break;
 492             case FN_QUIT:
 493                 *data_size = 0;
 494                 return buf_size;
 495                 break;
 496             default:
 497                 av_log(avctx, AV_LOG_ERROR, "unknown shorten function %d\n", cmd);
 498                 return -1;
 499                 break;
 500         }
 501     }
 502 frame_done:
 503     *data_size = (int8_t *)samples - (int8_t *)data;
 504
 505     //    s->last_blocksize = s->blocksize;
 506     s->bitindex = get_bits_count(&s->gb) - 8*((get_bits_count(&s->gb))/8);
 507     i= (get_bits_count(&s->gb))/8;
 508     if (i > buf_size) {
 509         av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
 510         s->bitstream_size=0;
 511         s->bitstream_index=0;
 512         return -1;
 513     }
 514     if (s->bitstream_size) {
 515         s->bitstream_index += i;
 516         s->bitstream_size  -= i;
 517         return input_buf_size;
 518     } else
 519         return i;
 520 }
 521
 522 static av_cold int shorten_decode_close(AVCodecContext *avctx)
 523 {
 524     ShortenContext *s = avctx->priv_data;
 525     int i;
 526
 527     for (i = 0; i < s->channels; i++) {
 528         s->decoded[i] -= s->nwrap;
 529         av_freep(&s->decoded[i]);
 530         av_freep(&s->offset[i]);
 531     }
 532     av_freep(&s->bitstream);
 533     av_freep(&s->coeffs);
 534     return 0;
 535 }
 536
 537 static void shorten_flush(AVCodecContext *avctx){
 538     ShortenContext *s = avctx->priv_data;
 539
 540     s->bitstream_size=
 541         s->bitstream_index= 0;
 542 }
 543
 544 AVCodec shorten_decoder = {
 545     "shorten",
 546     AVMEDIA_TYPE_AUDIO,
 547     CODEC_ID_SHORTEN,
 548     sizeof(ShortenContext),
 549     shorten_decode_init,
 550     NULL,
 551     shorten_decode_close,
 552     shorten_decode_frame,
 553     .flush= shorten_flush,
 554     .long_name= NULL_IF_CONFIG_SMALL("Shorten"),
 555 };