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 shorten.c
  24  * Shorten decoder
  25  * @author Jeff Muizelaar
  26  *
  27  */
  28
  29 #define DEBUG
  30 #include <limits.h>
  31 #include "avcodec.h"
  32 #include "bitstream.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     uint8_t *bitstream;
  87     int bitstream_size;
  88     int bitstream_index;
  89     unsigned int allocated_bitstream_size;
  90     int header_size;
  91     uint8_t header[OUT_BUFFER_SIZE];
  92     int version;
  93     int cur_chan;
  94     int bitshift;
  95     int nmean;
  96     int internal_ftype;
  97     int nwrap;
  98     int blocksize;
  99     int bitindex;
 100     int32_t lpcqoffset;
 101 } ShortenContext;
 102
 103 static av_cold int shorten_decode_init(AVCodecContext * avctx)
 104 {
 105     ShortenContext *s = avctx->priv_data;
 106     s->avctx = avctx;
 107     avctx->sample_fmt = SAMPLE_FMT_S16;
 108
 109     return 0;
 110 }
 111
 112 static int allocate_buffers(ShortenContext *s)
 113 {
 114     int i, chan;
 115     for (chan=0; chan<s->channels; chan++) {
 116         if(FFMAX(1, s->nmean) >= UINT_MAX/sizeof(int32_t)){
 117             av_log(s->avctx, AV_LOG_ERROR, "nmean too large\n");
 118             return -1;
 119         }
 120         if(s->blocksize + s->nwrap >= UINT_MAX/sizeof(int32_t) || s->blocksize + s->nwrap <= (unsigned)s->nwrap){
 121             av_log(s->avctx, AV_LOG_ERROR, "s->blocksize + s->nwrap too large\n");
 122             return -1;
 123         }
 124
 125         s->offset[chan] = av_realloc(s->offset[chan], sizeof(int32_t)*FFMAX(1, s->nmean));
 126
 127         s->decoded[chan] = av_realloc(s->decoded[chan], sizeof(int32_t)*(s->blocksize + s->nwrap));
 128         for (i=0; i<s->nwrap; i++)
 129             s->decoded[chan][i] = 0;
 130         s->decoded[chan] += s->nwrap;
 131     }
 132     return 0;
 133 }
 134
 135
 136 static inline unsigned int get_uint(ShortenContext *s, int k)
 137 {
 138     if (s->version != 0)
 139         k = get_ur_golomb_shorten(&s->gb, ULONGSIZE);
 140     return get_ur_golomb_shorten(&s->gb, k);
 141 }
 142
 143
 144 static void fix_bitshift(ShortenContext *s, int32_t *buffer)
 145 {
 146     int i;
 147
 148     if (s->bitshift != 0)
 149         for (i = 0; i < s->blocksize; i++)
 150             buffer[s->nwrap + i] <<= s->bitshift;
 151 }
 152
 153
 154 static void init_offset(ShortenContext *s)
 155 {
 156     int32_t mean = 0;
 157     int  chan, i;
 158     int nblock = FFMAX(1, s->nmean);
 159     /* initialise offset */
 160     switch (s->internal_ftype)
 161     {
 162         case TYPE_S16HL:
 163         case TYPE_S16LH:
 164             mean = 0;
 165             break;
 166         default:
 167             av_log(s->avctx, AV_LOG_ERROR, "unknown audio type");
 168             abort();
 169     }
 170
 171     for (chan = 0; chan < s->channels; chan++)
 172         for (i = 0; i < nblock; i++)
 173             s->offset[chan][i] = mean;
 174 }
 175
 176 static inline int get_le32(GetBitContext *gb)
 177 {
 178     return bswap_32(get_bits_long(gb, 32));
 179 }
 180
 181 static inline short get_le16(GetBitContext *gb)
 182 {
 183     return bswap_16(get_bits_long(gb, 16));
 184 }
 185
 186 static int decode_wave_header(AVCodecContext *avctx, uint8_t *header, int header_size)
 187 {
 188     GetBitContext hb;
 189     int len;
 190     int chunk_size;
 191     short wave_format;
 192
 193     init_get_bits(&hb, header, header_size*8);
 194     if (get_le32(&hb) != MKTAG('R','I','F','F')) {
 195         av_log(avctx, AV_LOG_ERROR, "missing RIFF tag\n");
 196         return -1;
 197     }
 198
 199     chunk_size = get_le32(&hb);
 200
 201     if (get_le32(&hb) != MKTAG('W','A','V','E')) {
 202         av_log(avctx, AV_LOG_ERROR, "missing WAVE tag\n");
 203         return -1;
 204     }
 205
 206     while (get_le32(&hb) != MKTAG('f','m','t',' ')) {
 207         len = get_le32(&hb);
 208         skip_bits(&hb, 8*len);
 209     }
 210     len = get_le32(&hb);
 211
 212     if (len < 16) {
 213         av_log(avctx, AV_LOG_ERROR, "fmt chunk was too short\n");
 214         return -1;
 215     }
 216
 217     wave_format = get_le16(&hb);
 218
 219     switch (wave_format) {
 220         case WAVE_FORMAT_PCM:
 221             break;
 222         default:
 223             av_log(avctx, AV_LOG_ERROR, "unsupported wave format\n");
 224             return -1;
 225     }
 226
 227     avctx->channels = get_le16(&hb);
 228     avctx->sample_rate = get_le32(&hb);
 229     avctx->bit_rate = get_le32(&hb) * 8;
 230     avctx->block_align = get_le16(&hb);
 231     avctx->bits_per_coded_sample = get_le16(&hb);
 232
 233     if (avctx->bits_per_coded_sample != 16) {
 234         av_log(avctx, AV_LOG_ERROR, "unsupported number of bits per sample\n");
 235         return -1;
 236     }
 237
 238     len -= 16;
 239     if (len > 0)
 240         av_log(avctx, AV_LOG_INFO, "%d header bytes unparsed\n", len);
 241
 242     return 0;
 243 }
 244
 245 static int16_t * interleave_buffer(int16_t *samples, int nchan, int blocksize, int32_t **buffer) {
 246     int i, chan;
 247     for (i=0; i<blocksize; i++)
 248         for (chan=0; chan < nchan; chan++)
 249             *samples++ = FFMIN(buffer[chan][i], 32768);
 250     return samples;
 251 }
 252
 253 static void decode_subframe_lpc(ShortenContext *s, int channel, int residual_size, int pred_order)
 254 {
 255     int sum, i, j;
 256     int coeffs[pred_order];
 257
 258     for (i=0; i<pred_order; i++)
 259         coeffs[i] = get_sr_golomb_shorten(&s->gb, LPCQUANT);
 260
 261     for (i=0; i < s->blocksize; i++) {
 262         sum = s->lpcqoffset;
 263         for (j=0; j<pred_order; j++)
 264             sum += coeffs[j] * s->decoded[channel][i-j-1];
 265         s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + (sum >> LPCQUANT);
 266     }
 267 }
 268
 269
 270 static int shorten_decode_frame(AVCodecContext *avctx,
 271         void *data, int *data_size,
 272         const uint8_t *buf, int buf_size)
 273 {
 274     ShortenContext *s = avctx->priv_data;
 275     int i, input_buf_size = 0;
 276     int16_t *samples = data;
 277     if(s->max_framesize == 0){
 278         s->max_framesize= 1024; // should hopefully be enough for the first header
 279         s->bitstream= av_fast_realloc(s->bitstream, &s->allocated_bitstream_size, s->max_framesize);
 280     }
 281
 282     if(1 && s->max_framesize){//FIXME truncated
 283         buf_size= FFMIN(buf_size, s->max_framesize - s->bitstream_size);
 284         input_buf_size= buf_size;
 285
 286         if(s->bitstream_index + s->bitstream_size + buf_size > s->allocated_bitstream_size){
 287             //                printf("memmove\n");
 288             memmove(s->bitstream, &s->bitstream[s->bitstream_index], s->bitstream_size);
 289             s->bitstream_index=0;
 290         }
 291         memcpy(&s->bitstream[s->bitstream_index + s->bitstream_size], buf, buf_size);
 292         buf= &s->bitstream[s->bitstream_index];
 293         buf_size += s->bitstream_size;
 294         s->bitstream_size= buf_size;
 295
 296         if(buf_size < s->max_framesize){
 297             //dprintf(avctx, "wanna more data ... %d\n", buf_size);
 298             *data_size = 0;
 299             return input_buf_size;
 300         }
 301     }
 302     init_get_bits(&s->gb, buf, buf_size*8);
 303     skip_bits(&s->gb, s->bitindex);
 304     if (!s->blocksize)
 305     {
 306         int maxnlpc = 0;
 307         /* shorten signature */
 308         if (get_bits_long(&s->gb, 32) != bswap_32(ff_get_fourcc("ajkg"))) {
 309             av_log(s->avctx, AV_LOG_ERROR, "missing shorten magic 'ajkg'\n");
 310             return -1;
 311         }
 312
 313         s->lpcqoffset = 0;
 314         s->blocksize = DEFAULT_BLOCK_SIZE;
 315         s->channels = 1;
 316         s->nmean = -1;
 317         s->version = get_bits(&s->gb, 8);
 318         s->internal_ftype = get_uint(s, TYPESIZE);
 319
 320         s->channels = get_uint(s, CHANSIZE);
 321         if (s->channels > MAX_CHANNELS) {
 322             av_log(s->avctx, AV_LOG_ERROR, "too many channels: %d\n", s->channels);
 323             return -1;
 324         }
 325
 326         /* get blocksize if version > 0 */
 327         if (s->version > 0) {
 328             int skip_bytes;
 329             s->blocksize = get_uint(s, av_log2(DEFAULT_BLOCK_SIZE));
 330             maxnlpc = get_uint(s, LPCQSIZE);
 331             s->nmean = get_uint(s, 0);
 332
 333             skip_bytes = get_uint(s, NSKIPSIZE);
 334             for (i=0; i<skip_bytes; i++) {
 335                 skip_bits(&s->gb, 8);
 336             }
 337         }
 338         s->nwrap = FFMAX(NWRAP, maxnlpc);
 339
 340         if (allocate_buffers(s))
 341             return -1;
 342
 343         init_offset(s);
 344
 345         if (s->version > 1)
 346             s->lpcqoffset = V2LPCQOFFSET;
 347
 348         if (get_ur_golomb_shorten(&s->gb, FNSIZE) != FN_VERBATIM) {
 349             av_log(s->avctx, AV_LOG_ERROR, "missing verbatim section at beginning of stream\n");
 350             return -1;
 351         }
 352
 353         s->header_size = get_ur_golomb_shorten(&s->gb, VERBATIM_CKSIZE_SIZE);
 354         if (s->header_size >= OUT_BUFFER_SIZE || s->header_size < CANONICAL_HEADER_SIZE) {
 355             av_log(s->avctx, AV_LOG_ERROR, "header is wrong size: %d\n", s->header_size);
 356             return -1;
 357         }
 358
 359         for (i=0; i<s->header_size; i++)
 360             s->header[i] = (char)get_ur_golomb_shorten(&s->gb, VERBATIM_BYTE_SIZE);
 361
 362         if (decode_wave_header(avctx, s->header, s->header_size) < 0)
 363             return -1;
 364
 365         s->cur_chan = 0;
 366         s->bitshift = 0;
 367     }
 368     else
 369     {
 370         int cmd;
 371         int len;
 372         cmd = get_ur_golomb_shorten(&s->gb, FNSIZE);
 373         switch (cmd) {
 374             case FN_ZERO:
 375             case FN_DIFF0:
 376             case FN_DIFF1:
 377             case FN_DIFF2:
 378             case FN_DIFF3:
 379             case FN_QLPC:
 380                 {
 381                     int residual_size = 0;
 382                     int channel = s->cur_chan;
 383                     int32_t coffset;
 384                     if (cmd != FN_ZERO) {
 385                         residual_size = get_ur_golomb_shorten(&s->gb, ENERGYSIZE);
 386                         /* this is a hack as version 0 differed in defintion of get_sr_golomb_shorten */
 387                         if (s->version == 0)
 388                             residual_size--;
 389                     }
 390
 391                     if (s->nmean == 0)
 392                         coffset = s->offset[channel][0];
 393                     else {
 394                         int32_t sum = (s->version < 2) ? 0 : s->nmean / 2;
 395                         for (i=0; i<s->nmean; i++)
 396                             sum += s->offset[channel][i];
 397                         coffset = sum / s->nmean;
 398                         if (s->version >= 2)
 399                             coffset >>= FFMIN(1, s->bitshift);
 400                     }
 401                     switch (cmd) {
 402                         case FN_ZERO:
 403                             for (i=0; i<s->blocksize; i++)
 404                                 s->decoded[channel][i] = 0;
 405                             break;
 406                         case FN_DIFF0:
 407                             for (i=0; i<s->blocksize; i++)
 408                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + coffset;
 409                             break;
 410                         case FN_DIFF1:
 411                             for (i=0; i<s->blocksize; i++)
 412                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + s->decoded[channel][i - 1];
 413                             break;
 414                         case FN_DIFF2:
 415                             for (i=0; i<s->blocksize; i++)
 416                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + 2*s->decoded[channel][i-1]
 417                                                                                                       -   s->decoded[channel][i-2];
 418                             break;
 419                         case FN_DIFF3:
 420                             for (i=0; i<s->blocksize; i++)
 421                                 s->decoded[channel][i] = get_sr_golomb_shorten(&s->gb, residual_size) + 3*s->decoded[channel][i-1]
 422                                                                                                       - 3*s->decoded[channel][i-2]
 423                                                                                                       +   s->decoded[channel][i-3];
 424                             break;
 425                         case FN_QLPC:
 426                             {
 427                                 int pred_order = get_ur_golomb_shorten(&s->gb, LPCQSIZE);
 428                                 for (i=0; i<pred_order; i++)
 429                                     s->decoded[channel][i - pred_order] -= coffset;
 430                                 decode_subframe_lpc(s, channel, residual_size, pred_order);
 431                                 if (coffset != 0)
 432                                     for (i=0; i < s->blocksize; i++)
 433                                         s->decoded[channel][i] += coffset;
 434                             }
 435                     }
 436                     if (s->nmean > 0) {
 437                         int32_t sum = (s->version < 2) ? 0 : s->blocksize / 2;
 438                         for (i=0; i<s->blocksize; i++)
 439                             sum += s->decoded[channel][i];
 440
 441                         for (i=1; i<s->nmean; i++)
 442                             s->offset[channel][i-1] = s->offset[channel][i];
 443
 444                         if (s->version < 2)
 445                             s->offset[channel][s->nmean - 1] = sum / s->blocksize;
 446                         else
 447                             s->offset[channel][s->nmean - 1] = (sum / s->blocksize) << s->bitshift;
 448                     }
 449                     for (i=-s->nwrap; i<0; i++)
 450                         s->decoded[channel][i] = s->decoded[channel][i + s->blocksize];
 451
 452                     fix_bitshift(s, s->decoded[channel]);
 453
 454                     s->cur_chan++;
 455                     if (s->cur_chan == s->channels) {
 456                         samples = interleave_buffer(samples, s->channels, s->blocksize, s->decoded);
 457                         s->cur_chan = 0;
 458                         goto frame_done;
 459                     }
 460                     break;
 461                 }
 462                 break;
 463             case FN_VERBATIM:
 464                 len = get_ur_golomb_shorten(&s->gb, VERBATIM_CKSIZE_SIZE);
 465                 while (len--) {
 466                     get_ur_golomb_shorten(&s->gb, VERBATIM_BYTE_SIZE);
 467                 }
 468                 break;
 469             case FN_BITSHIFT:
 470                 s->bitshift = get_ur_golomb_shorten(&s->gb, BITSHIFTSIZE);
 471                 break;
 472             case FN_BLOCKSIZE:
 473                 s->blocksize = get_uint(s, av_log2(s->blocksize));
 474                 break;
 475             case FN_QUIT:
 476                 *data_size = 0;
 477                 return buf_size;
 478                 break;
 479             default:
 480                 av_log(avctx, AV_LOG_ERROR, "unknown shorten function %d\n", cmd);
 481                 return -1;
 482                 break;
 483         }
 484     }
 485 frame_done:
 486     *data_size = (int8_t *)samples - (int8_t *)data;
 487
 488     //    s->last_blocksize = s->blocksize;
 489     s->bitindex = get_bits_count(&s->gb) - 8*((get_bits_count(&s->gb))/8);
 490     i= (get_bits_count(&s->gb))/8;
 491     if (i > buf_size) {
 492         av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", i - buf_size);
 493         s->bitstream_size=0;
 494         s->bitstream_index=0;
 495         return -1;
 496     }
 497     if (s->bitstream_size) {
 498         s->bitstream_index += i;
 499         s->bitstream_size  -= i;
 500         return input_buf_size;
 501     } else
 502         return i;
 503 }
 504
 505 static av_cold int shorten_decode_close(AVCodecContext *avctx)
 506 {
 507     ShortenContext *s = avctx->priv_data;
 508     int i;
 509
 510     for (i = 0; i < s->channels; i++) {
 511         s->decoded[i] -= s->nwrap;
 512         av_freep(&s->decoded[i]);
 513         av_freep(&s->offset[i]);
 514     }
 515     av_freep(&s->bitstream);
 516     return 0;
 517 }
 518
 519 static void shorten_flush(AVCodecContext *avctx){
 520     ShortenContext *s = avctx->priv_data;
 521
 522     s->bitstream_size=
 523         s->bitstream_index= 0;
 524 }
 525
 526 AVCodec shorten_decoder = {
 527     "shorten",
 528     CODEC_TYPE_AUDIO,
 529     CODEC_ID_SHORTEN,
 530     sizeof(ShortenContext),
 531     shorten_decode_init,
 532     NULL,
 533     shorten_decode_close,
 534     shorten_decode_frame,
 535     .flush= shorten_flush,
 536     .long_name= NULL_IF_CONFIG_SMALL("Shorten"),
 537 };