libavcodec/ralf.c

   1 /*
   2  * RealAudio Lossless decoder
   3  *
   4  * Copyright (c) 2012 Konstantin Shishkov
   5  *
   6  * This file is part of Libav.
   7  *
   8  * Libav is free software; you can redistribute it and/or
   9  * modify it under the terms of the GNU Lesser General Public
  10  * License as published by the Free Software Foundation; either
  11  * version 2.1 of the License, or (at your option) any later version.
  12  *
  13  * Libav is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  16  * Lesser General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU Lesser General Public
  19  * License along with Libav; if not, write to the Free Software
  20  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  21  */
  22
  23 /**
  24  * @file
  25  * This is a decoder for Real Audio Lossless format.
  26  * Dedicated to the mastermind behind it, Ralph Wiggum.
  27  */
  28
  29 #include "libavutil/channel_layout.h"
  30 #include "avcodec.h"
  31 #include "get_bits.h"
  32 #include "golomb.h"
  33 #include "internal.h"
  34 #include "unary.h"
  35 #include "ralfdata.h"
  36
  37 #define FILTER_NONE 0
  38 #define FILTER_RAW  642
  39
  40 typedef struct VLCSet {
  41     VLC filter_params;
  42     VLC bias;
  43     VLC coding_mode;
  44     VLC filter_coeffs[10][11];
  45     VLC short_codes[15];
  46     VLC long_codes[125];
  47 } VLCSet;
  48
  49 #define RALF_MAX_PKT_SIZE 8192
  50
  51 typedef struct RALFContext {
  52     int version;
  53     int max_frame_size;
  54     VLCSet sets[3];
  55     int32_t channel_data[2][4096];
  56
  57     int     filter_params;   ///< combined filter parameters for the current channel data
  58     int     filter_length;   ///< length of the filter for the current channel data
  59     int     filter_bits;     ///< filter precision for the current channel data
  60     int32_t filter[64];
  61
  62     int     bias[2];         ///< a constant value added to channel data after filtering
  63
  64     int num_blocks;          ///< number of blocks inside the frame
  65     int sample_offset;
  66     int block_size[1 << 12]; ///< size of the blocks
  67     int block_pts[1 << 12];  ///< block start time (in milliseconds)
  68
  69     uint8_t pkt[16384];
  70     int     has_pkt;
  71 } RALFContext;
  72
  73 #define MAX_ELEMS 644 // no RALF table uses more than that
  74
  75 static int init_ralf_vlc(VLC *vlc, const uint8_t *data, int elems)
  76 {
  77     uint8_t  lens[MAX_ELEMS];
  78     uint16_t codes[MAX_ELEMS];
  79     int counts[17], prefixes[18];
  80     int i, cur_len;
  81     int max_bits = 0;
  82     int nb = 0;
  83
  84     for (i = 0; i <= 16; i++)
  85         counts[i] = 0;
  86     for (i = 0; i < elems; i++) {
  87         cur_len  = (nb ? *data & 0xF : *data >> 4) + 1;
  88         counts[cur_len]++;
  89         max_bits = FFMAX(max_bits, cur_len);
  90         lens[i]  = cur_len;
  91         data    += nb;
  92         nb      ^= 1;
  93     }
  94     prefixes[1] = 0;
  95     for (i = 1; i <= 16; i++)
  96         prefixes[i + 1] = (prefixes[i] + counts[i]) << 1;
  97
  98     for (i = 0; i < elems; i++)
  99         codes[i] = prefixes[lens[i]]++;
 100
 101     return ff_init_vlc_sparse(vlc, FFMIN(max_bits, 9), elems,
 102                               lens, 1, 1, codes, 2, 2, NULL, 0, 0, 0);
 103 }
 104
 105 static av_cold int decode_close(AVCodecContext *avctx)
 106 {
 107     RALFContext *ctx = avctx->priv_data;
 108     int i, j, k;
 109
 110     for (i = 0; i < 3; i++) {
 111         ff_free_vlc(&ctx->sets[i].filter_params);
 112         ff_free_vlc(&ctx->sets[i].bias);
 113         ff_free_vlc(&ctx->sets[i].coding_mode);
 114         for (j = 0; j < 10; j++)
 115             for (k = 0; k < 11; k++)
 116                 ff_free_vlc(&ctx->sets[i].filter_coeffs[j][k]);
 117         for (j = 0; j < 15; j++)
 118             ff_free_vlc(&ctx->sets[i].short_codes[j]);
 119         for (j = 0; j < 125; j++)
 120             ff_free_vlc(&ctx->sets[i].long_codes[j]);
 121     }
 122
 123     return 0;
 124 }
 125
 126 static av_cold int decode_init(AVCodecContext *avctx)
 127 {
 128     RALFContext *ctx = avctx->priv_data;
 129     int i, j, k;
 130     int ret;
 131
 132     if (avctx->extradata_size < 24 || memcmp(avctx->extradata, "LSD:", 4)) {
 133         av_log(avctx, AV_LOG_ERROR, "Extradata is not groovy, dude\n");
 134         return AVERROR_INVALIDDATA;
 135     }
 136
 137     ctx->version = AV_RB16(avctx->extradata + 4);
 138     if (ctx->version != 0x103) {
 139         av_log_ask_for_sample(avctx, "unknown version %X\n", ctx->version);
 140         return AVERROR_PATCHWELCOME;
 141     }
 142
 143     avctx->channels    = AV_RB16(avctx->extradata + 8);
 144     avctx->sample_rate = AV_RB32(avctx->extradata + 12);
 145     if (avctx->channels < 1 || avctx->channels > 2
 146         || avctx->sample_rate < 8000 || avctx->sample_rate > 96000) {
 147         av_log(avctx, AV_LOG_ERROR, "Invalid coding parameters %d Hz %d ch\n",
 148                avctx->sample_rate, avctx->channels);
 149         return AVERROR_INVALIDDATA;
 150     }
 151     avctx->sample_fmt     = AV_SAMPLE_FMT_S16P;
 152     avctx->channel_layout = (avctx->channels == 2) ? AV_CH_LAYOUT_STEREO
 153                                                    : AV_CH_LAYOUT_MONO;
 154
 155     ctx->max_frame_size = AV_RB32(avctx->extradata + 16);
 156     if (ctx->max_frame_size > (1 << 20) || !ctx->max_frame_size) {
 157         av_log(avctx, AV_LOG_ERROR, "invalid frame size %d\n",
 158                ctx->max_frame_size);
 159     }
 160     ctx->max_frame_size = FFMAX(ctx->max_frame_size, avctx->sample_rate);
 161
 162     for (i = 0; i < 3; i++) {
 163         ret = init_ralf_vlc(&ctx->sets[i].filter_params, filter_param_def[i],
 164                             FILTERPARAM_ELEMENTS);
 165         if (ret < 0) {
 166             decode_close(avctx);
 167             return ret;
 168         }
 169         ret = init_ralf_vlc(&ctx->sets[i].bias, bias_def[i], BIAS_ELEMENTS);
 170         if (ret < 0) {
 171             decode_close(avctx);
 172             return ret;
 173         }
 174         ret = init_ralf_vlc(&ctx->sets[i].coding_mode, coding_mode_def[i],
 175                             CODING_MODE_ELEMENTS);
 176         if (ret < 0) {
 177             decode_close(avctx);
 178             return ret;
 179         }
 180         for (j = 0; j < 10; j++) {
 181             for (k = 0; k < 11; k++) {
 182                 ret = init_ralf_vlc(&ctx->sets[i].filter_coeffs[j][k],
 183                                     filter_coeffs_def[i][j][k],
 184                                     FILTER_COEFFS_ELEMENTS);
 185                 if (ret < 0) {
 186                     decode_close(avctx);
 187                     return ret;
 188                 }
 189             }
 190         }
 191         for (j = 0; j < 15; j++) {
 192             ret = init_ralf_vlc(&ctx->sets[i].short_codes[j],
 193                                 short_codes_def[i][j], SHORT_CODES_ELEMENTS);
 194             if (ret < 0) {
 195                 decode_close(avctx);
 196                 return ret;
 197             }
 198         }
 199         for (j = 0; j < 125; j++) {
 200             ret = init_ralf_vlc(&ctx->sets[i].long_codes[j],
 201                                 long_codes_def[i][j], LONG_CODES_ELEMENTS);
 202             if (ret < 0) {
 203                 decode_close(avctx);
 204                 return ret;
 205             }
 206         }
 207     }
 208
 209     return 0;
 210 }
 211
 212 static inline int extend_code(GetBitContext *gb, int val, int range, int bits)
 213 {
 214     if (val == 0) {
 215         val = -range - get_ue_golomb(gb);
 216     } else if (val == range * 2) {
 217         val =  range + get_ue_golomb(gb);
 218     } else {
 219         val -= range;
 220     }
 221     if (bits)
 222         val = (val << bits) | get_bits(gb, bits);
 223     return val;
 224 }
 225
 226 static int decode_channel(RALFContext *ctx, GetBitContext *gb, int ch,
 227                           int length, int mode, int bits)
 228 {
 229     int i, t;
 230     int code_params;
 231     VLCSet *set = ctx->sets + mode;
 232     VLC *code_vlc; int range, range2, add_bits;
 233     int *dst = ctx->channel_data[ch];
 234
 235     ctx->filter_params = get_vlc2(gb, set->filter_params.table, 9, 2);
 236     ctx->filter_bits   = (ctx->filter_params - 2) >> 6;
 237     ctx->filter_length = ctx->filter_params - (ctx->filter_bits << 6) - 1;
 238
 239     if (ctx->filter_params == FILTER_RAW) {
 240         for (i = 0; i < length; i++)
 241             dst[i] = get_bits(gb, bits);
 242         ctx->bias[ch] = 0;
 243         return 0;
 244     }
 245
 246     ctx->bias[ch] = get_vlc2(gb, set->bias.table, 9, 2);
 247     ctx->bias[ch] = extend_code(gb, ctx->bias[ch], 127, 4);
 248
 249     if (ctx->filter_params == FILTER_NONE) {
 250         memset(dst, 0, sizeof(*dst) * length);
 251         return 0;
 252     }
 253
 254     if (ctx->filter_params > 1) {
 255         int cmode = 0, coeff = 0;
 256         VLC *vlc = set->filter_coeffs[ctx->filter_bits] + 5;
 257
 258         add_bits = ctx->filter_bits;
 259
 260         for (i = 0; i < ctx->filter_length; i++) {
 261             t = get_vlc2(gb, vlc[cmode].table, vlc[cmode].bits, 2);
 262             t = extend_code(gb, t, 21, add_bits);
 263             if (!cmode)
 264                 coeff -= 12 << add_bits;
 265             coeff = t - coeff;
 266             ctx->filter[i] = coeff;
 267
 268             cmode = coeff >> add_bits;
 269             if (cmode < 0) {
 270                 cmode = -1 - av_log2(-cmode);
 271                 if (cmode < -5)
 272                     cmode = -5;
 273             } else if (cmode > 0) {
 274                 cmode =  1 + av_log2(cmode);
 275                 if (cmode > 5)
 276                     cmode = 5;
 277             }
 278         }
 279     }
 280
 281     code_params = get_vlc2(gb, set->coding_mode.table, set->coding_mode.bits, 2);
 282     if (code_params >= 15) {
 283         add_bits = av_clip((code_params / 5 - 3) / 2, 0, 10);
 284         if (add_bits > 9 && (code_params % 5) != 2)
 285             add_bits--;
 286         range    = 10;
 287         range2   = 21;
 288         code_vlc = set->long_codes + code_params - 15;
 289     } else {
 290         add_bits = 0;
 291         range    = 6;
 292         range2   = 13;
 293         code_vlc = set->short_codes + code_params;
 294     }
 295
 296     for (i = 0; i < length; i += 2) {
 297         int code1, code2;
 298
 299         t = get_vlc2(gb, code_vlc->table, code_vlc->bits, 2);
 300         code1 = t / range2;
 301         code2 = t % range2;
 302         dst[i]     = extend_code(gb, code1, range, 0) << add_bits;
 303         dst[i + 1] = extend_code(gb, code2, range, 0) << add_bits;
 304         if (add_bits) {
 305             dst[i]     |= get_bits(gb, add_bits);
 306             dst[i + 1] |= get_bits(gb, add_bits);
 307         }
 308     }
 309
 310     return 0;
 311 }
 312
 313 static void apply_lpc(RALFContext *ctx, int ch, int length, int bits)
 314 {
 315     int i, j, acc;
 316     int *audio = ctx->channel_data[ch];
 317     int bias = 1 << (ctx->filter_bits - 1);
 318     int max_clip = (1 << bits) - 1, min_clip = -max_clip - 1;
 319
 320     for (i = 1; i < length; i++) {
 321         int flen = FFMIN(ctx->filter_length, i);
 322
 323         acc = 0;
 324         for (j = 0; j < flen; j++)
 325             acc += ctx->filter[j] * audio[i - j - 1];
 326         if (acc < 0) {
 327             acc = (acc + bias - 1) >> ctx->filter_bits;
 328             acc = FFMAX(acc, min_clip);
 329         } else {
 330             acc = (acc + bias) >> ctx->filter_bits;
 331             acc = FFMIN(acc, max_clip);
 332         }
 333         audio[i] += acc;
 334     }
 335 }
 336
 337 static int decode_block(AVCodecContext *avctx, GetBitContext *gb,
 338                         int16_t *dst0, int16_t *dst1)
 339 {
 340     RALFContext *ctx = avctx->priv_data;
 341     int len, ch, ret;
 342     int dmode, mode[2], bits[2];
 343     int *ch0, *ch1;
 344     int i, t, t2;
 345
 346     len = 12 - get_unary(gb, 0, 6);
 347
 348     if (len <= 7) len ^= 1; // codes for length = 6 and 7 are swapped
 349     len = 1 << len;
 350
 351     if (ctx->sample_offset + len > ctx->max_frame_size) {
 352         av_log(avctx, AV_LOG_ERROR,
 353                "Decoder's stomach is crying, it ate too many samples\n");
 354         return AVERROR_INVALIDDATA;
 355     }
 356
 357     if (avctx->channels > 1)
 358         dmode = get_bits(gb, 2) + 1;
 359     else
 360         dmode = 0;
 361
 362     mode[0] = (dmode == 4) ? 1 : 0;
 363     mode[1] = (dmode >= 2) ? 2 : 0;
 364     bits[0] = 16;
 365     bits[1] = (mode[1] == 2) ? 17 : 16;
 366
 367     for (ch = 0; ch < avctx->channels; ch++) {
 368         if ((ret = decode_channel(ctx, gb, ch, len, mode[ch], bits[ch])) < 0)
 369             return ret;
 370         if (ctx->filter_params > 1 && ctx->filter_params != FILTER_RAW) {
 371             ctx->filter_bits += 3;
 372             apply_lpc(ctx, ch, len, bits[ch]);
 373         }
 374         if (get_bits_left(gb) < 0)
 375             return AVERROR_INVALIDDATA;
 376     }
 377     ch0 = ctx->channel_data[0];
 378     ch1 = ctx->channel_data[1];
 379     switch (dmode) {
 380     case 0:
 381         for (i = 0; i < len; i++)
 382             dst0[i] = ch0[i] + ctx->bias[0];
 383         break;
 384     case 1:
 385         for (i = 0; i < len; i++) {
 386             dst0[i] = ch0[i] + ctx->bias[0];
 387             dst1[i] = ch1[i] + ctx->bias[1];
 388         }
 389         break;
 390     case 2:
 391         for (i = 0; i < len; i++) {
 392             ch0[i] += ctx->bias[0];
 393             dst0[i] = ch0[i];
 394             dst1[i] = ch0[i] - (ch1[i] + ctx->bias[1]);
 395         }
 396         break;
 397     case 3:
 398         for (i = 0; i < len; i++) {
 399             t  = ch0[i] + ctx->bias[0];
 400             t2 = ch1[i] + ctx->bias[1];
 401             dst0[i] = t + t2;
 402             dst1[i] = t;
 403         }
 404         break;
 405     case 4:
 406         for (i = 0; i < len; i++) {
 407             t  =   ch1[i] + ctx->bias[1];
 408             t2 = ((ch0[i] + ctx->bias[0]) << 1) | (t & 1);
 409             dst0[i] = (t2 + t) / 2;
 410             dst1[i] = (t2 - t) / 2;
 411         }
 412         break;
 413     }
 414
 415     ctx->sample_offset += len;
 416
 417     return 0;
 418 }
 419
 420 static int decode_frame(AVCodecContext *avctx, void *data, int *got_frame_ptr,
 421                         AVPacket *avpkt)
 422 {
 423     RALFContext *ctx = avctx->priv_data;
 424     AVFrame *frame   = data;
 425     int16_t *samples0;
 426     int16_t *samples1;
 427     int ret;
 428     GetBitContext gb;
 429     int table_size, table_bytes, i;
 430     const uint8_t *src, *block_pointer;
 431     int src_size;
 432     int bytes_left;
 433
 434     if (ctx->has_pkt) {
 435         ctx->has_pkt = 0;
 436         table_bytes = (AV_RB16(avpkt->data) + 7) >> 3;
 437         if (table_bytes + 3 > avpkt->size || avpkt->size > RALF_MAX_PKT_SIZE) {
 438             av_log(avctx, AV_LOG_ERROR, "Wrong packet's breath smells of wrong data!\n");
 439             return AVERROR_INVALIDDATA;
 440         }
 441         if (memcmp(ctx->pkt, avpkt->data, 2 + table_bytes)) {
 442             av_log(avctx, AV_LOG_ERROR, "Wrong packet tails are wrong!\n");
 443             return AVERROR_INVALIDDATA;
 444         }
 445
 446         src      = ctx->pkt;
 447         src_size = RALF_MAX_PKT_SIZE + avpkt->size;
 448         memcpy(ctx->pkt + RALF_MAX_PKT_SIZE, avpkt->data + 2 + table_bytes,
 449                avpkt->size - 2 - table_bytes);
 450     } else {
 451         if (avpkt->size == RALF_MAX_PKT_SIZE) {
 452             memcpy(ctx->pkt, avpkt->data, avpkt->size);
 453             ctx->has_pkt   = 1;
 454             *got_frame_ptr = 0;
 455
 456             return avpkt->size;
 457         }
 458         src      = avpkt->data;
 459         src_size = avpkt->size;
 460     }
 461
 462     frame->nb_samples = ctx->max_frame_size;
 463     if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
 464         av_log(avctx, AV_LOG_ERROR, "Me fail get_buffer()? That's unpossible!\n");
 465         return ret;
 466     }
 467     samples0 = (int16_t *)frame->data[0];
 468     samples1 = (int16_t *)frame->data[1];
 469
 470     if (src_size < 5) {
 471         av_log(avctx, AV_LOG_ERROR, "too short packets are too short!\n");
 472         return AVERROR_INVALIDDATA;
 473     }
 474     table_size  = AV_RB16(src);
 475     table_bytes = (table_size + 7) >> 3;
 476     if (src_size < table_bytes + 3) {
 477         av_log(avctx, AV_LOG_ERROR, "short packets are short!\n");
 478         return AVERROR_INVALIDDATA;
 479     }
 480     init_get_bits(&gb, src + 2, table_size);
 481     ctx->num_blocks = 0;
 482     while (get_bits_left(&gb) > 0) {
 483         ctx->block_size[ctx->num_blocks] = get_bits(&gb, 15);
 484         if (get_bits1(&gb)) {
 485             ctx->block_pts[ctx->num_blocks] = get_bits(&gb, 9);
 486         } else {
 487             ctx->block_pts[ctx->num_blocks] = 0;
 488         }
 489         ctx->num_blocks++;
 490     }
 491
 492     block_pointer = src      + table_bytes + 2;
 493     bytes_left    = src_size - table_bytes - 2;
 494     ctx->sample_offset = 0;
 495     for (i = 0; i < ctx->num_blocks; i++) {
 496         if (bytes_left < ctx->block_size[i]) {
 497             av_log(avctx, AV_LOG_ERROR, "I'm pedaling backwards\n");
 498             break;
 499         }
 500         init_get_bits(&gb, block_pointer, ctx->block_size[i] * 8);
 501         if (decode_block(avctx, &gb, samples0 + ctx->sample_offset,
 502                                      samples1 + ctx->sample_offset) < 0) {
 503             av_log(avctx, AV_LOG_ERROR, "Sir, I got carsick in your office. Not decoding the rest of packet.\n");
 504             break;
 505         }
 506         block_pointer += ctx->block_size[i];
 507         bytes_left    -= ctx->block_size[i];
 508     }
 509
 510     frame->nb_samples = ctx->sample_offset;
 511     *got_frame_ptr    = ctx->sample_offset > 0;
 512
 513     return avpkt->size;
 514 }
 515
 516 static void decode_flush(AVCodecContext *avctx)
 517 {
 518     RALFContext *ctx = avctx->priv_data;
 519
 520     ctx->has_pkt = 0;
 521 }
 522
 523
 524 AVCodec ff_ralf_decoder = {
 525     .name           = "ralf",
 526     .type           = AVMEDIA_TYPE_AUDIO,
 527     .id             = AV_CODEC_ID_RALF,
 528     .priv_data_size = sizeof(RALFContext),
 529     .init           = decode_init,
 530     .close          = decode_close,
 531     .decode         = decode_frame,
 532     .flush          = decode_flush,
 533     .capabilities   = CODEC_CAP_DR1,
 534     .long_name      = NULL_IF_CONFIG_SMALL("RealAudio Lossless"),
 535     .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16P,
 536                                                       AV_SAMPLE_FMT_NONE },
 537 };