libavcodec/binkaudio.c

   1 /*
   2  * Bink Audio decoder
   3  * Copyright (c) 2007-2011 Peter Ross (pross@xvid.org)
   4  * Copyright (c) 2009 Daniel Verkamp (daniel@drv.nu)
   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  * Bink Audio decoder
  26  *
  27  * Technical details here:
  28  *  http://wiki.multimedia.cx/index.php?title=Bink_Audio
  29  */
  30
  31 #include "libavutil/channel_layout.h"
  32 #include "avcodec.h"
  33 #define BITSTREAM_READER_LE
  34 #include "get_bits.h"
  35 #include "dsputil.h"
  36 #include "dct.h"
  37 #include "rdft.h"
  38 #include "fmtconvert.h"
  39 #include "internal.h"
  40 #include "libavutil/intfloat.h"
  41
  42 extern const uint16_t ff_wma_critical_freqs[25];
  43
  44 static float quant_table[96];
  45
  46 #define MAX_CHANNELS 2
  47 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
  48
  49 typedef struct {
  50     AVFrame frame;
  51     GetBitContext gb;
  52     int version_b;          ///< Bink version 'b'
  53     int first;
  54     int channels;
  55     int frame_len;          ///< transform size (samples)
  56     int overlap_len;        ///< overlap size (samples)
  57     int block_size;
  58     int num_bands;
  59     unsigned int *bands;
  60     float root;
  61     DECLARE_ALIGNED(32, FFTSample, coeffs)[BINK_BLOCK_MAX_SIZE];
  62     float previous[MAX_CHANNELS][BINK_BLOCK_MAX_SIZE / 16];  ///< coeffs from previous audio block
  63     uint8_t *packet_buffer;
  64     union {
  65         RDFTContext rdft;
  66         DCTContext dct;
  67     } trans;
  68 } BinkAudioContext;
  69
  70
  71 static av_cold int decode_init(AVCodecContext *avctx)
  72 {
  73     BinkAudioContext *s = avctx->priv_data;
  74     int sample_rate = avctx->sample_rate;
  75     int sample_rate_half;
  76     int i;
  77     int frame_len_bits;
  78
  79     /* determine frame length */
  80     if (avctx->sample_rate < 22050) {
  81         frame_len_bits = 9;
  82     } else if (avctx->sample_rate < 44100) {
  83         frame_len_bits = 10;
  84     } else {
  85         frame_len_bits = 11;
  86     }
  87
  88     if (avctx->channels > MAX_CHANNELS) {
  89         av_log(avctx, AV_LOG_ERROR, "too many channels: %d\n", avctx->channels);
  90         return -1;
  91     }
  92     avctx->channel_layout = avctx->channels == 1 ? AV_CH_LAYOUT_MONO :
  93                                                    AV_CH_LAYOUT_STEREO;
  94
  95     s->version_b = avctx->extradata && avctx->extradata[3] == 'b';
  96
  97     if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT) {
  98         // audio is already interleaved for the RDFT format variant
  99         avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
 100         sample_rate  *= avctx->channels;
 101         s->channels = 1;
 102         if (!s->version_b)
 103             frame_len_bits += av_log2(avctx->channels);
 104     } else {
 105         s->channels = avctx->channels;
 106         avctx->sample_fmt = AV_SAMPLE_FMT_FLTP;
 107     }
 108
 109     s->frame_len     = 1 << frame_len_bits;
 110     s->overlap_len   = s->frame_len / 16;
 111     s->block_size    = (s->frame_len - s->overlap_len) * s->channels;
 112     sample_rate_half = (sample_rate + 1) / 2;
 113     if (avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
 114         s->root = 2.0 / (sqrt(s->frame_len) * 32768.0);
 115     else
 116         s->root = s->frame_len / (sqrt(s->frame_len) * 32768.0);
 117     for (i = 0; i < 96; i++) {
 118         /* constant is result of 0.066399999/log10(M_E) */
 119         quant_table[i] = expf(i * 0.15289164787221953823f) * s->root;
 120     }
 121
 122     /* calculate number of bands */
 123     for (s->num_bands = 1; s->num_bands < 25; s->num_bands++)
 124         if (sample_rate_half <= ff_wma_critical_freqs[s->num_bands - 1])
 125             break;
 126
 127     s->bands = av_malloc((s->num_bands + 1) * sizeof(*s->bands));
 128     if (!s->bands)
 129         return AVERROR(ENOMEM);
 130
 131     /* populate bands data */
 132     s->bands[0] = 2;
 133     for (i = 1; i < s->num_bands; i++)
 134         s->bands[i] = (ff_wma_critical_freqs[i - 1] * s->frame_len / sample_rate_half) & ~1;
 135     s->bands[s->num_bands] = s->frame_len;
 136
 137     s->first = 1;
 138
 139     if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
 140         ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
 141     else if (CONFIG_BINKAUDIO_DCT_DECODER)
 142         ff_dct_init(&s->trans.dct, frame_len_bits, DCT_III);
 143     else
 144         return -1;
 145
 146     avcodec_get_frame_defaults(&s->frame);
 147     avctx->coded_frame = &s->frame;
 148
 149     return 0;
 150 }
 151
 152 static float get_float(GetBitContext *gb)
 153 {
 154     int power = get_bits(gb, 5);
 155     float f = ldexpf(get_bits_long(gb, 23), power - 23);
 156     if (get_bits1(gb))
 157         f = -f;
 158     return f;
 159 }
 160
 161 static const uint8_t rle_length_tab[16] = {
 162     2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
 163 };
 164
 165 /**
 166  * Decode Bink Audio block
 167  * @param[out] out Output buffer (must contain s->block_size elements)
 168  * @return 0 on success, negative error code on failure
 169  */
 170 static int decode_block(BinkAudioContext *s, float **out, int use_dct)
 171 {
 172     int ch, i, j, k;
 173     float q, quant[25];
 174     int width, coeff;
 175     GetBitContext *gb = &s->gb;
 176
 177     if (use_dct)
 178         skip_bits(gb, 2);
 179
 180     for (ch = 0; ch < s->channels; ch++) {
 181         FFTSample *coeffs = out[ch];
 182
 183         if (s->version_b) {
 184             if (get_bits_left(gb) < 64)
 185                 return AVERROR_INVALIDDATA;
 186             coeffs[0] = av_int2float(get_bits_long(gb, 32)) * s->root;
 187             coeffs[1] = av_int2float(get_bits_long(gb, 32)) * s->root;
 188         } else {
 189             if (get_bits_left(gb) < 58)
 190                 return AVERROR_INVALIDDATA;
 191             coeffs[0] = get_float(gb) * s->root;
 192             coeffs[1] = get_float(gb) * s->root;
 193         }
 194
 195         if (get_bits_left(gb) < s->num_bands * 8)
 196             return AVERROR_INVALIDDATA;
 197         for (i = 0; i < s->num_bands; i++) {
 198             int value = get_bits(gb, 8);
 199             quant[i]  = quant_table[FFMIN(value, 95)];
 200         }
 201
 202         k = 0;
 203         q = quant[0];
 204
 205         // parse coefficients
 206         i = 2;
 207         while (i < s->frame_len) {
 208             if (s->version_b) {
 209                 j = i + 16;
 210             } else {
 211                 int v = get_bits1(gb);
 212                 if (v) {
 213                     v = get_bits(gb, 4);
 214                     j = i + rle_length_tab[v] * 8;
 215                 } else {
 216                     j = i + 8;
 217                 }
 218             }
 219
 220             j = FFMIN(j, s->frame_len);
 221
 222             width = get_bits(gb, 4);
 223             if (width == 0) {
 224                 memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
 225                 i = j;
 226                 while (s->bands[k] < i)
 227                     q = quant[k++];
 228             } else {
 229                 while (i < j) {
 230                     if (s->bands[k] == i)
 231                         q = quant[k++];
 232                     coeff = get_bits(gb, width);
 233                     if (coeff) {
 234                         int v;
 235                         v = get_bits1(gb);
 236                         if (v)
 237                             coeffs[i] = -q * coeff;
 238                         else
 239                             coeffs[i] =  q * coeff;
 240                     } else {
 241                         coeffs[i] = 0.0f;
 242                     }
 243                     i++;
 244                 }
 245             }
 246         }
 247
 248         if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
 249             coeffs[0] /= 0.5;
 250             s->trans.dct.dct_calc(&s->trans.dct,  coeffs);
 251         }
 252         else if (CONFIG_BINKAUDIO_RDFT_DECODER)
 253             s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
 254     }
 255
 256     for (ch = 0; ch < s->channels; ch++) {
 257         int j;
 258         int count = s->overlap_len * s->channels;
 259         if (!s->first) {
 260             j = ch;
 261             for (i = 0; i < s->overlap_len; i++, j += s->channels)
 262                 out[ch][i] = (s->previous[ch][i] * (count - j) +
 263                                       out[ch][i] *          j) / count;
 264         }
 265         memcpy(s->previous[ch], &out[ch][s->frame_len - s->overlap_len],
 266                s->overlap_len * sizeof(*s->previous[ch]));
 267     }
 268
 269     s->first = 0;
 270
 271     return 0;
 272 }
 273
 274 static av_cold int decode_end(AVCodecContext *avctx)
 275 {
 276     BinkAudioContext * s = avctx->priv_data;
 277     av_freep(&s->bands);
 278     av_freep(&s->packet_buffer);
 279     if (CONFIG_BINKAUDIO_RDFT_DECODER && avctx->codec->id == AV_CODEC_ID_BINKAUDIO_RDFT)
 280         ff_rdft_end(&s->trans.rdft);
 281     else if (CONFIG_BINKAUDIO_DCT_DECODER)
 282         ff_dct_end(&s->trans.dct);
 283
 284     return 0;
 285 }
 286
 287 static void get_bits_align32(GetBitContext *s)
 288 {
 289     int n = (-get_bits_count(s)) & 31;
 290     if (n) skip_bits(s, n);
 291 }
 292
 293 static int decode_frame(AVCodecContext *avctx, void *data,
 294                         int *got_frame_ptr, AVPacket *avpkt)
 295 {
 296     BinkAudioContext *s = avctx->priv_data;
 297     GetBitContext *gb = &s->gb;
 298     int ret, consumed = 0;
 299
 300     if (!get_bits_left(gb)) {
 301         uint8_t *buf;
 302         /* handle end-of-stream */
 303         if (!avpkt->size) {
 304             *got_frame_ptr = 0;
 305             return 0;
 306         }
 307         if (avpkt->size < 4) {
 308             av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
 309             return AVERROR_INVALIDDATA;
 310         }
 311         buf = av_realloc(s->packet_buffer, avpkt->size + FF_INPUT_BUFFER_PADDING_SIZE);
 312         if (!buf)
 313             return AVERROR(ENOMEM);
 314         s->packet_buffer = buf;
 315         memcpy(s->packet_buffer, avpkt->data, avpkt->size);
 316         init_get_bits(gb, s->packet_buffer, avpkt->size * 8);
 317         consumed = avpkt->size;
 318
 319         /* skip reported size */
 320         skip_bits_long(gb, 32);
 321     }
 322
 323     /* get output buffer */
 324     s->frame.nb_samples = s->frame_len;
 325     if ((ret = ff_get_buffer(avctx, &s->frame)) < 0) {
 326         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
 327         return ret;
 328     }
 329
 330     if (decode_block(s, (float **)s->frame.extended_data,
 331                      avctx->codec->id == AV_CODEC_ID_BINKAUDIO_DCT)) {
 332         av_log(avctx, AV_LOG_ERROR, "Incomplete packet\n");
 333         return AVERROR_INVALIDDATA;
 334     }
 335     get_bits_align32(gb);
 336
 337     s->frame.nb_samples = s->block_size / avctx->channels;
 338     *got_frame_ptr   = 1;
 339     *(AVFrame *)data = s->frame;
 340
 341     return consumed;
 342 }
 343
 344 AVCodec ff_binkaudio_rdft_decoder = {
 345     .name           = "binkaudio_rdft",
 346     .type           = AVMEDIA_TYPE_AUDIO,
 347     .id             = AV_CODEC_ID_BINKAUDIO_RDFT,
 348     .priv_data_size = sizeof(BinkAudioContext),
 349     .init           = decode_init,
 350     .close          = decode_end,
 351     .decode         = decode_frame,
 352     .capabilities   = CODEC_CAP_DELAY | CODEC_CAP_DR1,
 353     .long_name      = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)")
 354 };
 355
 356 AVCodec ff_binkaudio_dct_decoder = {
 357     .name           = "binkaudio_dct",
 358     .type           = AVMEDIA_TYPE_AUDIO,
 359     .id             = AV_CODEC_ID_BINKAUDIO_DCT,
 360     .priv_data_size = sizeof(BinkAudioContext),
 361     .init           = decode_init,
 362     .close          = decode_end,
 363     .decode         = decode_frame,
 364     .capabilities   = CODEC_CAP_DELAY | CODEC_CAP_DR1,
 365     .long_name      = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)")
 366 };