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