binkaudio.c
Go to the documentation of this file.
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 
31 #include "avcodec.h"
32 #define BITSTREAM_READER_LE
33 #include "get_bits.h"
34 #include "dsputil.h"
35 #include "dct.h"
36 #include "rdft.h"
37 #include "fmtconvert.h"
38 #include "libavutil/intfloat.h"
39 
40 extern const uint16_t ff_wma_critical_freqs[25];
41 
42 static float quant_table[96];
43 
44 #define MAX_CHANNELS 2
45 #define BINK_BLOCK_MAX_SIZE (MAX_CHANNELS << 11)
46 
47 typedef struct {
52  int version_b;
53  int first;
54  int channels;
55  int frame_len;
58  int num_bands;
59  unsigned int *bands;
60  float root;
62  DECLARE_ALIGNED(16, int16_t, previous)[BINK_BLOCK_MAX_SIZE / 16];
63  DECLARE_ALIGNED(16, int16_t, current)[BINK_BLOCK_MAX_SIZE / 16];
64  float *coeffs_ptr[MAX_CHANNELS];
65  float *prev_ptr[MAX_CHANNELS];
66  uint8_t *packet_buffer;
67  union {
70  } trans;
72 
73 
75 {
76  BinkAudioContext *s = avctx->priv_data;
77  int sample_rate = avctx->sample_rate;
78  int sample_rate_half;
79  int i;
80  int frame_len_bits;
81 
82  dsputil_init(&s->dsp, avctx);
83  ff_fmt_convert_init(&s->fmt_conv, avctx);
84 
85  /* determine frame length */
86  if (avctx->sample_rate < 22050) {
87  frame_len_bits = 9;
88  } else if (avctx->sample_rate < 44100) {
89  frame_len_bits = 10;
90  } else {
91  frame_len_bits = 11;
92  }
93 
94  if (avctx->channels > MAX_CHANNELS) {
95  av_log(avctx, AV_LOG_ERROR, "too many channels: %d\n", avctx->channels);
96  return -1;
97  }
98 
99  s->version_b = avctx->extradata && avctx->extradata[3] == 'b';
100 
101  if (avctx->codec->id == CODEC_ID_BINKAUDIO_RDFT) {
102  // audio is already interleaved for the RDFT format variant
103  sample_rate *= avctx->channels;
104  s->channels = 1;
105  if (!s->version_b)
106  frame_len_bits += av_log2(avctx->channels);
107  } else {
108  s->channels = avctx->channels;
109  }
110 
111  s->frame_len = 1 << frame_len_bits;
112  s->overlap_len = s->frame_len / 16;
113  s->block_size = (s->frame_len - s->overlap_len) * s->channels;
114  sample_rate_half = (sample_rate + 1) / 2;
115  s->root = 2.0 / sqrt(s->frame_len);
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  avctx->sample_fmt = AV_SAMPLE_FMT_S16;
138 
139  for (i = 0; i < s->channels; i++) {
140  s->coeffs_ptr[i] = s->coeffs + i * s->frame_len;
141  s->prev_ptr[i] = s->coeffs_ptr[i] + s->frame_len - s->overlap_len;
142  }
143 
145  ff_rdft_init(&s->trans.rdft, frame_len_bits, DFT_C2R);
147  ff_dct_init(&s->trans.dct, frame_len_bits, DCT_III);
148  else
149  return -1;
150 
152  avctx->coded_frame = &s->frame;
153 
154  return 0;
155 }
156 
157 static float get_float(GetBitContext *gb)
158 {
159  int power = get_bits(gb, 5);
160  float f = ldexpf(get_bits_long(gb, 23), power - 23);
161  if (get_bits1(gb))
162  f = -f;
163  return f;
164 }
165 
166 static const uint8_t rle_length_tab[16] = {
167  2, 3, 4, 5, 6, 8, 9, 10, 11, 12, 13, 14, 15, 16, 32, 64
168 };
169 
170 #define GET_BITS_SAFE(out, nbits) do { \
171  if (get_bits_left(gb) < nbits) \
172  return AVERROR_INVALIDDATA; \
173  out = get_bits(gb, nbits); \
174 } while (0)
175 
181 static int decode_block(BinkAudioContext *s, int16_t *out, int use_dct)
182 {
183  int ch, i, j, k;
184  float q, quant[25];
185  int width, coeff;
186  GetBitContext *gb = &s->gb;
187 
188  if (use_dct)
189  skip_bits(gb, 2);
190 
191  for (ch = 0; ch < s->channels; ch++) {
192  FFTSample *coeffs = s->coeffs_ptr[ch];
193  if (s->version_b) {
194  if (get_bits_left(gb) < 64)
195  return AVERROR_INVALIDDATA;
196  coeffs[0] = av_int2float(get_bits_long(gb, 32)) * s->root;
197  coeffs[1] = av_int2float(get_bits_long(gb, 32)) * s->root;
198  } else {
199  if (get_bits_left(gb) < 58)
200  return AVERROR_INVALIDDATA;
201  coeffs[0] = get_float(gb) * s->root;
202  coeffs[1] = get_float(gb) * s->root;
203  }
204 
205  if (get_bits_left(gb) < s->num_bands * 8)
206  return AVERROR_INVALIDDATA;
207  for (i = 0; i < s->num_bands; i++) {
208  int value = get_bits(gb, 8);
209  quant[i] = quant_table[FFMIN(value, 95)];
210  }
211 
212  k = 0;
213  q = quant[0];
214 
215  // parse coefficients
216  i = 2;
217  while (i < s->frame_len) {
218  if (s->version_b) {
219  j = i + 16;
220  } else {
221  int v;
222  GET_BITS_SAFE(v, 1);
223  if (v) {
224  GET_BITS_SAFE(v, 4);
225  j = i + rle_length_tab[v] * 8;
226  } else {
227  j = i + 8;
228  }
229  }
230 
231  j = FFMIN(j, s->frame_len);
232 
233  GET_BITS_SAFE(width, 4);
234  if (width == 0) {
235  memset(coeffs + i, 0, (j - i) * sizeof(*coeffs));
236  i = j;
237  while (s->bands[k] < i)
238  q = quant[k++];
239  } else {
240  while (i < j) {
241  if (s->bands[k] == i)
242  q = quant[k++];
243  GET_BITS_SAFE(coeff, width);
244  if (coeff) {
245  int v;
246  GET_BITS_SAFE(v, 1);
247  if (v)
248  coeffs[i] = -q * coeff;
249  else
250  coeffs[i] = q * coeff;
251  } else {
252  coeffs[i] = 0.0f;
253  }
254  i++;
255  }
256  }
257  }
258 
259  if (CONFIG_BINKAUDIO_DCT_DECODER && use_dct) {
260  coeffs[0] /= 0.5;
261  s->trans.dct.dct_calc(&s->trans.dct, coeffs);
262  s->dsp.vector_fmul_scalar(coeffs, coeffs, s->frame_len / 2, s->frame_len);
263  }
265  s->trans.rdft.rdft_calc(&s->trans.rdft, coeffs);
266  }
267 
269  (const float **)s->prev_ptr,
270  s->overlap_len, s->channels);
271  s->fmt_conv.float_to_int16_interleave(out, (const float **)s->coeffs_ptr,
272  s->frame_len - s->overlap_len,
273  s->channels);
274 
275  if (!s->first) {
276  int count = s->overlap_len * s->channels;
277  int shift = av_log2(count);
278  for (i = 0; i < count; i++) {
279  out[i] = (s->previous[i] * (count - i) + out[i] * i) >> shift;
280  }
281  }
282 
283  memcpy(s->previous, s->current,
284  s->overlap_len * s->channels * sizeof(*s->previous));
285 
286  s->first = 0;
287 
288  return 0;
289 }
290 
292 {
293  BinkAudioContext * s = avctx->priv_data;
294  av_freep(&s->bands);
295  av_freep(&s->packet_buffer);
297  ff_rdft_end(&s->trans.rdft);
299  ff_dct_end(&s->trans.dct);
300 
301  return 0;
302 }
303 
305 {
306  int n = (-get_bits_count(s)) & 31;
307  if (n) skip_bits(s, n);
308 }
309 
310 static int decode_frame(AVCodecContext *avctx, void *data,
311  int *got_frame_ptr, AVPacket *avpkt)
312 {
313  BinkAudioContext *s = avctx->priv_data;
314  int16_t *samples;
315  GetBitContext *gb = &s->gb;
316  int ret, consumed = 0;
317 
318  if (!get_bits_left(gb)) {
319  uint8_t *buf;
320  /* handle end-of-stream */
321  if (!avpkt->size) {
322  *got_frame_ptr = 0;
323  return 0;
324  }
325  if (avpkt->size < 4) {
326  av_log(avctx, AV_LOG_ERROR, "Packet is too small\n");
327  return AVERROR_INVALIDDATA;
328  }
330  if (!buf)
331  return AVERROR(ENOMEM);
332  s->packet_buffer = buf;
333  memcpy(s->packet_buffer, avpkt->data, avpkt->size);
334  init_get_bits(gb, s->packet_buffer, avpkt->size * 8);
335  consumed = avpkt->size;
336 
337  /* skip reported size */
338  skip_bits_long(gb, 32);
339  }
340 
341  /* get output buffer */
342  s->frame.nb_samples = s->block_size / avctx->channels;
343  if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
344  av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
345  return ret;
346  }
347  samples = (int16_t *)s->frame.data[0];
348 
349  if (decode_block(s, samples, avctx->codec->id == CODEC_ID_BINKAUDIO_DCT)) {
350  av_log(avctx, AV_LOG_ERROR, "Incomplete packet\n");
351  return AVERROR_INVALIDDATA;
352  }
353  get_bits_align32(gb);
354 
355  *got_frame_ptr = 1;
356  *(AVFrame *)data = s->frame;
357 
358  return consumed;
359 }
360 
362  .name = "binkaudio_rdft",
363  .type = AVMEDIA_TYPE_AUDIO,
365  .priv_data_size = sizeof(BinkAudioContext),
366  .init = decode_init,
367  .close = decode_end,
368  .decode = decode_frame,
369  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_DR1,
370  .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (RDFT)")
371 };
372 
374  .name = "binkaudio_dct",
375  .type = AVMEDIA_TYPE_AUDIO,
377  .priv_data_size = sizeof(BinkAudioContext),
378  .init = decode_init,
379  .close = decode_end,
380  .decode = decode_frame,
381  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_DR1,
382  .long_name = NULL_IF_CONFIG_SMALL("Bink Audio (DCT)")
383 };