FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
wmaenc.c
Go to the documentation of this file.
1 /*
2  * WMA compatible encoder
3  * Copyright (c) 2007 Michael Niedermayer
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 #include "avcodec.h"
23 #include "internal.h"
24 #include "wma.h"
25 #include "libavutil/avassert.h"
26 
27 
28 static int encode_init(AVCodecContext * avctx){
29  WMACodecContext *s = avctx->priv_data;
30  int i, flags1, flags2, block_align;
31  uint8_t *extradata;
32 
33  s->avctx = avctx;
34 
35  if(avctx->channels > MAX_CHANNELS) {
36  av_log(avctx, AV_LOG_ERROR, "too many channels: got %i, need %i or fewer\n",
37  avctx->channels, MAX_CHANNELS);
38  return AVERROR(EINVAL);
39  }
40 
41  if (avctx->sample_rate > 48000) {
42  av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz\n",
43  avctx->sample_rate);
44  return AVERROR(EINVAL);
45  }
46 
47  if(avctx->bit_rate < 24*1000) {
48  av_log(avctx, AV_LOG_ERROR, "bitrate too low: got %i, need 24000 or higher\n",
49  avctx->bit_rate);
50  return AVERROR(EINVAL);
51  }
52 
53 #if FF_API_OLD_ENCODE_AUDIO
54  if (!(avctx->coded_frame = avcodec_alloc_frame()))
55  return AVERROR(ENOMEM);
56 #endif
57 
58  /* extract flag infos */
59  flags1 = 0;
60  flags2 = 1;
61  if (avctx->codec->id == AV_CODEC_ID_WMAV1) {
62  extradata= av_malloc(4);
63  avctx->extradata_size= 4;
64  AV_WL16(extradata, flags1);
65  AV_WL16(extradata+2, flags2);
66  } else if (avctx->codec->id == AV_CODEC_ID_WMAV2) {
67  extradata= av_mallocz(10);
68  avctx->extradata_size= 10;
69  AV_WL32(extradata, flags1);
70  AV_WL16(extradata+4, flags2);
71  }else
72  av_assert0(0);
73  avctx->extradata= extradata;
74  s->use_exp_vlc = flags2 & 0x0001;
75  s->use_bit_reservoir = flags2 & 0x0002;
76  s->use_variable_block_len = flags2 & 0x0004;
77  if (avctx->channels == 2)
78  s->ms_stereo = 1;
79 
80  ff_wma_init(avctx, flags2);
81 
82  /* init MDCT */
83  for(i = 0; i < s->nb_block_sizes; i++)
84  ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0);
85 
86  block_align = avctx->bit_rate * (int64_t)s->frame_len /
87  (avctx->sample_rate * 8);
88  block_align = FFMIN(block_align, MAX_CODED_SUPERFRAME_SIZE);
89  avctx->block_align = block_align;
90 
91  avctx->frame_size = avctx->delay = s->frame_len;
92 
93  return 0;
94 }
95 
96 
97 static void apply_window_and_mdct(AVCodecContext * avctx, const AVFrame *frame)
98 {
99  WMACodecContext *s = avctx->priv_data;
100  float **audio = (float **)frame->extended_data;
101  int len = frame->nb_samples;
102  int window_index= s->frame_len_bits - s->block_len_bits;
103  FFTContext *mdct = &s->mdct_ctx[window_index];
104  int ch;
105  const float * win = s->windows[window_index];
106  int window_len = 1 << s->block_len_bits;
107  float n = 2.0 * 32768.0 / window_len;
108 
109  for (ch = 0; ch < avctx->channels; ch++) {
110  memcpy(s->output, s->frame_out[ch], window_len * sizeof(*s->output));
111  s->fdsp.vector_fmul_scalar(s->frame_out[ch], audio[ch], n, len);
112  s->fdsp.vector_fmul_reverse(&s->output[window_len], s->frame_out[ch], win, len);
113  s->fdsp.vector_fmul(s->frame_out[ch], s->frame_out[ch], win, len);
114  mdct->mdct_calc(mdct, s->coefs[ch], s->output);
115  }
116 }
117 
118 //FIXME use for decoding too
119 static void init_exp(WMACodecContext *s, int ch, const int *exp_param){
120  int n;
121  const uint16_t *ptr;
122  float v, *q, max_scale, *q_end;
123 
124  ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
125  q = s->exponents[ch];
126  q_end = q + s->block_len;
127  max_scale = 0;
128  while (q < q_end) {
129  /* XXX: use a table */
130  v = pow(10, *exp_param++ * (1.0 / 16.0));
131  max_scale= FFMAX(max_scale, v);
132  n = *ptr++;
133  do {
134  *q++ = v;
135  } while (--n);
136  }
137  s->max_exponent[ch] = max_scale;
138 }
139 
140 static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param){
141  int last_exp;
142  const uint16_t *ptr;
143  float *q, *q_end;
144 
145  ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
146  q = s->exponents[ch];
147  q_end = q + s->block_len;
148  if (s->version == 1) {
149  last_exp= *exp_param++;
150  av_assert0(last_exp-10 >= 0 && last_exp-10 < 32);
151  put_bits(&s->pb, 5, last_exp - 10);
152  q+= *ptr++;
153  }else
154  last_exp = 36;
155  while (q < q_end) {
156  int exp = *exp_param++;
157  int code = exp - last_exp + 60;
158  av_assert1(code >= 0 && code < 120);
160  /* XXX: use a table */
161  q+= *ptr++;
162  last_exp= exp;
163  }
164 }
165 
166 static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain){
167  int v, bsize, ch, coef_nb_bits, parse_exponents;
168  float mdct_norm;
169  int nb_coefs[MAX_CHANNELS];
170  static const int fixed_exp[25]={20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20,20};
171 
172  //FIXME remove duplication relative to decoder
173  if (s->use_variable_block_len) {
174  av_assert0(0); //FIXME not implemented
175  }else{
176  /* fixed block len */
180  }
181 
182  s->block_len = 1 << s->block_len_bits;
183 // assert((s->block_pos + s->block_len) <= s->frame_len);
184  bsize = s->frame_len_bits - s->block_len_bits;
185 
186  //FIXME factor
187  v = s->coefs_end[bsize] - s->coefs_start;
188  for (ch = 0; ch < s->avctx->channels; ch++)
189  nb_coefs[ch] = v;
190  {
191  int n4 = s->block_len / 2;
192  mdct_norm = 1.0 / (float)n4;
193  if (s->version == 1) {
194  mdct_norm *= sqrt(n4);
195  }
196  }
197 
198  if (s->avctx->channels == 2) {
199  put_bits(&s->pb, 1, !!s->ms_stereo);
200  }
201 
202  for (ch = 0; ch < s->avctx->channels; ch++) {
203  s->channel_coded[ch] = 1; //FIXME only set channel_coded when needed, instead of always
204  if (s->channel_coded[ch]) {
205  init_exp(s, ch, fixed_exp);
206  }
207  }
208 
209  for (ch = 0; ch < s->avctx->channels; ch++) {
210  if (s->channel_coded[ch]) {
211  WMACoef *coefs1;
212  float *coefs, *exponents, mult;
213  int i, n;
214 
215  coefs1 = s->coefs1[ch];
216  exponents = s->exponents[ch];
217  mult = pow(10, total_gain * 0.05) / s->max_exponent[ch];
218  mult *= mdct_norm;
219  coefs = src_coefs[ch];
220  if (s->use_noise_coding && 0) {
221  av_assert0(0); //FIXME not implemented
222  } else {
223  coefs += s->coefs_start;
224  n = nb_coefs[ch];
225  for(i = 0;i < n; i++){
226  double t= *coefs++ / (exponents[i] * mult);
227  if(t<-32768 || t>32767)
228  return -1;
229 
230  coefs1[i] = lrint(t);
231  }
232  }
233  }
234  }
235 
236  v = 0;
237  for (ch = 0; ch < s->avctx->channels; ch++) {
238  int a = s->channel_coded[ch];
239  put_bits(&s->pb, 1, a);
240  v |= a;
241  }
242 
243  if (!v)
244  return 1;
245 
246  for(v= total_gain-1; v>=127; v-= 127)
247  put_bits(&s->pb, 7, 127);
248  put_bits(&s->pb, 7, v);
249 
250  coef_nb_bits= ff_wma_total_gain_to_bits(total_gain);
251 
252  if (s->use_noise_coding) {
253  for (ch = 0; ch < s->avctx->channels; ch++) {
254  if (s->channel_coded[ch]) {
255  int i, n;
256  n = s->exponent_high_sizes[bsize];
257  for(i=0;i<n;i++) {
258  put_bits(&s->pb, 1, s->high_band_coded[ch][i]= 0);
259  if (0)
260  nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
261  }
262  }
263  }
264  }
265 
266  parse_exponents = 1;
267  if (s->block_len_bits != s->frame_len_bits) {
268  put_bits(&s->pb, 1, parse_exponents);
269  }
270 
271  if (parse_exponents) {
272  for (ch = 0; ch < s->avctx->channels; ch++) {
273  if (s->channel_coded[ch]) {
274  if (s->use_exp_vlc) {
275  encode_exp_vlc(s, ch, fixed_exp);
276  } else {
277  av_assert0(0); //FIXME not implemented
278 // encode_exp_lsp(s, ch);
279  }
280  }
281  }
282  } else {
283  av_assert0(0); //FIXME not implemented
284  }
285 
286  for (ch = 0; ch < s->avctx->channels; ch++) {
287  if (s->channel_coded[ch]) {
288  int run, tindex;
289  WMACoef *ptr, *eptr;
290  tindex = (ch == 1 && s->ms_stereo);
291  ptr = &s->coefs1[ch][0];
292  eptr = ptr + nb_coefs[ch];
293 
294  run=0;
295  for(;ptr < eptr; ptr++){
296  if(*ptr){
297  int level= *ptr;
298  int abs_level= FFABS(level);
299  int code= 0;
300  if(abs_level <= s->coef_vlcs[tindex]->max_level){
301  if(run < s->coef_vlcs[tindex]->levels[abs_level-1])
302  code= run + s->int_table[tindex][abs_level-1];
303  }
304 
305  av_assert2(code < s->coef_vlcs[tindex]->n);
306  put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]);
307 
308  if(code == 0){
309  if(1<<coef_nb_bits <= abs_level)
310  return -1;
311 
312  put_bits(&s->pb, coef_nb_bits, abs_level);
313  put_bits(&s->pb, s->frame_len_bits, run);
314  }
315  put_bits(&s->pb, 1, level < 0); //FIXME the sign is fliped somewhere
316  run=0;
317  }else{
318  run++;
319  }
320  }
321  if(run)
322  put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]);
323  }
324  if (s->version == 1 && s->avctx->channels >= 2) {
326  }
327  }
328  return 0;
329 }
330 
331 static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], uint8_t *buf, int buf_size, int total_gain){
332  init_put_bits(&s->pb, buf, buf_size);
333 
334  if (s->use_bit_reservoir) {
335  av_assert0(0);//FIXME not implemented
336  }else{
337  if(encode_block(s, src_coefs, total_gain) < 0)
338  return INT_MAX;
339  }
340 
342 
343  return put_bits_count(&s->pb) / 8 - s->avctx->block_align;
344 }
345 
346 static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt,
347  const AVFrame *frame, int *got_packet_ptr)
348 {
349  WMACodecContext *s = avctx->priv_data;
350  int i, total_gain, ret, error;
351 
352  s->block_len_bits= s->frame_len_bits; //required by non variable block len
353  s->block_len = 1 << s->block_len_bits;
354 
355  apply_window_and_mdct(avctx, frame);
356 
357  if (s->ms_stereo) {
358  float a, b;
359  int i;
360 
361  for(i = 0; i < s->block_len; i++) {
362  a = s->coefs[0][i]*0.5;
363  b = s->coefs[1][i]*0.5;
364  s->coefs[0][i] = a + b;
365  s->coefs[1][i] = a - b;
366  }
367  }
368 
369  if ((ret = ff_alloc_packet2(avctx, avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE)) < 0)
370  return ret;
371 
372  total_gain= 128;
373  for(i=64; i; i>>=1){
374  error = encode_frame(s, s->coefs, avpkt->data, avpkt->size,
375  total_gain - i);
376  if(error<=0)
377  total_gain-= i;
378  }
379 
380  while(total_gain <= 128 && error > 0)
381  error = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain++);
382  av_assert0((put_bits_count(&s->pb) & 7) == 0);
383  i= avctx->block_align - (put_bits_count(&s->pb)+7)/8;
384  av_assert0(i>=0);
385  while(i--)
386  put_bits(&s->pb, 8, 'N');
387 
388  flush_put_bits(&s->pb);
389  av_assert0(put_bits_ptr(&s->pb) - s->pb.buf == avctx->block_align);
390 
391  if (frame->pts != AV_NOPTS_VALUE)
392  avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
393 
394  avpkt->size = avctx->block_align;
395  *got_packet_ptr = 1;
396  return 0;
397 }
398 
399 #if CONFIG_WMAV1_ENCODER
400 AVCodec ff_wmav1_encoder = {
401  .name = "wmav1",
402  .type = AVMEDIA_TYPE_AUDIO,
403  .id = AV_CODEC_ID_WMAV1,
404  .priv_data_size = sizeof(WMACodecContext),
405  .init = encode_init,
406  .encode2 = encode_superframe,
407  .close = ff_wma_end,
408  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
410  .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"),
411 };
412 #endif
413 #if CONFIG_WMAV2_ENCODER
414 AVCodec ff_wmav2_encoder = {
415  .name = "wmav2",
416  .type = AVMEDIA_TYPE_AUDIO,
417  .id = AV_CODEC_ID_WMAV2,
418  .priv_data_size = sizeof(WMACodecContext),
419  .init = encode_init,
420  .encode2 = encode_superframe,
421  .close = ff_wma_end,
422  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
424  .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"),
425 };
426 #endif