FFmpeg
 All Data Structures Namespaces 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 "libavutil/attributes.h"
23 #include "avcodec.h"
24 #include "internal.h"
25 #include "wma.h"
26 #include "libavutil/avassert.h"
27 
28 
30 {
31  WMACodecContext *s = avctx->priv_data;
32  int i, flags1, flags2, block_align;
33  uint8_t *extradata;
34 
35  s->avctx = avctx;
36 
37  if(avctx->channels > MAX_CHANNELS) {
38  av_log(avctx, AV_LOG_ERROR, "too many channels: got %i, need %i or fewer\n",
39  avctx->channels, MAX_CHANNELS);
40  return AVERROR(EINVAL);
41  }
42 
43  if (avctx->sample_rate > 48000) {
44  av_log(avctx, AV_LOG_ERROR, "sample rate is too high: %d > 48kHz\n",
45  avctx->sample_rate);
46  return AVERROR(EINVAL);
47  }
48 
49  if(avctx->bit_rate < 24*1000) {
50  av_log(avctx, AV_LOG_ERROR, "bitrate too low: got %i, need 24000 or higher\n",
51  avctx->bit_rate);
52  return AVERROR(EINVAL);
53  }
54 
55  /* extract flag infos */
56  flags1 = 0;
57  flags2 = 1;
58  if (avctx->codec->id == AV_CODEC_ID_WMAV1) {
59  extradata= av_malloc(4);
60  avctx->extradata_size= 4;
61  AV_WL16(extradata, flags1);
62  AV_WL16(extradata+2, flags2);
63  } else if (avctx->codec->id == AV_CODEC_ID_WMAV2) {
64  extradata= av_mallocz(10);
65  avctx->extradata_size= 10;
66  AV_WL32(extradata, flags1);
67  AV_WL16(extradata+4, flags2);
68  }else
69  av_assert0(0);
70  avctx->extradata= extradata;
71  s->use_exp_vlc = flags2 & 0x0001;
72  s->use_bit_reservoir = flags2 & 0x0002;
73  s->use_variable_block_len = flags2 & 0x0004;
74  if (avctx->channels == 2)
75  s->ms_stereo = 1;
76 
77  ff_wma_init(avctx, flags2);
78 
79  /* init MDCT */
80  for(i = 0; i < s->nb_block_sizes; i++)
81  ff_mdct_init(&s->mdct_ctx[i], s->frame_len_bits - i + 1, 0, 1.0);
82 
83  block_align = avctx->bit_rate * (int64_t)s->frame_len /
84  (avctx->sample_rate * 8);
85  block_align = FFMIN(block_align, MAX_CODED_SUPERFRAME_SIZE);
86  avctx->block_align = block_align;
87 
88  avctx->frame_size = avctx->delay = s->frame_len;
89 
90  return 0;
91 }
92 
93 
94 static void apply_window_and_mdct(AVCodecContext * avctx, const AVFrame *frame)
95 {
96  WMACodecContext *s = avctx->priv_data;
97  float **audio = (float **)frame->extended_data;
98  int len = frame->nb_samples;
99  int window_index= s->frame_len_bits - s->block_len_bits;
100  FFTContext *mdct = &s->mdct_ctx[window_index];
101  int ch;
102  const float * win = s->windows[window_index];
103  int window_len = 1 << s->block_len_bits;
104  float n = 2.0 * 32768.0 / window_len;
105 
106  for (ch = 0; ch < avctx->channels; ch++) {
107  memcpy(s->output, s->frame_out[ch], window_len * sizeof(*s->output));
108  s->fdsp.vector_fmul_scalar(s->frame_out[ch], audio[ch], n, len);
109  s->fdsp.vector_fmul_reverse(&s->output[window_len], s->frame_out[ch], win, len);
110  s->fdsp.vector_fmul(s->frame_out[ch], s->frame_out[ch], win, len);
111  mdct->mdct_calc(mdct, s->coefs[ch], s->output);
112  }
113 }
114 
115 //FIXME use for decoding too
116 static void init_exp(WMACodecContext *s, int ch, const int *exp_param){
117  int n;
118  const uint16_t *ptr;
119  float v, *q, max_scale, *q_end;
120 
121  ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
122  q = s->exponents[ch];
123  q_end = q + s->block_len;
124  max_scale = 0;
125  while (q < q_end) {
126  /* XXX: use a table */
127  v = pow(10, *exp_param++ * (1.0 / 16.0));
128  max_scale= FFMAX(max_scale, v);
129  n = *ptr++;
130  do {
131  *q++ = v;
132  } while (--n);
133  }
134  s->max_exponent[ch] = max_scale;
135 }
136 
137 static void encode_exp_vlc(WMACodecContext *s, int ch, const int *exp_param){
138  int last_exp;
139  const uint16_t *ptr;
140  float *q, *q_end;
141 
142  ptr = s->exponent_bands[s->frame_len_bits - s->block_len_bits];
143  q = s->exponents[ch];
144  q_end = q + s->block_len;
145  if (s->version == 1) {
146  last_exp= *exp_param++;
147  av_assert0(last_exp-10 >= 0 && last_exp-10 < 32);
148  put_bits(&s->pb, 5, last_exp - 10);
149  q+= *ptr++;
150  }else
151  last_exp = 36;
152  while (q < q_end) {
153  int exp = *exp_param++;
154  int code = exp - last_exp + 60;
155  av_assert1(code >= 0 && code < 120);
157  /* XXX: use a table */
158  q+= *ptr++;
159  last_exp= exp;
160  }
161 }
162 
163 static int encode_block(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], int total_gain){
164  int v, bsize, ch, coef_nb_bits, parse_exponents;
165  float mdct_norm;
166  int nb_coefs[MAX_CHANNELS];
167  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};
168 
169  //FIXME remove duplication relative to decoder
170  if (s->use_variable_block_len) {
171  av_assert0(0); //FIXME not implemented
172  }else{
173  /* fixed block len */
177  }
178 
179  s->block_len = 1 << s->block_len_bits;
180 // av_assert0((s->block_pos + s->block_len) <= s->frame_len);
181  bsize = s->frame_len_bits - s->block_len_bits;
182 
183  //FIXME factor
184  v = s->coefs_end[bsize] - s->coefs_start;
185  for (ch = 0; ch < s->avctx->channels; ch++)
186  nb_coefs[ch] = v;
187  {
188  int n4 = s->block_len / 2;
189  mdct_norm = 1.0 / (float)n4;
190  if (s->version == 1) {
191  mdct_norm *= sqrt(n4);
192  }
193  }
194 
195  if (s->avctx->channels == 2) {
196  put_bits(&s->pb, 1, !!s->ms_stereo);
197  }
198 
199  for (ch = 0; ch < s->avctx->channels; ch++) {
200  s->channel_coded[ch] = 1; //FIXME only set channel_coded when needed, instead of always
201  if (s->channel_coded[ch]) {
202  init_exp(s, ch, fixed_exp);
203  }
204  }
205 
206  for (ch = 0; ch < s->avctx->channels; ch++) {
207  if (s->channel_coded[ch]) {
208  WMACoef *coefs1;
209  float *coefs, *exponents, mult;
210  int i, n;
211 
212  coefs1 = s->coefs1[ch];
213  exponents = s->exponents[ch];
214  mult = pow(10, total_gain * 0.05) / s->max_exponent[ch];
215  mult *= mdct_norm;
216  coefs = src_coefs[ch];
217  if (s->use_noise_coding && 0) {
218  av_assert0(0); //FIXME not implemented
219  } else {
220  coefs += s->coefs_start;
221  n = nb_coefs[ch];
222  for(i = 0;i < n; i++){
223  double t= *coefs++ / (exponents[i] * mult);
224  if(t<-32768 || t>32767)
225  return -1;
226 
227  coefs1[i] = lrint(t);
228  }
229  }
230  }
231  }
232 
233  v = 0;
234  for (ch = 0; ch < s->avctx->channels; ch++) {
235  int a = s->channel_coded[ch];
236  put_bits(&s->pb, 1, a);
237  v |= a;
238  }
239 
240  if (!v)
241  return 1;
242 
243  for(v= total_gain-1; v>=127; v-= 127)
244  put_bits(&s->pb, 7, 127);
245  put_bits(&s->pb, 7, v);
246 
247  coef_nb_bits= ff_wma_total_gain_to_bits(total_gain);
248 
249  if (s->use_noise_coding) {
250  for (ch = 0; ch < s->avctx->channels; ch++) {
251  if (s->channel_coded[ch]) {
252  int i, n;
253  n = s->exponent_high_sizes[bsize];
254  for(i=0;i<n;i++) {
255  put_bits(&s->pb, 1, s->high_band_coded[ch][i]= 0);
256  if (0)
257  nb_coefs[ch] -= s->exponent_high_bands[bsize][i];
258  }
259  }
260  }
261  }
262 
263  parse_exponents = 1;
264  if (s->block_len_bits != s->frame_len_bits) {
265  put_bits(&s->pb, 1, parse_exponents);
266  }
267 
268  if (parse_exponents) {
269  for (ch = 0; ch < s->avctx->channels; ch++) {
270  if (s->channel_coded[ch]) {
271  if (s->use_exp_vlc) {
272  encode_exp_vlc(s, ch, fixed_exp);
273  } else {
274  av_assert0(0); //FIXME not implemented
275 // encode_exp_lsp(s, ch);
276  }
277  }
278  }
279  } else {
280  av_assert0(0); //FIXME not implemented
281  }
282 
283  for (ch = 0; ch < s->avctx->channels; ch++) {
284  if (s->channel_coded[ch]) {
285  int run, tindex;
286  WMACoef *ptr, *eptr;
287  tindex = (ch == 1 && s->ms_stereo);
288  ptr = &s->coefs1[ch][0];
289  eptr = ptr + nb_coefs[ch];
290 
291  run=0;
292  for(;ptr < eptr; ptr++){
293  if(*ptr){
294  int level= *ptr;
295  int abs_level= FFABS(level);
296  int code= 0;
297  if(abs_level <= s->coef_vlcs[tindex]->max_level){
298  if(run < s->coef_vlcs[tindex]->levels[abs_level-1])
299  code= run + s->int_table[tindex][abs_level-1];
300  }
301 
302  av_assert2(code < s->coef_vlcs[tindex]->n);
303  put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[code], s->coef_vlcs[tindex]->huffcodes[code]);
304 
305  if(code == 0){
306  if(1<<coef_nb_bits <= abs_level)
307  return -1;
308 
309  put_bits(&s->pb, coef_nb_bits, abs_level);
310  put_bits(&s->pb, s->frame_len_bits, run);
311  }
312  put_bits(&s->pb, 1, level < 0); //FIXME the sign is fliped somewhere
313  run=0;
314  }else{
315  run++;
316  }
317  }
318  if(run)
319  put_bits(&s->pb, s->coef_vlcs[tindex]->huffbits[1], s->coef_vlcs[tindex]->huffcodes[1]);
320  }
321  if (s->version == 1 && s->avctx->channels >= 2) {
323  }
324  }
325  return 0;
326 }
327 
328 static int encode_frame(WMACodecContext *s, float (*src_coefs)[BLOCK_MAX_SIZE], uint8_t *buf, int buf_size, int total_gain){
329  init_put_bits(&s->pb, buf, buf_size);
330 
331  if (s->use_bit_reservoir) {
332  av_assert0(0);//FIXME not implemented
333  }else{
334  if(encode_block(s, src_coefs, total_gain) < 0)
335  return INT_MAX;
336  }
337 
339 
340  return put_bits_count(&s->pb) / 8 - s->avctx->block_align;
341 }
342 
343 static int encode_superframe(AVCodecContext *avctx, AVPacket *avpkt,
344  const AVFrame *frame, int *got_packet_ptr)
345 {
346  WMACodecContext *s = avctx->priv_data;
347  int i, total_gain, ret, error;
348 
349  s->block_len_bits= s->frame_len_bits; //required by non variable block len
350  s->block_len = 1 << s->block_len_bits;
351 
352  apply_window_and_mdct(avctx, frame);
353 
354  if (s->ms_stereo) {
355  float a, b;
356  int i;
357 
358  for(i = 0; i < s->block_len; i++) {
359  a = s->coefs[0][i]*0.5;
360  b = s->coefs[1][i]*0.5;
361  s->coefs[0][i] = a + b;
362  s->coefs[1][i] = a - b;
363  }
364  }
365 
366  if ((ret = ff_alloc_packet2(avctx, avpkt, 2 * MAX_CODED_SUPERFRAME_SIZE)) < 0)
367  return ret;
368 
369  total_gain= 128;
370  for(i=64; i; i>>=1){
371  error = encode_frame(s, s->coefs, avpkt->data, avpkt->size,
372  total_gain - i);
373  if(error<=0)
374  total_gain-= i;
375  }
376 
377  while(total_gain <= 128 && error > 0)
378  error = encode_frame(s, s->coefs, avpkt->data, avpkt->size, total_gain++);
379  if (error > 0) {
380  av_log(avctx, AV_LOG_ERROR, "Invalid input data or requested bitrate too low, cannot encode\n");
381  avpkt->size = 0;
382  return AVERROR(EINVAL);
383  }
384  av_assert0((put_bits_count(&s->pb) & 7) == 0);
385  i= avctx->block_align - (put_bits_count(&s->pb)+7)/8;
386  av_assert0(i>=0);
387  while(i--)
388  put_bits(&s->pb, 8, 'N');
389 
390  flush_put_bits(&s->pb);
391  av_assert0(put_bits_ptr(&s->pb) - s->pb.buf == avctx->block_align);
392 
393  if (frame->pts != AV_NOPTS_VALUE)
394  avpkt->pts = frame->pts - ff_samples_to_time_base(avctx, avctx->delay);
395 
396  avpkt->size = avctx->block_align;
397  *got_packet_ptr = 1;
398  return 0;
399 }
400 
401 #if CONFIG_WMAV1_ENCODER
402 AVCodec ff_wmav1_encoder = {
403  .name = "wmav1",
404  .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 1"),
405  .type = AVMEDIA_TYPE_AUDIO,
406  .id = AV_CODEC_ID_WMAV1,
407  .priv_data_size = sizeof(WMACodecContext),
408  .init = encode_init,
409  .encode2 = encode_superframe,
410  .close = ff_wma_end,
411  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
413 };
414 #endif
415 #if CONFIG_WMAV2_ENCODER
416 AVCodec ff_wmav2_encoder = {
417  .name = "wmav2",
418  .long_name = NULL_IF_CONFIG_SMALL("Windows Media Audio 2"),
419  .type = AVMEDIA_TYPE_AUDIO,
420  .id = AV_CODEC_ID_WMAV2,
421  .priv_data_size = sizeof(WMACodecContext),
422  .init = encode_init,
423  .encode2 = encode_superframe,
424  .close = ff_wma_end,
425  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
427 };
428 #endif