FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
vorbisenc.c
Go to the documentation of this file.
1 /*
2  * copyright (c) 2006 Oded Shimon <ods15@ods15.dyndns.org>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 /**
22  * @file
23  * Native Vorbis encoder.
24  * @author Oded Shimon <ods15@ods15.dyndns.org>
25  */
26 
27 #include <float.h>
28 #include "avcodec.h"
29 #include "internal.h"
30 #include "fft.h"
31 #include "vorbis.h"
32 #include "vorbis_enc_data.h"
33 
34 #define BITSTREAM_WRITER_LE
35 #include "put_bits.h"
36 
37 #undef NDEBUG
38 #include <assert.h>
39 
40 typedef struct {
41  int nentries;
42  uint8_t *lens;
43  uint32_t *codewords;
44  int ndimensions;
45  float min;
46  float delta;
47  int seq_p;
48  int lookup;
49  int *quantlist;
50  float *dimensions;
51  float *pow2;
52 } vorbis_enc_codebook;
53 
54 typedef struct {
55  int dim;
56  int subclass;
57  int masterbook;
58  int *books;
59 } vorbis_enc_floor_class;
60 
61 typedef struct {
62  int partitions;
63  int *partition_to_class;
64  int nclasses;
65  vorbis_enc_floor_class *classes;
66  int multiplier;
67  int rangebits;
68  int values;
69  vorbis_floor1_entry *list;
70 } vorbis_enc_floor;
71 
72 typedef struct {
73  int type;
74  int begin;
75  int end;
76  int partition_size;
77  int classifications;
78  int classbook;
79  int8_t (*books)[8];
80  float (*maxes)[2];
81 } vorbis_enc_residue;
82 
83 typedef struct {
84  int submaps;
85  int *mux;
86  int *floor;
87  int *residue;
88  int coupling_steps;
89  int *magnitude;
90  int *angle;
91 } vorbis_enc_mapping;
92 
93 typedef struct {
94  int blockflag;
95  int mapping;
96 } vorbis_enc_mode;
97 
98 typedef struct {
99  int channels;
100  int sample_rate;
101  int log2_blocksize[2];
102  FFTContext mdct[2];
103  const float *win[2];
104  int have_saved;
105  float *saved;
106  float *samples;
107  float *floor; // also used for tmp values for mdct
108  float *coeffs; // also used for residue after floor
109  float quality;
110 
111  int ncodebooks;
112  vorbis_enc_codebook *codebooks;
113 
114  int nfloors;
115  vorbis_enc_floor *floors;
116 
117  int nresidues;
118  vorbis_enc_residue *residues;
119 
120  int nmappings;
121  vorbis_enc_mapping *mappings;
122 
123  int nmodes;
124  vorbis_enc_mode *modes;
125 
126  int64_t next_pts;
127 } vorbis_enc_context;
128 
129 #define MAX_CHANNELS 2
130 #define MAX_CODEBOOK_DIM 8
131 
132 #define MAX_FLOOR_CLASS_DIM 4
133 #define NUM_FLOOR_PARTITIONS 8
134 #define MAX_FLOOR_VALUES (MAX_FLOOR_CLASS_DIM*NUM_FLOOR_PARTITIONS+2)
135 
136 #define RESIDUE_SIZE 1600
137 #define RESIDUE_PART_SIZE 32
138 #define NUM_RESIDUE_PARTITIONS (RESIDUE_SIZE/RESIDUE_PART_SIZE)
139 
140 static inline int put_codeword(PutBitContext *pb, vorbis_enc_codebook *cb,
141  int entry)
142 {
143  assert(entry >= 0);
144  assert(entry < cb->nentries);
145  assert(cb->lens[entry]);
146  if (pb->size_in_bits - put_bits_count(pb) < cb->lens[entry])
147  return AVERROR(EINVAL);
148  put_bits(pb, cb->lens[entry], cb->codewords[entry]);
149  return 0;
150 }
151 
152 static int cb_lookup_vals(int lookup, int dimensions, int entries)
153 {
154  if (lookup == 1)
155  return ff_vorbis_nth_root(entries, dimensions);
156  else if (lookup == 2)
157  return dimensions *entries;
158  return 0;
159 }
160 
161 static int ready_codebook(vorbis_enc_codebook *cb)
162 {
163  int i;
164 
165  ff_vorbis_len2vlc(cb->lens, cb->codewords, cb->nentries);
166 
167  if (!cb->lookup) {
168  cb->pow2 = cb->dimensions = NULL;
169  } else {
170  int vals = cb_lookup_vals(cb->lookup, cb->ndimensions, cb->nentries);
171  cb->dimensions = av_malloc(sizeof(float) * cb->nentries * cb->ndimensions);
172  cb->pow2 = av_mallocz(sizeof(float) * cb->nentries);
173  if (!cb->dimensions || !cb->pow2)
174  return AVERROR(ENOMEM);
175  for (i = 0; i < cb->nentries; i++) {
176  float last = 0;
177  int j;
178  int div = 1;
179  for (j = 0; j < cb->ndimensions; j++) {
180  int off;
181  if (cb->lookup == 1)
182  off = (i / div) % vals; // lookup type 1
183  else
184  off = i * cb->ndimensions + j; // lookup type 2
185 
186  cb->dimensions[i * cb->ndimensions + j] = last + cb->min + cb->quantlist[off] * cb->delta;
187  if (cb->seq_p)
188  last = cb->dimensions[i * cb->ndimensions + j];
189  cb->pow2[i] += cb->dimensions[i * cb->ndimensions + j] * cb->dimensions[i * cb->ndimensions + j];
190  div *= vals;
191  }
192  cb->pow2[i] /= 2.;
193  }
194  }
195  return 0;
196 }
197 
198 static int ready_residue(vorbis_enc_residue *rc, vorbis_enc_context *venc)
199 {
200  int i;
201  assert(rc->type == 2);
202  rc->maxes = av_mallocz(sizeof(float[2]) * rc->classifications);
203  if (!rc->maxes)
204  return AVERROR(ENOMEM);
205  for (i = 0; i < rc->classifications; i++) {
206  int j;
207  vorbis_enc_codebook * cb;
208  for (j = 0; j < 8; j++)
209  if (rc->books[i][j] != -1)
210  break;
211  if (j == 8) // zero
212  continue;
213  cb = &venc->codebooks[rc->books[i][j]];
214  assert(cb->ndimensions >= 2);
215  assert(cb->lookup);
216 
217  for (j = 0; j < cb->nentries; j++) {
218  float a;
219  if (!cb->lens[j])
220  continue;
221  a = fabs(cb->dimensions[j * cb->ndimensions]);
222  if (a > rc->maxes[i][0])
223  rc->maxes[i][0] = a;
224  a = fabs(cb->dimensions[j * cb->ndimensions + 1]);
225  if (a > rc->maxes[i][1])
226  rc->maxes[i][1] = a;
227  }
228  }
229  // small bias
230  for (i = 0; i < rc->classifications; i++) {
231  rc->maxes[i][0] += 0.8;
232  rc->maxes[i][1] += 0.8;
233  }
234  return 0;
235 }
236 
237 static int create_vorbis_context(vorbis_enc_context *venc,
238  AVCodecContext *avctx)
239 {
240  vorbis_enc_floor *fc;
241  vorbis_enc_residue *rc;
242  vorbis_enc_mapping *mc;
243  int i, book, ret;
244 
245  venc->channels = avctx->channels;
246  venc->sample_rate = avctx->sample_rate;
247  venc->log2_blocksize[0] = venc->log2_blocksize[1] = 11;
248 
249  venc->ncodebooks = FF_ARRAY_ELEMS(cvectors);
250  venc->codebooks = av_malloc(sizeof(vorbis_enc_codebook) * venc->ncodebooks);
251  if (!venc->codebooks)
252  return AVERROR(ENOMEM);
253 
254  // codebook 0..14 - floor1 book, values 0..255
255  // codebook 15 residue masterbook
256  // codebook 16..29 residue
257  for (book = 0; book < venc->ncodebooks; book++) {
258  vorbis_enc_codebook *cb = &venc->codebooks[book];
259  int vals;
260  cb->ndimensions = cvectors[book].dim;
261  cb->nentries = cvectors[book].real_len;
262  cb->min = cvectors[book].min;
263  cb->delta = cvectors[book].delta;
264  cb->lookup = cvectors[book].lookup;
265  cb->seq_p = 0;
266 
267  cb->lens = av_malloc(sizeof(uint8_t) * cb->nentries);
268  cb->codewords = av_malloc(sizeof(uint32_t) * cb->nentries);
269  if (!cb->lens || !cb->codewords)
270  return AVERROR(ENOMEM);
271  memcpy(cb->lens, cvectors[book].clens, cvectors[book].len);
272  memset(cb->lens + cvectors[book].len, 0, cb->nentries - cvectors[book].len);
273 
274  if (cb->lookup) {
275  vals = cb_lookup_vals(cb->lookup, cb->ndimensions, cb->nentries);
276  cb->quantlist = av_malloc(sizeof(int) * vals);
277  if (!cb->quantlist)
278  return AVERROR(ENOMEM);
279  for (i = 0; i < vals; i++)
280  cb->quantlist[i] = cvectors[book].quant[i];
281  } else {
282  cb->quantlist = NULL;
283  }
284  if ((ret = ready_codebook(cb)) < 0)
285  return ret;
286  }
287 
288  venc->nfloors = 1;
289  venc->floors = av_malloc(sizeof(vorbis_enc_floor) * venc->nfloors);
290  if (!venc->floors)
291  return AVERROR(ENOMEM);
292 
293  // just 1 floor
294  fc = &venc->floors[0];
295  fc->partitions = NUM_FLOOR_PARTITIONS;
296  fc->partition_to_class = av_malloc(sizeof(int) * fc->partitions);
297  if (!fc->partition_to_class)
298  return AVERROR(ENOMEM);
299  fc->nclasses = 0;
300  for (i = 0; i < fc->partitions; i++) {
301  static const int a[] = {0, 1, 2, 2, 3, 3, 4, 4};
302  fc->partition_to_class[i] = a[i];
303  fc->nclasses = FFMAX(fc->nclasses, fc->partition_to_class[i]);
304  }
305  fc->nclasses++;
306  fc->classes = av_malloc(sizeof(vorbis_enc_floor_class) * fc->nclasses);
307  if (!fc->classes)
308  return AVERROR(ENOMEM);
309  for (i = 0; i < fc->nclasses; i++) {
310  vorbis_enc_floor_class * c = &fc->classes[i];
311  int j, books;
312  c->dim = floor_classes[i].dim;
313  c->subclass = floor_classes[i].subclass;
314  c->masterbook = floor_classes[i].masterbook;
315  books = (1 << c->subclass);
316  c->books = av_malloc(sizeof(int) * books);
317  if (!c->books)
318  return AVERROR(ENOMEM);
319  for (j = 0; j < books; j++)
320  c->books[j] = floor_classes[i].nbooks[j];
321  }
322  fc->multiplier = 2;
323  fc->rangebits = venc->log2_blocksize[0] - 1;
324 
325  fc->values = 2;
326  for (i = 0; i < fc->partitions; i++)
327  fc->values += fc->classes[fc->partition_to_class[i]].dim;
328 
329  fc->list = av_malloc(sizeof(vorbis_floor1_entry) * fc->values);
330  if (!fc->list)
331  return AVERROR(ENOMEM);
332  fc->list[0].x = 0;
333  fc->list[1].x = 1 << fc->rangebits;
334  for (i = 2; i < fc->values; i++) {
335  static const int a[] = {
336  93, 23,372, 6, 46,186,750, 14, 33, 65,
337  130,260,556, 3, 10, 18, 28, 39, 55, 79,
338  111,158,220,312,464,650,850
339  };
340  fc->list[i].x = a[i - 2];
341  }
342  if (ff_vorbis_ready_floor1_list(avctx, fc->list, fc->values))
343  return AVERROR_BUG;
344 
345  venc->nresidues = 1;
346  venc->residues = av_malloc(sizeof(vorbis_enc_residue) * venc->nresidues);
347  if (!venc->residues)
348  return AVERROR(ENOMEM);
349 
350  // single residue
351  rc = &venc->residues[0];
352  rc->type = 2;
353  rc->begin = 0;
354  rc->end = 1600;
355  rc->partition_size = 32;
356  rc->classifications = 10;
357  rc->classbook = 15;
358  rc->books = av_malloc(sizeof(*rc->books) * rc->classifications);
359  if (!rc->books)
360  return AVERROR(ENOMEM);
361  {
362  static const int8_t a[10][8] = {
363  { -1, -1, -1, -1, -1, -1, -1, -1, },
364  { -1, -1, 16, -1, -1, -1, -1, -1, },
365  { -1, -1, 17, -1, -1, -1, -1, -1, },
366  { -1, -1, 18, -1, -1, -1, -1, -1, },
367  { -1, -1, 19, -1, -1, -1, -1, -1, },
368  { -1, -1, 20, -1, -1, -1, -1, -1, },
369  { -1, -1, 21, -1, -1, -1, -1, -1, },
370  { 22, 23, -1, -1, -1, -1, -1, -1, },
371  { 24, 25, -1, -1, -1, -1, -1, -1, },
372  { 26, 27, 28, -1, -1, -1, -1, -1, },
373  };
374  memcpy(rc->books, a, sizeof a);
375  }
376  if ((ret = ready_residue(rc, venc)) < 0)
377  return ret;
378 
379  venc->nmappings = 1;
380  venc->mappings = av_malloc(sizeof(vorbis_enc_mapping) * venc->nmappings);
381  if (!venc->mappings)
382  return AVERROR(ENOMEM);
383 
384  // single mapping
385  mc = &venc->mappings[0];
386  mc->submaps = 1;
387  mc->mux = av_malloc(sizeof(int) * venc->channels);
388  if (!mc->mux)
389  return AVERROR(ENOMEM);
390  for (i = 0; i < venc->channels; i++)
391  mc->mux[i] = 0;
392  mc->floor = av_malloc(sizeof(int) * mc->submaps);
393  mc->residue = av_malloc(sizeof(int) * mc->submaps);
394  if (!mc->floor || !mc->residue)
395  return AVERROR(ENOMEM);
396  for (i = 0; i < mc->submaps; i++) {
397  mc->floor[i] = 0;
398  mc->residue[i] = 0;
399  }
400  mc->coupling_steps = venc->channels == 2 ? 1 : 0;
401  mc->magnitude = av_malloc(sizeof(int) * mc->coupling_steps);
402  mc->angle = av_malloc(sizeof(int) * mc->coupling_steps);
403  if (!mc->magnitude || !mc->angle)
404  return AVERROR(ENOMEM);
405  if (mc->coupling_steps) {
406  mc->magnitude[0] = 0;
407  mc->angle[0] = 1;
408  }
409 
410  venc->nmodes = 1;
411  venc->modes = av_malloc(sizeof(vorbis_enc_mode) * venc->nmodes);
412  if (!venc->modes)
413  return AVERROR(ENOMEM);
414 
415  // single mode
416  venc->modes[0].blockflag = 0;
417  venc->modes[0].mapping = 0;
418 
419  venc->have_saved = 0;
420  venc->saved = av_malloc(sizeof(float) * venc->channels * (1 << venc->log2_blocksize[1]) / 2);
421  venc->samples = av_malloc(sizeof(float) * venc->channels * (1 << venc->log2_blocksize[1]));
422  venc->floor = av_malloc(sizeof(float) * venc->channels * (1 << venc->log2_blocksize[1]) / 2);
423  venc->coeffs = av_malloc(sizeof(float) * venc->channels * (1 << venc->log2_blocksize[1]) / 2);
424  if (!venc->saved || !venc->samples || !venc->floor || !venc->coeffs)
425  return AVERROR(ENOMEM);
426 
427  venc->win[0] = ff_vorbis_vwin[venc->log2_blocksize[0] - 6];
428  venc->win[1] = ff_vorbis_vwin[venc->log2_blocksize[1] - 6];
429 
430  if ((ret = ff_mdct_init(&venc->mdct[0], venc->log2_blocksize[0], 0, 1.0)) < 0)
431  return ret;
432  if ((ret = ff_mdct_init(&venc->mdct[1], venc->log2_blocksize[1], 0, 1.0)) < 0)
433  return ret;
434 
435  return 0;
436 }
437 
438 static void put_float(PutBitContext *pb, float f)
439 {
440  int exp, mant;
441  uint32_t res = 0;
442  mant = (int)ldexp(frexp(f, &exp), 20);
443  exp += 788 - 20;
444  if (mant < 0) {
445  res |= (1U << 31);
446  mant = -mant;
447  }
448  res |= mant | (exp << 21);
449  put_bits32(pb, res);
450 }
451 
452 static void put_codebook_header(PutBitContext *pb, vorbis_enc_codebook *cb)
453 {
454  int i;
455  int ordered = 0;
456 
457  put_bits(pb, 24, 0x564342); //magic
458  put_bits(pb, 16, cb->ndimensions);
459  put_bits(pb, 24, cb->nentries);
460 
461  for (i = 1; i < cb->nentries; i++)
462  if (cb->lens[i] < cb->lens[i-1])
463  break;
464  if (i == cb->nentries)
465  ordered = 1;
466 
467  put_bits(pb, 1, ordered);
468  if (ordered) {
469  int len = cb->lens[0];
470  put_bits(pb, 5, len - 1);
471  i = 0;
472  while (i < cb->nentries) {
473  int j;
474  for (j = 0; j+i < cb->nentries; j++)
475  if (cb->lens[j+i] != len)
476  break;
477  put_bits(pb, ilog(cb->nentries - i), j);
478  i += j;
479  len++;
480  }
481  } else {
482  int sparse = 0;
483  for (i = 0; i < cb->nentries; i++)
484  if (!cb->lens[i])
485  break;
486  if (i != cb->nentries)
487  sparse = 1;
488  put_bits(pb, 1, sparse);
489 
490  for (i = 0; i < cb->nentries; i++) {
491  if (sparse)
492  put_bits(pb, 1, !!cb->lens[i]);
493  if (cb->lens[i])
494  put_bits(pb, 5, cb->lens[i] - 1);
495  }
496  }
497 
498  put_bits(pb, 4, cb->lookup);
499  if (cb->lookup) {
500  int tmp = cb_lookup_vals(cb->lookup, cb->ndimensions, cb->nentries);
501  int bits = ilog(cb->quantlist[0]);
502 
503  for (i = 1; i < tmp; i++)
504  bits = FFMAX(bits, ilog(cb->quantlist[i]));
505 
506  put_float(pb, cb->min);
507  put_float(pb, cb->delta);
508 
509  put_bits(pb, 4, bits - 1);
510  put_bits(pb, 1, cb->seq_p);
511 
512  for (i = 0; i < tmp; i++)
513  put_bits(pb, bits, cb->quantlist[i]);
514  }
515 }
516 
517 static void put_floor_header(PutBitContext *pb, vorbis_enc_floor *fc)
518 {
519  int i;
520 
521  put_bits(pb, 16, 1); // type, only floor1 is supported
522 
523  put_bits(pb, 5, fc->partitions);
524 
525  for (i = 0; i < fc->partitions; i++)
526  put_bits(pb, 4, fc->partition_to_class[i]);
527 
528  for (i = 0; i < fc->nclasses; i++) {
529  int j, books;
530 
531  put_bits(pb, 3, fc->classes[i].dim - 1);
532  put_bits(pb, 2, fc->classes[i].subclass);
533 
534  if (fc->classes[i].subclass)
535  put_bits(pb, 8, fc->classes[i].masterbook);
536 
537  books = (1 << fc->classes[i].subclass);
538 
539  for (j = 0; j < books; j++)
540  put_bits(pb, 8, fc->classes[i].books[j] + 1);
541  }
542 
543  put_bits(pb, 2, fc->multiplier - 1);
544  put_bits(pb, 4, fc->rangebits);
545 
546  for (i = 2; i < fc->values; i++)
547  put_bits(pb, fc->rangebits, fc->list[i].x);
548 }
549 
550 static void put_residue_header(PutBitContext *pb, vorbis_enc_residue *rc)
551 {
552  int i;
553 
554  put_bits(pb, 16, rc->type);
555 
556  put_bits(pb, 24, rc->begin);
557  put_bits(pb, 24, rc->end);
558  put_bits(pb, 24, rc->partition_size - 1);
559  put_bits(pb, 6, rc->classifications - 1);
560  put_bits(pb, 8, rc->classbook);
561 
562  for (i = 0; i < rc->classifications; i++) {
563  int j, tmp = 0;
564  for (j = 0; j < 8; j++)
565  tmp |= (rc->books[i][j] != -1) << j;
566 
567  put_bits(pb, 3, tmp & 7);
568  put_bits(pb, 1, tmp > 7);
569 
570  if (tmp > 7)
571  put_bits(pb, 5, tmp >> 3);
572  }
573 
574  for (i = 0; i < rc->classifications; i++) {
575  int j;
576  for (j = 0; j < 8; j++)
577  if (rc->books[i][j] != -1)
578  put_bits(pb, 8, rc->books[i][j]);
579  }
580 }
581 
582 static int put_main_header(vorbis_enc_context *venc, uint8_t **out)
583 {
584  int i;
585  PutBitContext pb;
586  uint8_t buffer[50000] = {0}, *p = buffer;
587  int buffer_len = sizeof buffer;
588  int len, hlens[3];
589 
590  // identification header
591  init_put_bits(&pb, p, buffer_len);
592  put_bits(&pb, 8, 1); //magic
593  for (i = 0; "vorbis"[i]; i++)
594  put_bits(&pb, 8, "vorbis"[i]);
595  put_bits32(&pb, 0); // version
596  put_bits(&pb, 8, venc->channels);
597  put_bits32(&pb, venc->sample_rate);
598  put_bits32(&pb, 0); // bitrate
599  put_bits32(&pb, 0); // bitrate
600  put_bits32(&pb, 0); // bitrate
601  put_bits(&pb, 4, venc->log2_blocksize[0]);
602  put_bits(&pb, 4, venc->log2_blocksize[1]);
603  put_bits(&pb, 1, 1); // framing
604 
605  flush_put_bits(&pb);
606  hlens[0] = put_bits_count(&pb) >> 3;
607  buffer_len -= hlens[0];
608  p += hlens[0];
609 
610  // comment header
611  init_put_bits(&pb, p, buffer_len);
612  put_bits(&pb, 8, 3); //magic
613  for (i = 0; "vorbis"[i]; i++)
614  put_bits(&pb, 8, "vorbis"[i]);
615  put_bits32(&pb, 0); // vendor length TODO
616  put_bits32(&pb, 0); // amount of comments
617  put_bits(&pb, 1, 1); // framing
618 
619  flush_put_bits(&pb);
620  hlens[1] = put_bits_count(&pb) >> 3;
621  buffer_len -= hlens[1];
622  p += hlens[1];
623 
624  // setup header
625  init_put_bits(&pb, p, buffer_len);
626  put_bits(&pb, 8, 5); //magic
627  for (i = 0; "vorbis"[i]; i++)
628  put_bits(&pb, 8, "vorbis"[i]);
629 
630  // codebooks
631  put_bits(&pb, 8, venc->ncodebooks - 1);
632  for (i = 0; i < venc->ncodebooks; i++)
633  put_codebook_header(&pb, &venc->codebooks[i]);
634 
635  // time domain, reserved, zero
636  put_bits(&pb, 6, 0);
637  put_bits(&pb, 16, 0);
638 
639  // floors
640  put_bits(&pb, 6, venc->nfloors - 1);
641  for (i = 0; i < venc->nfloors; i++)
642  put_floor_header(&pb, &venc->floors[i]);
643 
644  // residues
645  put_bits(&pb, 6, venc->nresidues - 1);
646  for (i = 0; i < venc->nresidues; i++)
647  put_residue_header(&pb, &venc->residues[i]);
648 
649  // mappings
650  put_bits(&pb, 6, venc->nmappings - 1);
651  for (i = 0; i < venc->nmappings; i++) {
652  vorbis_enc_mapping *mc = &venc->mappings[i];
653  int j;
654  put_bits(&pb, 16, 0); // mapping type
655 
656  put_bits(&pb, 1, mc->submaps > 1);
657  if (mc->submaps > 1)
658  put_bits(&pb, 4, mc->submaps - 1);
659 
660  put_bits(&pb, 1, !!mc->coupling_steps);
661  if (mc->coupling_steps) {
662  put_bits(&pb, 8, mc->coupling_steps - 1);
663  for (j = 0; j < mc->coupling_steps; j++) {
664  put_bits(&pb, ilog(venc->channels - 1), mc->magnitude[j]);
665  put_bits(&pb, ilog(venc->channels - 1), mc->angle[j]);
666  }
667  }
668 
669  put_bits(&pb, 2, 0); // reserved
670 
671  if (mc->submaps > 1)
672  for (j = 0; j < venc->channels; j++)
673  put_bits(&pb, 4, mc->mux[j]);
674 
675  for (j = 0; j < mc->submaps; j++) {
676  put_bits(&pb, 8, 0); // reserved time configuration
677  put_bits(&pb, 8, mc->floor[j]);
678  put_bits(&pb, 8, mc->residue[j]);
679  }
680  }
681 
682  // modes
683  put_bits(&pb, 6, venc->nmodes - 1);
684  for (i = 0; i < venc->nmodes; i++) {
685  put_bits(&pb, 1, venc->modes[i].blockflag);
686  put_bits(&pb, 16, 0); // reserved window type
687  put_bits(&pb, 16, 0); // reserved transform type
688  put_bits(&pb, 8, venc->modes[i].mapping);
689  }
690 
691  put_bits(&pb, 1, 1); // framing
692 
693  flush_put_bits(&pb);
694  hlens[2] = put_bits_count(&pb) >> 3;
695 
696  len = hlens[0] + hlens[1] + hlens[2];
697  p = *out = av_mallocz(64 + len + len/255);
698  if (!p)
699  return AVERROR(ENOMEM);
700 
701  *p++ = 2;
702  p += av_xiphlacing(p, hlens[0]);
703  p += av_xiphlacing(p, hlens[1]);
704  buffer_len = 0;
705  for (i = 0; i < 3; i++) {
706  memcpy(p, buffer + buffer_len, hlens[i]);
707  p += hlens[i];
708  buffer_len += hlens[i];
709  }
710 
711  return p - *out;
712 }
713 
714 static float get_floor_average(vorbis_enc_floor * fc, float *coeffs, int i)
715 {
716  int begin = fc->list[fc->list[FFMAX(i-1, 0)].sort].x;
717  int end = fc->list[fc->list[FFMIN(i+1, fc->values - 1)].sort].x;
718  int j;
719  float average = 0;
720 
721  for (j = begin; j < end; j++)
722  average += fabs(coeffs[j]);
723  return average / (end - begin);
724 }
725 
726 static void floor_fit(vorbis_enc_context *venc, vorbis_enc_floor *fc,
727  float *coeffs, uint16_t *posts, int samples)
728 {
729  int range = 255 / fc->multiplier + 1;
730  int i;
731  float tot_average = 0.;
732  float averages[MAX_FLOOR_VALUES];
733  for (i = 0; i < fc->values; i++) {
734  averages[i] = get_floor_average(fc, coeffs, i);
735  tot_average += averages[i];
736  }
737  tot_average /= fc->values;
738  tot_average /= venc->quality;
739 
740  for (i = 0; i < fc->values; i++) {
741  int position = fc->list[fc->list[i].sort].x;
742  float average = averages[i];
743  int j;
744 
745  average = sqrt(tot_average * average) * pow(1.25f, position*0.005f); // MAGIC!
746  for (j = 0; j < range - 1; j++)
747  if (ff_vorbis_floor1_inverse_db_table[j * fc->multiplier] > average)
748  break;
749  posts[fc->list[i].sort] = j;
750  }
751 }
752 
753 static int render_point(int x0, int y0, int x1, int y1, int x)
754 {
755  return y0 + (x - x0) * (y1 - y0) / (x1 - x0);
756 }
757 
758 static int floor_encode(vorbis_enc_context *venc, vorbis_enc_floor *fc,
759  PutBitContext *pb, uint16_t *posts,
760  float *floor, int samples)
761 {
762  int range = 255 / fc->multiplier + 1;
763  int coded[MAX_FLOOR_VALUES]; // first 2 values are unused
764  int i, counter;
765 
766  if (pb->size_in_bits - put_bits_count(pb) < 1 + 2 * ilog(range - 1))
767  return AVERROR(EINVAL);
768  put_bits(pb, 1, 1); // non zero
769  put_bits(pb, ilog(range - 1), posts[0]);
770  put_bits(pb, ilog(range - 1), posts[1]);
771  coded[0] = coded[1] = 1;
772 
773  for (i = 2; i < fc->values; i++) {
774  int predicted = render_point(fc->list[fc->list[i].low].x,
775  posts[fc->list[i].low],
776  fc->list[fc->list[i].high].x,
777  posts[fc->list[i].high],
778  fc->list[i].x);
779  int highroom = range - predicted;
780  int lowroom = predicted;
781  int room = FFMIN(highroom, lowroom);
782  if (predicted == posts[i]) {
783  coded[i] = 0; // must be used later as flag!
784  continue;
785  } else {
786  if (!coded[fc->list[i].low ])
787  coded[fc->list[i].low ] = -1;
788  if (!coded[fc->list[i].high])
789  coded[fc->list[i].high] = -1;
790  }
791  if (posts[i] > predicted) {
792  if (posts[i] - predicted > room)
793  coded[i] = posts[i] - predicted + lowroom;
794  else
795  coded[i] = (posts[i] - predicted) << 1;
796  } else {
797  if (predicted - posts[i] > room)
798  coded[i] = predicted - posts[i] + highroom - 1;
799  else
800  coded[i] = ((predicted - posts[i]) << 1) - 1;
801  }
802  }
803 
804  counter = 2;
805  for (i = 0; i < fc->partitions; i++) {
806  vorbis_enc_floor_class * c = &fc->classes[fc->partition_to_class[i]];
807  int k, cval = 0, csub = 1<<c->subclass;
808  if (c->subclass) {
809  vorbis_enc_codebook * book = &venc->codebooks[c->masterbook];
810  int cshift = 0;
811  for (k = 0; k < c->dim; k++) {
812  int l;
813  for (l = 0; l < csub; l++) {
814  int maxval = 1;
815  if (c->books[l] != -1)
816  maxval = venc->codebooks[c->books[l]].nentries;
817  // coded could be -1, but this still works, cause that is 0
818  if (coded[counter + k] < maxval)
819  break;
820  }
821  assert(l != csub);
822  cval |= l << cshift;
823  cshift += c->subclass;
824  }
825  if (put_codeword(pb, book, cval))
826  return AVERROR(EINVAL);
827  }
828  for (k = 0; k < c->dim; k++) {
829  int book = c->books[cval & (csub-1)];
830  int entry = coded[counter++];
831  cval >>= c->subclass;
832  if (book == -1)
833  continue;
834  if (entry == -1)
835  entry = 0;
836  if (put_codeword(pb, &venc->codebooks[book], entry))
837  return AVERROR(EINVAL);
838  }
839  }
840 
841  ff_vorbis_floor1_render_list(fc->list, fc->values, posts, coded,
842  fc->multiplier, floor, samples);
843 
844  return 0;
845 }
846 
847 static float *put_vector(vorbis_enc_codebook *book, PutBitContext *pb,
848  float *num)
849 {
850  int i, entry = -1;
851  float distance = FLT_MAX;
852  assert(book->dimensions);
853  for (i = 0; i < book->nentries; i++) {
854  float * vec = book->dimensions + i * book->ndimensions, d = book->pow2[i];
855  int j;
856  if (!book->lens[i])
857  continue;
858  for (j = 0; j < book->ndimensions; j++)
859  d -= vec[j] * num[j];
860  if (distance > d) {
861  entry = i;
862  distance = d;
863  }
864  }
865  if (put_codeword(pb, book, entry))
866  return NULL;
867  return &book->dimensions[entry * book->ndimensions];
868 }
869 
870 static int residue_encode(vorbis_enc_context *venc, vorbis_enc_residue *rc,
871  PutBitContext *pb, float *coeffs, int samples,
872  int real_ch)
873 {
874  int pass, i, j, p, k;
875  int psize = rc->partition_size;
876  int partitions = (rc->end - rc->begin) / psize;
877  int channels = (rc->type == 2) ? 1 : real_ch;
878  int classes[MAX_CHANNELS][NUM_RESIDUE_PARTITIONS];
879  int classwords = venc->codebooks[rc->classbook].ndimensions;
880 
881  assert(rc->type == 2);
882  assert(real_ch == 2);
883  for (p = 0; p < partitions; p++) {
884  float max1 = 0., max2 = 0.;
885  int s = rc->begin + p * psize;
886  for (k = s; k < s + psize; k += 2) {
887  max1 = FFMAX(max1, fabs(coeffs[ k / real_ch]));
888  max2 = FFMAX(max2, fabs(coeffs[samples + k / real_ch]));
889  }
890 
891  for (i = 0; i < rc->classifications - 1; i++)
892  if (max1 < rc->maxes[i][0] && max2 < rc->maxes[i][1])
893  break;
894  classes[0][p] = i;
895  }
896 
897  for (pass = 0; pass < 8; pass++) {
898  p = 0;
899  while (p < partitions) {
900  if (pass == 0)
901  for (j = 0; j < channels; j++) {
902  vorbis_enc_codebook * book = &venc->codebooks[rc->classbook];
903  int entry = 0;
904  for (i = 0; i < classwords; i++) {
905  entry *= rc->classifications;
906  entry += classes[j][p + i];
907  }
908  if (put_codeword(pb, book, entry))
909  return AVERROR(EINVAL);
910  }
911  for (i = 0; i < classwords && p < partitions; i++, p++) {
912  for (j = 0; j < channels; j++) {
913  int nbook = rc->books[classes[j][p]][pass];
914  vorbis_enc_codebook * book = &venc->codebooks[nbook];
915  float *buf = coeffs + samples*j + rc->begin + p*psize;
916  if (nbook == -1)
917  continue;
918 
919  assert(rc->type == 0 || rc->type == 2);
920  assert(!(psize % book->ndimensions));
921 
922  if (rc->type == 0) {
923  for (k = 0; k < psize; k += book->ndimensions) {
924  int l;
925  float *a = put_vector(book, pb, &buf[k]);
926  if (!a)
927  return AVERROR(EINVAL);
928  for (l = 0; l < book->ndimensions; l++)
929  buf[k + l] -= a[l];
930  }
931  } else {
932  int s = rc->begin + p * psize, a1, b1;
933  a1 = (s % real_ch) * samples;
934  b1 = s / real_ch;
935  s = real_ch * samples;
936  for (k = 0; k < psize; k += book->ndimensions) {
937  int dim, a2 = a1, b2 = b1;
938  float vec[MAX_CODEBOOK_DIM], *pv = vec;
939  for (dim = book->ndimensions; dim--; ) {
940  *pv++ = coeffs[a2 + b2];
941  if ((a2 += samples) == s) {
942  a2 = 0;
943  b2++;
944  }
945  }
946  pv = put_vector(book, pb, vec);
947  if (!pv)
948  return AVERROR(EINVAL);
949  for (dim = book->ndimensions; dim--; ) {
950  coeffs[a1 + b1] -= *pv++;
951  if ((a1 += samples) == s) {
952  a1 = 0;
953  b1++;
954  }
955  }
956  }
957  }
958  }
959  }
960  }
961  }
962  return 0;
963 }
964 
965 static int apply_window_and_mdct(vorbis_enc_context *venc,
966  float **audio, int samples)
967 {
968  int i, channel;
969  const float * win = venc->win[0];
970  int window_len = 1 << (venc->log2_blocksize[0] - 1);
971  float n = (float)(1 << venc->log2_blocksize[0]) / 4.;
972  // FIXME use dsp
973 
974  if (!venc->have_saved && !samples)
975  return 0;
976 
977  if (venc->have_saved) {
978  for (channel = 0; channel < venc->channels; channel++)
979  memcpy(venc->samples + channel * window_len * 2,
980  venc->saved + channel * window_len, sizeof(float) * window_len);
981  } else {
982  for (channel = 0; channel < venc->channels; channel++)
983  memset(venc->samples + channel * window_len * 2, 0,
984  sizeof(float) * window_len);
985  }
986 
987  if (samples) {
988  for (channel = 0; channel < venc->channels; channel++) {
989  float * offset = venc->samples + channel*window_len*2 + window_len;
990  for (i = 0; i < samples; i++)
991  offset[i] = audio[channel][i] / n * win[window_len - i - 1];
992  }
993  } else {
994  for (channel = 0; channel < venc->channels; channel++)
995  memset(venc->samples + channel * window_len * 2 + window_len,
996  0, sizeof(float) * window_len);
997  }
998 
999  for (channel = 0; channel < venc->channels; channel++)
1000  venc->mdct[0].mdct_calc(&venc->mdct[0], venc->coeffs + channel * window_len,
1001  venc->samples + channel * window_len * 2);
1002 
1003  if (samples) {
1004  for (channel = 0; channel < venc->channels; channel++) {
1005  float *offset = venc->saved + channel * window_len;
1006  for (i = 0; i < samples; i++)
1007  offset[i] = audio[channel][i] / n * win[i];
1008  }
1009  venc->have_saved = 1;
1010  } else {
1011  venc->have_saved = 0;
1012  }
1013  return 1;
1014 }
1015 
1016 static int vorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
1017  const AVFrame *frame, int *got_packet_ptr)
1018 {
1019  vorbis_enc_context *venc = avctx->priv_data;
1020  float **audio = frame ? (float **)frame->extended_data : NULL;
1021  int samples = frame ? frame->nb_samples : 0;
1022  vorbis_enc_mode *mode;
1023  vorbis_enc_mapping *mapping;
1024  PutBitContext pb;
1025  int i, ret;
1026 
1027  if (!apply_window_and_mdct(venc, audio, samples))
1028  return 0;
1029  samples = 1 << (venc->log2_blocksize[0] - 1);
1030 
1031  if ((ret = ff_alloc_packet2(avctx, avpkt, 8192)) < 0)
1032  return ret;
1033 
1034  init_put_bits(&pb, avpkt->data, avpkt->size);
1035 
1036  if (pb.size_in_bits - put_bits_count(&pb) < 1 + ilog(venc->nmodes - 1)) {
1037  av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
1038  return AVERROR(EINVAL);
1039  }
1040 
1041  put_bits(&pb, 1, 0); // magic bit
1042 
1043  put_bits(&pb, ilog(venc->nmodes - 1), 0); // 0 bits, the mode
1044 
1045  mode = &venc->modes[0];
1046  mapping = &venc->mappings[mode->mapping];
1047  if (mode->blockflag) {
1048  put_bits(&pb, 1, 0);
1049  put_bits(&pb, 1, 0);
1050  }
1051 
1052  for (i = 0; i < venc->channels; i++) {
1053  vorbis_enc_floor *fc = &venc->floors[mapping->floor[mapping->mux[i]]];
1054  uint16_t posts[MAX_FLOOR_VALUES];
1055  floor_fit(venc, fc, &venc->coeffs[i * samples], posts, samples);
1056  if (floor_encode(venc, fc, &pb, posts, &venc->floor[i * samples], samples)) {
1057  av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
1058  return AVERROR(EINVAL);
1059  }
1060  }
1061 
1062  for (i = 0; i < venc->channels * samples; i++)
1063  venc->coeffs[i] /= venc->floor[i];
1064 
1065  for (i = 0; i < mapping->coupling_steps; i++) {
1066  float *mag = venc->coeffs + mapping->magnitude[i] * samples;
1067  float *ang = venc->coeffs + mapping->angle[i] * samples;
1068  int j;
1069  for (j = 0; j < samples; j++) {
1070  float a = ang[j];
1071  ang[j] -= mag[j];
1072  if (mag[j] > 0)
1073  ang[j] = -ang[j];
1074  if (ang[j] < 0)
1075  mag[j] = a;
1076  }
1077  }
1078 
1079  if (residue_encode(venc, &venc->residues[mapping->residue[mapping->mux[0]]],
1080  &pb, venc->coeffs, samples, venc->channels)) {
1081  av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
1082  return AVERROR(EINVAL);
1083  }
1084 
1085  flush_put_bits(&pb);
1086  avpkt->size = put_bits_count(&pb) >> 3;
1087 
1088  avpkt->duration = ff_samples_to_time_base(avctx, avctx->frame_size);
1089  if (frame) {
1090  if (frame->pts != AV_NOPTS_VALUE)
1091  avpkt->pts = ff_samples_to_time_base(avctx, frame->pts);
1092  } else
1093  avpkt->pts = venc->next_pts;
1094  if (avpkt->pts != AV_NOPTS_VALUE)
1095  venc->next_pts = avpkt->pts + avpkt->duration;
1096 
1097  *got_packet_ptr = 1;
1098  return 0;
1099 }
1100 
1101 
1102 static av_cold int vorbis_encode_close(AVCodecContext *avctx)
1103 {
1104  vorbis_enc_context *venc = avctx->priv_data;
1105  int i;
1106 
1107  if (venc->codebooks)
1108  for (i = 0; i < venc->ncodebooks; i++) {
1109  av_freep(&venc->codebooks[i].lens);
1110  av_freep(&venc->codebooks[i].codewords);
1111  av_freep(&venc->codebooks[i].quantlist);
1112  av_freep(&venc->codebooks[i].dimensions);
1113  av_freep(&venc->codebooks[i].pow2);
1114  }
1115  av_freep(&venc->codebooks);
1116 
1117  if (venc->floors)
1118  for (i = 0; i < venc->nfloors; i++) {
1119  int j;
1120  if (venc->floors[i].classes)
1121  for (j = 0; j < venc->floors[i].nclasses; j++)
1122  av_freep(&venc->floors[i].classes[j].books);
1123  av_freep(&venc->floors[i].classes);
1124  av_freep(&venc->floors[i].partition_to_class);
1125  av_freep(&venc->floors[i].list);
1126  }
1127  av_freep(&venc->floors);
1128 
1129  if (venc->residues)
1130  for (i = 0; i < venc->nresidues; i++) {
1131  av_freep(&venc->residues[i].books);
1132  av_freep(&venc->residues[i].maxes);
1133  }
1134  av_freep(&venc->residues);
1135 
1136  if (venc->mappings)
1137  for (i = 0; i < venc->nmappings; i++) {
1138  av_freep(&venc->mappings[i].mux);
1139  av_freep(&venc->mappings[i].floor);
1140  av_freep(&venc->mappings[i].residue);
1141  av_freep(&venc->mappings[i].magnitude);
1142  av_freep(&venc->mappings[i].angle);
1143  }
1144  av_freep(&venc->mappings);
1145 
1146  av_freep(&venc->modes);
1147 
1148  av_freep(&venc->saved);
1149  av_freep(&venc->samples);
1150  av_freep(&venc->floor);
1151  av_freep(&venc->coeffs);
1152 
1153  ff_mdct_end(&venc->mdct[0]);
1154  ff_mdct_end(&venc->mdct[1]);
1155 
1156  av_freep(&avctx->extradata);
1157 
1158  return 0 ;
1159 }
1160 
1161 static av_cold int vorbis_encode_init(AVCodecContext *avctx)
1162 {
1163  vorbis_enc_context *venc = avctx->priv_data;
1164  int ret;
1165 
1166  if (avctx->channels != 2) {
1167  av_log(avctx, AV_LOG_ERROR, "Current FFmpeg Vorbis encoder only supports 2 channels.\n");
1168  return -1;
1169  }
1170 
1171  if ((ret = create_vorbis_context(venc, avctx)) < 0)
1172  goto error;
1173 
1174  avctx->bit_rate = 0;
1175  if (avctx->flags & CODEC_FLAG_QSCALE)
1176  venc->quality = avctx->global_quality / (float)FF_QP2LAMBDA;
1177  else
1178  venc->quality = 8;
1179  venc->quality *= venc->quality;
1180 
1181  if ((ret = put_main_header(venc, (uint8_t**)&avctx->extradata)) < 0)
1182  goto error;
1183  avctx->extradata_size = ret;
1184 
1185  avctx->frame_size = 1 << (venc->log2_blocksize[0] - 1);
1186 
1187  return 0;
1188 error:
1189  vorbis_encode_close(avctx);
1190  return ret;
1191 }
1192 
1193 AVCodec ff_vorbis_encoder = {
1194  .name = "vorbis",
1195  .type = AVMEDIA_TYPE_AUDIO,
1196  .id = AV_CODEC_ID_VORBIS,
1197  .priv_data_size = sizeof(vorbis_enc_context),
1198  .init = vorbis_encode_init,
1199  .encode2 = vorbis_encode_frame,
1200  .close = vorbis_encode_close,
1201  .capabilities = CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
1202  .sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
1203  AV_SAMPLE_FMT_NONE },
1204  .long_name = NULL_IF_CONFIG_SMALL("Vorbis"),
1205 };
Generated on Wed Jul 10 2013 23:48:07 for FFmpeg by   doxygen 1.8.2