avformat/mpeg: demux ivtv captions
[ffmpeg.git] / libavcodec / vorbisenc.c
blob62684a7d8c15891539f21cf09f7d1dcecb340776
1 /*
2 * copyright (c) 2006 Oded Shimon <ods15@ods15.dyndns.org>
4 * This file is part of FFmpeg.
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.
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.
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
21 /**
22 * @file
23 * Native Vorbis encoder.
24 * @author Oded Shimon <ods15@ods15.dyndns.org>
27 #include <float.h>
28 #include "libavutil/float_dsp.h"
29 #include "libavutil/mem.h"
30 #include "libavutil/tx.h"
32 #include "avcodec.h"
33 #include "codec_internal.h"
34 #include "encode.h"
35 #include "mathops.h"
36 #include "vorbis.h"
37 #include "vorbis_data.h"
38 #include "vorbis_enc_data.h"
40 #include "audio_frame_queue.h"
41 #include "libavfilter/bufferqueue.h"
43 #define BITSTREAM_WRITER_LE
44 #include "put_bits.h"
46 #undef NDEBUG
47 #include <assert.h>
49 typedef struct vorbis_enc_codebook {
50 int nentries;
51 uint8_t *lens;
52 uint32_t *codewords;
53 int ndimensions;
54 float min;
55 float delta;
56 int seq_p;
57 int lookup;
58 int *quantlist;
59 float *dimensions;
60 float *pow2;
61 } vorbis_enc_codebook;
63 typedef struct vorbis_enc_floor_class {
64 int dim;
65 int subclass;
66 int masterbook;
67 int *books;
68 } vorbis_enc_floor_class;
70 typedef struct vorbis_enc_floor {
71 int partitions;
72 int *partition_to_class;
73 int nclasses;
74 vorbis_enc_floor_class *classes;
75 int multiplier;
76 int rangebits;
77 int values;
78 vorbis_floor1_entry *list;
79 } vorbis_enc_floor;
81 typedef struct vorbis_enc_residue {
82 int type;
83 int begin;
84 int end;
85 int partition_size;
86 int classifications;
87 int classbook;
88 int8_t (*books)[8];
89 float (*maxes)[2];
90 } vorbis_enc_residue;
92 typedef struct vorbis_enc_mapping {
93 int submaps;
94 int *mux;
95 int *floor;
96 int *residue;
97 int coupling_steps;
98 int *magnitude;
99 int *angle;
100 } vorbis_enc_mapping;
102 typedef struct vorbis_enc_mode {
103 int blockflag;
104 int mapping;
105 } vorbis_enc_mode;
107 typedef struct vorbis_enc_context {
108 int channels;
109 int sample_rate;
110 int log2_blocksize[2];
111 AVTXContext *mdct[2];
112 av_tx_fn mdct_fn[2];
113 const float *win[2];
114 int have_saved;
115 float *saved;
116 float *samples;
117 float *floor; // also used for tmp values for mdct
118 float *coeffs; // also used for residue after floor
119 float *scratch; // used for tmp values for psy model
120 float quality;
122 AudioFrameQueue afq;
123 struct FFBufQueue bufqueue;
125 int ncodebooks;
126 vorbis_enc_codebook *codebooks;
128 int nfloors;
129 vorbis_enc_floor *floors;
131 int nresidues;
132 vorbis_enc_residue *residues;
134 int nmappings;
135 vorbis_enc_mapping *mappings;
137 int nmodes;
138 vorbis_enc_mode *modes;
140 int64_t next_pts;
142 AVFloatDSPContext *fdsp;
143 } vorbis_enc_context;
145 #define MAX_CHANNELS 2
146 #define MAX_CODEBOOK_DIM 8
148 #define MAX_FLOOR_CLASS_DIM 4
149 #define NUM_FLOOR_PARTITIONS 8
150 #define MAX_FLOOR_VALUES (MAX_FLOOR_CLASS_DIM*NUM_FLOOR_PARTITIONS+2)
152 #define RESIDUE_SIZE 1600
153 #define RESIDUE_PART_SIZE 32
154 #define NUM_RESIDUE_PARTITIONS (RESIDUE_SIZE/RESIDUE_PART_SIZE)
156 static inline int put_codeword(PutBitContext *pb, vorbis_enc_codebook *cb,
157 int entry)
159 av_assert2(entry >= 0);
160 av_assert2(entry < cb->nentries);
161 av_assert2(cb->lens[entry]);
162 if (put_bits_left(pb) < cb->lens[entry])
163 return AVERROR(EINVAL);
164 put_bits(pb, cb->lens[entry], cb->codewords[entry]);
165 return 0;
168 static int cb_lookup_vals(int lookup, int dimensions, int entries)
170 if (lookup == 1)
171 return ff_vorbis_nth_root(entries, dimensions);
172 else if (lookup == 2)
173 return dimensions *entries;
174 return 0;
177 static int ready_codebook(vorbis_enc_codebook *cb)
179 int i;
181 ff_vorbis_len2vlc(cb->lens, cb->codewords, cb->nentries);
183 if (!cb->lookup) {
184 cb->pow2 = cb->dimensions = NULL;
185 } else {
186 int vals = cb_lookup_vals(cb->lookup, cb->ndimensions, cb->nentries);
187 cb->dimensions = av_malloc_array(cb->nentries, sizeof(float) * cb->ndimensions);
188 cb->pow2 = av_calloc(cb->nentries, sizeof(*cb->pow2));
189 if (!cb->dimensions || !cb->pow2)
190 return AVERROR(ENOMEM);
191 for (i = 0; i < cb->nentries; i++) {
192 float last = 0;
193 int j;
194 int div = 1;
195 for (j = 0; j < cb->ndimensions; j++) {
196 int off;
197 if (cb->lookup == 1)
198 off = (i / div) % vals; // lookup type 1
199 else
200 off = i * cb->ndimensions + j; // lookup type 2
202 cb->dimensions[i * cb->ndimensions + j] = last + cb->min + cb->quantlist[off] * cb->delta;
203 if (cb->seq_p)
204 last = cb->dimensions[i * cb->ndimensions + j];
205 cb->pow2[i] += cb->dimensions[i * cb->ndimensions + j] * cb->dimensions[i * cb->ndimensions + j];
206 div *= vals;
208 cb->pow2[i] /= 2.0;
211 return 0;
214 static int ready_residue(vorbis_enc_residue *rc, vorbis_enc_context *venc)
216 int i;
217 av_assert0(rc->type == 2);
218 rc->maxes = av_calloc(rc->classifications, sizeof(*rc->maxes));
219 if (!rc->maxes)
220 return AVERROR(ENOMEM);
221 for (i = 0; i < rc->classifications; i++) {
222 int j;
223 vorbis_enc_codebook * cb;
224 for (j = 0; j < 8; j++)
225 if (rc->books[i][j] != -1)
226 break;
227 if (j == 8) // zero
228 continue;
229 cb = &venc->codebooks[rc->books[i][j]];
230 assert(cb->ndimensions >= 2);
231 assert(cb->lookup);
233 for (j = 0; j < cb->nentries; j++) {
234 float a;
235 if (!cb->lens[j])
236 continue;
237 a = fabs(cb->dimensions[j * cb->ndimensions]);
238 if (a > rc->maxes[i][0])
239 rc->maxes[i][0] = a;
240 a = fabs(cb->dimensions[j * cb->ndimensions + 1]);
241 if (a > rc->maxes[i][1])
242 rc->maxes[i][1] = a;
245 // small bias
246 for (i = 0; i < rc->classifications; i++) {
247 rc->maxes[i][0] += 0.8;
248 rc->maxes[i][1] += 0.8;
250 return 0;
253 static av_cold int dsp_init(AVCodecContext *avctx, vorbis_enc_context *venc)
255 int ret = 0;
256 float scale = 1.0f;
258 venc->fdsp = avpriv_float_dsp_alloc(avctx->flags & AV_CODEC_FLAG_BITEXACT);
259 if (!venc->fdsp)
260 return AVERROR(ENOMEM);
262 // init windows
263 venc->win[0] = ff_vorbis_vwin[venc->log2_blocksize[0] - 6];
264 venc->win[1] = ff_vorbis_vwin[venc->log2_blocksize[1] - 6];
266 if ((ret = av_tx_init(&venc->mdct[0], &venc->mdct_fn[0], AV_TX_FLOAT_MDCT,
267 0, 1 << (venc->log2_blocksize[0] - 1), &scale, 0)) < 0)
268 return ret;
269 if ((ret = av_tx_init(&venc->mdct[1], &venc->mdct_fn[1], AV_TX_FLOAT_MDCT,
270 0, 1 << (venc->log2_blocksize[1] - 1), &scale, 0)) < 0)
271 return ret;
273 return 0;
276 static int create_vorbis_context(vorbis_enc_context *venc,
277 AVCodecContext *avctx)
279 vorbis_enc_floor *fc;
280 vorbis_enc_residue *rc;
281 vorbis_enc_mapping *mc;
282 const uint8_t *clens, *quant;
283 int i, book, ret;
285 venc->channels = avctx->ch_layout.nb_channels;
286 venc->sample_rate = avctx->sample_rate;
287 venc->log2_blocksize[0] = venc->log2_blocksize[1] = 11;
289 venc->ncodebooks = FF_ARRAY_ELEMS(cvectors);
290 venc->codebooks = av_mallocz(sizeof(vorbis_enc_codebook) * venc->ncodebooks);
291 if (!venc->codebooks)
292 return AVERROR(ENOMEM);
294 // codebook 0..14 - floor1 book, values 0..255
295 // codebook 15 residue masterbook
296 // codebook 16..29 residue
297 clens = codebooks;
298 quant = quant_tables;
299 for (book = 0; book < venc->ncodebooks; book++) {
300 vorbis_enc_codebook *cb = &venc->codebooks[book];
301 int vals;
302 cb->ndimensions = cvectors[book].dim;
303 cb->nentries = cvectors[book].real_len;
304 cb->min = cvectors[book].min;
305 cb->delta = cvectors[book].delta;
306 cb->lookup = cvectors[book].lookup;
307 cb->seq_p = 0;
309 cb->lens = av_malloc_array(cb->nentries, sizeof(uint8_t));
310 cb->codewords = av_malloc_array(cb->nentries, sizeof(uint32_t));
311 if (!cb->lens || !cb->codewords)
312 return AVERROR(ENOMEM);
313 memcpy(cb->lens, clens, cvectors[book].len);
314 memset(cb->lens + cvectors[book].len, 0, cb->nentries - cvectors[book].len);
315 clens += cvectors[book].len;
317 if (cb->lookup) {
318 vals = cb_lookup_vals(cb->lookup, cb->ndimensions, cb->nentries);
319 cb->quantlist = av_malloc_array(vals, sizeof(int));
320 if (!cb->quantlist)
321 return AVERROR(ENOMEM);
322 for (i = 0; i < vals; i++)
323 cb->quantlist[i] = *quant++;
324 } else {
325 cb->quantlist = NULL;
327 if ((ret = ready_codebook(cb)) < 0)
328 return ret;
331 venc->nfloors = 1;
332 venc->floors = av_mallocz(sizeof(vorbis_enc_floor) * venc->nfloors);
333 if (!venc->floors)
334 return AVERROR(ENOMEM);
336 // just 1 floor
337 fc = &venc->floors[0];
338 fc->partitions = NUM_FLOOR_PARTITIONS;
339 fc->partition_to_class = av_malloc(sizeof(int) * fc->partitions);
340 if (!fc->partition_to_class)
341 return AVERROR(ENOMEM);
342 fc->nclasses = 0;
343 for (i = 0; i < fc->partitions; i++) {
344 static const int a[] = {0, 1, 2, 2, 3, 3, 4, 4};
345 fc->partition_to_class[i] = a[i];
346 fc->nclasses = FFMAX(fc->nclasses, fc->partition_to_class[i]);
348 fc->nclasses++;
349 fc->classes = av_calloc(fc->nclasses, sizeof(vorbis_enc_floor_class));
350 if (!fc->classes)
351 return AVERROR(ENOMEM);
352 for (i = 0; i < fc->nclasses; i++) {
353 vorbis_enc_floor_class * c = &fc->classes[i];
354 int j, books;
355 c->dim = floor_classes[i].dim;
356 c->subclass = floor_classes[i].subclass;
357 c->masterbook = floor_classes[i].masterbook;
358 books = (1 << c->subclass);
359 c->books = av_malloc_array(books, sizeof(int));
360 if (!c->books)
361 return AVERROR(ENOMEM);
362 for (j = 0; j < books; j++)
363 c->books[j] = floor_classes[i].nbooks[j];
365 fc->multiplier = 2;
366 fc->rangebits = venc->log2_blocksize[1] - 1;
368 fc->values = 2;
369 for (i = 0; i < fc->partitions; i++)
370 fc->values += fc->classes[fc->partition_to_class[i]].dim;
372 fc->list = av_malloc_array(fc->values, sizeof(vorbis_floor1_entry));
373 if (!fc->list)
374 return AVERROR(ENOMEM);
375 fc->list[0].x = 0;
376 fc->list[1].x = 1 << fc->rangebits;
377 for (i = 2; i < fc->values; i++) {
378 static const int a[] = {
379 93, 23,372, 6, 46,186,750, 14, 33, 65,
380 130,260,556, 3, 10, 18, 28, 39, 55, 79,
381 111,158,220,312,464,650,850
383 fc->list[i].x = a[i - 2];
385 if (ff_vorbis_ready_floor1_list(avctx, fc->list, fc->values))
386 return AVERROR_BUG;
388 venc->nresidues = 1;
389 venc->residues = av_mallocz(sizeof(vorbis_enc_residue) * venc->nresidues);
390 if (!venc->residues)
391 return AVERROR(ENOMEM);
393 // single residue
394 rc = &venc->residues[0];
395 rc->type = 2;
396 rc->begin = 0;
397 rc->end = 1600;
398 rc->partition_size = 32;
399 rc->classifications = 10;
400 rc->classbook = 15;
401 rc->books = av_malloc(sizeof(*rc->books) * rc->classifications);
402 if (!rc->books)
403 return AVERROR(ENOMEM);
405 static const int8_t a[10][8] = {
406 { -1, -1, -1, -1, -1, -1, -1, -1, },
407 { -1, -1, 16, -1, -1, -1, -1, -1, },
408 { -1, -1, 17, -1, -1, -1, -1, -1, },
409 { -1, -1, 18, -1, -1, -1, -1, -1, },
410 { -1, -1, 19, -1, -1, -1, -1, -1, },
411 { -1, -1, 20, -1, -1, -1, -1, -1, },
412 { -1, -1, 21, -1, -1, -1, -1, -1, },
413 { 22, 23, -1, -1, -1, -1, -1, -1, },
414 { 24, 25, -1, -1, -1, -1, -1, -1, },
415 { 26, 27, 28, -1, -1, -1, -1, -1, },
417 memcpy(rc->books, a, sizeof a);
419 if ((ret = ready_residue(rc, venc)) < 0)
420 return ret;
422 venc->nmappings = 1;
423 venc->mappings = av_mallocz(sizeof(vorbis_enc_mapping) * venc->nmappings);
424 if (!venc->mappings)
425 return AVERROR(ENOMEM);
427 // single mapping
428 mc = &venc->mappings[0];
429 mc->submaps = 1;
430 mc->mux = av_malloc(sizeof(int) * venc->channels);
431 if (!mc->mux)
432 return AVERROR(ENOMEM);
433 for (i = 0; i < venc->channels; i++)
434 mc->mux[i] = 0;
435 mc->floor = av_malloc(sizeof(int) * mc->submaps);
436 mc->residue = av_malloc(sizeof(int) * mc->submaps);
437 if (!mc->floor || !mc->residue)
438 return AVERROR(ENOMEM);
439 for (i = 0; i < mc->submaps; i++) {
440 mc->floor[i] = 0;
441 mc->residue[i] = 0;
443 mc->coupling_steps = venc->channels == 2 ? 1 : 0;
444 mc->magnitude = av_malloc(sizeof(int) * mc->coupling_steps);
445 mc->angle = av_malloc(sizeof(int) * mc->coupling_steps);
446 if (!mc->magnitude || !mc->angle)
447 return AVERROR(ENOMEM);
448 if (mc->coupling_steps) {
449 mc->magnitude[0] = 0;
450 mc->angle[0] = 1;
453 venc->nmodes = 2;
454 venc->modes = av_malloc(sizeof(vorbis_enc_mode) * venc->nmodes);
455 if (!venc->modes)
456 return AVERROR(ENOMEM);
458 // Short block
459 venc->modes[0].blockflag = 0;
460 venc->modes[0].mapping = 0;
461 // Long block
462 venc->modes[1].blockflag = 1;
463 venc->modes[1].mapping = 0;
465 venc->have_saved = 0;
466 venc->saved = av_malloc_array(sizeof(float) * venc->channels, (1 << venc->log2_blocksize[1]) / 2);
467 venc->samples = av_malloc_array(sizeof(float) * venc->channels, (1 << venc->log2_blocksize[1]));
468 venc->floor = av_malloc_array(sizeof(float) * venc->channels, (1 << venc->log2_blocksize[1]) / 2);
469 venc->coeffs = av_malloc_array(sizeof(float) * venc->channels, (1 << venc->log2_blocksize[1]) / 2);
470 venc->scratch = av_malloc_array(sizeof(float) * venc->channels, (1 << venc->log2_blocksize[1]));
472 if (!venc->saved || !venc->samples || !venc->floor || !venc->coeffs || !venc->scratch)
473 return AVERROR(ENOMEM);
475 if ((ret = dsp_init(avctx, venc)) < 0)
476 return ret;
478 return 0;
481 static void put_float(PutBitContext *pb, float f)
483 int exp, mant;
484 uint32_t res = 0;
485 mant = (int)ldexp(frexp(f, &exp), 20);
486 exp += 788 - 20;
487 if (mant < 0) {
488 res |= (1U << 31);
489 mant = -mant;
491 res |= mant | (exp << 21);
492 put_bits32(pb, res);
495 static void put_codebook_header(PutBitContext *pb, vorbis_enc_codebook *cb)
497 int i;
498 int ordered = 0;
500 put_bits(pb, 24, 0x564342); //magic
501 put_bits(pb, 16, cb->ndimensions);
502 put_bits(pb, 24, cb->nentries);
504 for (i = 1; i < cb->nentries; i++)
505 if (cb->lens[i] < cb->lens[i-1])
506 break;
507 if (i == cb->nentries)
508 ordered = 1;
510 put_bits(pb, 1, ordered);
511 if (ordered) {
512 int len = cb->lens[0];
513 put_bits(pb, 5, len - 1);
514 i = 0;
515 while (i < cb->nentries) {
516 int j;
517 for (j = 0; j+i < cb->nentries; j++)
518 if (cb->lens[j+i] != len)
519 break;
520 put_bits(pb, ilog(cb->nentries - i), j);
521 i += j;
522 len++;
524 } else {
525 int sparse = 0;
526 for (i = 0; i < cb->nentries; i++)
527 if (!cb->lens[i])
528 break;
529 if (i != cb->nentries)
530 sparse = 1;
531 put_bits(pb, 1, sparse);
533 for (i = 0; i < cb->nentries; i++) {
534 if (sparse)
535 put_bits(pb, 1, !!cb->lens[i]);
536 if (cb->lens[i])
537 put_bits(pb, 5, cb->lens[i] - 1);
541 put_bits(pb, 4, cb->lookup);
542 if (cb->lookup) {
543 int tmp = cb_lookup_vals(cb->lookup, cb->ndimensions, cb->nentries);
544 int bits = ilog(cb->quantlist[0]);
546 for (i = 1; i < tmp; i++)
547 bits = FFMAX(bits, ilog(cb->quantlist[i]));
549 put_float(pb, cb->min);
550 put_float(pb, cb->delta);
552 put_bits(pb, 4, bits - 1);
553 put_bits(pb, 1, cb->seq_p);
555 for (i = 0; i < tmp; i++)
556 put_bits(pb, bits, cb->quantlist[i]);
560 static void put_floor_header(PutBitContext *pb, vorbis_enc_floor *fc)
562 int i;
564 put_bits(pb, 16, 1); // type, only floor1 is supported
566 put_bits(pb, 5, fc->partitions);
568 for (i = 0; i < fc->partitions; i++)
569 put_bits(pb, 4, fc->partition_to_class[i]);
571 for (i = 0; i < fc->nclasses; i++) {
572 int j, books;
574 put_bits(pb, 3, fc->classes[i].dim - 1);
575 put_bits(pb, 2, fc->classes[i].subclass);
577 if (fc->classes[i].subclass)
578 put_bits(pb, 8, fc->classes[i].masterbook);
580 books = (1 << fc->classes[i].subclass);
582 for (j = 0; j < books; j++)
583 put_bits(pb, 8, fc->classes[i].books[j] + 1);
586 put_bits(pb, 2, fc->multiplier - 1);
587 put_bits(pb, 4, fc->rangebits);
589 for (i = 2; i < fc->values; i++)
590 put_bits(pb, fc->rangebits, fc->list[i].x);
593 static void put_residue_header(PutBitContext *pb, vorbis_enc_residue *rc)
595 int i;
597 put_bits(pb, 16, rc->type);
599 put_bits(pb, 24, rc->begin);
600 put_bits(pb, 24, rc->end);
601 put_bits(pb, 24, rc->partition_size - 1);
602 put_bits(pb, 6, rc->classifications - 1);
603 put_bits(pb, 8, rc->classbook);
605 for (i = 0; i < rc->classifications; i++) {
606 int j, tmp = 0;
607 for (j = 0; j < 8; j++)
608 tmp |= (rc->books[i][j] != -1) << j;
610 put_bits(pb, 3, tmp & 7);
611 put_bits(pb, 1, tmp > 7);
613 if (tmp > 7)
614 put_bits(pb, 5, tmp >> 3);
617 for (i = 0; i < rc->classifications; i++) {
618 int j;
619 for (j = 0; j < 8; j++)
620 if (rc->books[i][j] != -1)
621 put_bits(pb, 8, rc->books[i][j]);
625 static int put_main_header(vorbis_enc_context *venc, uint8_t **out)
627 int i;
628 PutBitContext pb;
629 int len, hlens[3];
630 int buffer_len = 50000;
631 uint8_t *buffer = av_mallocz(buffer_len), *p = buffer;
632 if (!buffer)
633 return AVERROR(ENOMEM);
635 // identification header
636 init_put_bits(&pb, p, buffer_len);
637 put_bits(&pb, 8, 1); //magic
638 for (i = 0; "vorbis"[i]; i++)
639 put_bits(&pb, 8, "vorbis"[i]);
640 put_bits32(&pb, 0); // version
641 put_bits(&pb, 8, venc->channels);
642 put_bits32(&pb, venc->sample_rate);
643 put_bits32(&pb, 0); // bitrate
644 put_bits32(&pb, 0); // bitrate
645 put_bits32(&pb, 0); // bitrate
646 put_bits(&pb, 4, venc->log2_blocksize[0]);
647 put_bits(&pb, 4, venc->log2_blocksize[1]);
648 put_bits(&pb, 1, 1); // framing
650 flush_put_bits(&pb);
651 hlens[0] = put_bytes_output(&pb);
652 buffer_len -= hlens[0];
653 p += hlens[0];
655 // comment header
656 init_put_bits(&pb, p, buffer_len);
657 put_bits(&pb, 8, 3); //magic
658 for (i = 0; "vorbis"[i]; i++)
659 put_bits(&pb, 8, "vorbis"[i]);
660 put_bits32(&pb, 0); // vendor length TODO
661 put_bits32(&pb, 0); // amount of comments
662 put_bits(&pb, 1, 1); // framing
664 flush_put_bits(&pb);
665 hlens[1] = put_bytes_output(&pb);
666 buffer_len -= hlens[1];
667 p += hlens[1];
669 // setup header
670 init_put_bits(&pb, p, buffer_len);
671 put_bits(&pb, 8, 5); //magic
672 for (i = 0; "vorbis"[i]; i++)
673 put_bits(&pb, 8, "vorbis"[i]);
675 // codebooks
676 put_bits(&pb, 8, venc->ncodebooks - 1);
677 for (i = 0; i < venc->ncodebooks; i++)
678 put_codebook_header(&pb, &venc->codebooks[i]);
680 // time domain, reserved, zero
681 put_bits(&pb, 6, 0);
682 put_bits(&pb, 16, 0);
684 // floors
685 put_bits(&pb, 6, venc->nfloors - 1);
686 for (i = 0; i < venc->nfloors; i++)
687 put_floor_header(&pb, &venc->floors[i]);
689 // residues
690 put_bits(&pb, 6, venc->nresidues - 1);
691 for (i = 0; i < venc->nresidues; i++)
692 put_residue_header(&pb, &venc->residues[i]);
694 // mappings
695 put_bits(&pb, 6, venc->nmappings - 1);
696 for (i = 0; i < venc->nmappings; i++) {
697 vorbis_enc_mapping *mc = &venc->mappings[i];
698 int j;
699 put_bits(&pb, 16, 0); // mapping type
701 put_bits(&pb, 1, mc->submaps > 1);
702 if (mc->submaps > 1)
703 put_bits(&pb, 4, mc->submaps - 1);
705 put_bits(&pb, 1, !!mc->coupling_steps);
706 if (mc->coupling_steps) {
707 put_bits(&pb, 8, mc->coupling_steps - 1);
708 for (j = 0; j < mc->coupling_steps; j++) {
709 put_bits(&pb, ilog(venc->channels - 1), mc->magnitude[j]);
710 put_bits(&pb, ilog(venc->channels - 1), mc->angle[j]);
714 put_bits(&pb, 2, 0); // reserved
716 if (mc->submaps > 1)
717 for (j = 0; j < venc->channels; j++)
718 put_bits(&pb, 4, mc->mux[j]);
720 for (j = 0; j < mc->submaps; j++) {
721 put_bits(&pb, 8, 0); // reserved time configuration
722 put_bits(&pb, 8, mc->floor[j]);
723 put_bits(&pb, 8, mc->residue[j]);
727 // modes
728 put_bits(&pb, 6, venc->nmodes - 1);
729 for (i = 0; i < venc->nmodes; i++) {
730 put_bits(&pb, 1, venc->modes[i].blockflag);
731 put_bits(&pb, 16, 0); // reserved window type
732 put_bits(&pb, 16, 0); // reserved transform type
733 put_bits(&pb, 8, venc->modes[i].mapping);
736 put_bits(&pb, 1, 1); // framing
738 flush_put_bits(&pb);
739 hlens[2] = put_bytes_output(&pb);
741 len = hlens[0] + hlens[1] + hlens[2];
742 p = *out = av_mallocz(64 + len + len/255);
743 if (!p)
744 return AVERROR(ENOMEM);
746 *p++ = 2;
747 p += av_xiphlacing(p, hlens[0]);
748 p += av_xiphlacing(p, hlens[1]);
749 buffer_len = 0;
750 for (i = 0; i < 3; i++) {
751 memcpy(p, buffer + buffer_len, hlens[i]);
752 p += hlens[i];
753 buffer_len += hlens[i];
756 av_freep(&buffer);
757 return p - *out;
760 static float get_floor_average(vorbis_enc_floor * fc, float *coeffs, int i)
762 int begin = fc->list[fc->list[FFMAX(i-1, 0)].sort].x;
763 int end = fc->list[fc->list[FFMIN(i+1, fc->values - 1)].sort].x;
764 int j;
765 float average = 0;
767 for (j = begin; j < end; j++)
768 average += fabs(coeffs[j]);
769 return average / (end - begin);
772 static void floor_fit(vorbis_enc_context *venc, vorbis_enc_floor *fc,
773 float *coeffs, uint16_t *posts, int samples)
775 int range = 255 / fc->multiplier + 1;
776 int i;
777 float tot_average = 0.0;
778 float averages[MAX_FLOOR_VALUES];
779 for (i = 0; i < fc->values; i++) {
780 averages[i] = get_floor_average(fc, coeffs, i);
781 tot_average += averages[i];
783 tot_average /= fc->values;
784 tot_average /= venc->quality;
786 for (i = 0; i < fc->values; i++) {
787 int position = fc->list[fc->list[i].sort].x;
788 float average = averages[i];
789 int j;
791 average = sqrt(tot_average * average) * pow(1.25f, position*0.005f); // MAGIC!
792 for (j = 0; j < range - 1; j++)
793 if (ff_vorbis_floor1_inverse_db_table[j * fc->multiplier] > average)
794 break;
795 posts[fc->list[i].sort] = j;
799 static int render_point(int x0, int y0, int x1, int y1, int x)
801 return y0 + (x - x0) * (y1 - y0) / (x1 - x0);
804 static int floor_encode(vorbis_enc_context *venc, vorbis_enc_floor *fc,
805 PutBitContext *pb, uint16_t *posts,
806 float *floor, int samples)
808 int range = 255 / fc->multiplier + 1;
809 int coded[MAX_FLOOR_VALUES]; // first 2 values are unused
810 int i, counter;
812 if (put_bits_left(pb) < 1 + 2 * ilog(range - 1))
813 return AVERROR(EINVAL);
814 put_bits(pb, 1, 1); // non zero
815 put_bits(pb, ilog(range - 1), posts[0]);
816 put_bits(pb, ilog(range - 1), posts[1]);
817 coded[0] = coded[1] = 1;
819 for (i = 2; i < fc->values; i++) {
820 int predicted = render_point(fc->list[fc->list[i].low].x,
821 posts[fc->list[i].low],
822 fc->list[fc->list[i].high].x,
823 posts[fc->list[i].high],
824 fc->list[i].x);
825 int highroom = range - predicted;
826 int lowroom = predicted;
827 int room = FFMIN(highroom, lowroom);
828 if (predicted == posts[i]) {
829 coded[i] = 0; // must be used later as flag!
830 continue;
831 } else {
832 if (!coded[fc->list[i].low ])
833 coded[fc->list[i].low ] = -1;
834 if (!coded[fc->list[i].high])
835 coded[fc->list[i].high] = -1;
837 if (posts[i] > predicted) {
838 if (posts[i] - predicted > room)
839 coded[i] = posts[i] - predicted + lowroom;
840 else
841 coded[i] = (posts[i] - predicted) << 1;
842 } else {
843 if (predicted - posts[i] > room)
844 coded[i] = predicted - posts[i] + highroom - 1;
845 else
846 coded[i] = ((predicted - posts[i]) << 1) - 1;
850 counter = 2;
851 for (i = 0; i < fc->partitions; i++) {
852 vorbis_enc_floor_class * c = &fc->classes[fc->partition_to_class[i]];
853 int k, cval = 0, csub = 1<<c->subclass;
854 if (c->subclass) {
855 vorbis_enc_codebook * book = &venc->codebooks[c->masterbook];
856 int cshift = 0;
857 for (k = 0; k < c->dim; k++) {
858 int l;
859 for (l = 0; l < csub; l++) {
860 int maxval = 1;
861 if (c->books[l] != -1)
862 maxval = venc->codebooks[c->books[l]].nentries;
863 // coded could be -1, but this still works, cause that is 0
864 if (coded[counter + k] < maxval)
865 break;
867 assert(l != csub);
868 cval |= l << cshift;
869 cshift += c->subclass;
871 if (put_codeword(pb, book, cval))
872 return AVERROR(EINVAL);
874 for (k = 0; k < c->dim; k++) {
875 int book = c->books[cval & (csub-1)];
876 int entry = coded[counter++];
877 cval >>= c->subclass;
878 if (book == -1)
879 continue;
880 if (entry == -1)
881 entry = 0;
882 if (put_codeword(pb, &venc->codebooks[book], entry))
883 return AVERROR(EINVAL);
887 ff_vorbis_floor1_render_list(fc->list, fc->values, posts, coded,
888 fc->multiplier, floor, samples);
890 return 0;
893 static float *put_vector(vorbis_enc_codebook *book, PutBitContext *pb,
894 float *num)
896 int i, entry = -1;
897 float distance = FLT_MAX;
898 assert(book->dimensions);
899 for (i = 0; i < book->nentries; i++) {
900 float * vec = book->dimensions + i * book->ndimensions, d = book->pow2[i];
901 int j;
902 if (!book->lens[i])
903 continue;
904 for (j = 0; j < book->ndimensions; j++)
905 d -= vec[j] * num[j];
906 if (distance > d) {
907 entry = i;
908 distance = d;
911 if (put_codeword(pb, book, entry))
912 return NULL;
913 return &book->dimensions[entry * book->ndimensions];
916 static int residue_encode(vorbis_enc_context *venc, vorbis_enc_residue *rc,
917 PutBitContext *pb, float *coeffs, int samples,
918 int real_ch)
920 int pass, i, j, p, k;
921 int psize = rc->partition_size;
922 int partitions = (rc->end - rc->begin) / psize;
923 int channels = (rc->type == 2) ? 1 : real_ch;
924 int classes[MAX_CHANNELS][NUM_RESIDUE_PARTITIONS];
925 int classwords = venc->codebooks[rc->classbook].ndimensions;
927 av_assert0(rc->type == 2);
928 av_assert0(real_ch == 2);
929 for (p = 0; p < partitions; p++) {
930 float max1 = 0.0, max2 = 0.0;
931 int s = rc->begin + p * psize;
932 for (k = s; k < s + psize; k += 2) {
933 max1 = FFMAX(max1, fabs(coeffs[ k / real_ch]));
934 max2 = FFMAX(max2, fabs(coeffs[samples + k / real_ch]));
937 for (i = 0; i < rc->classifications - 1; i++)
938 if (max1 < rc->maxes[i][0] && max2 < rc->maxes[i][1])
939 break;
940 classes[0][p] = i;
943 for (pass = 0; pass < 8; pass++) {
944 p = 0;
945 while (p < partitions) {
946 if (pass == 0)
947 for (j = 0; j < channels; j++) {
948 vorbis_enc_codebook * book = &venc->codebooks[rc->classbook];
949 int entry = 0;
950 for (i = 0; i < classwords; i++) {
951 entry *= rc->classifications;
952 entry += classes[j][p + i];
954 if (put_codeword(pb, book, entry))
955 return AVERROR(EINVAL);
957 for (i = 0; i < classwords && p < partitions; i++, p++) {
958 for (j = 0; j < channels; j++) {
959 int nbook = rc->books[classes[j][p]][pass];
960 vorbis_enc_codebook * book = &venc->codebooks[nbook];
961 float *buf = coeffs + samples*j + rc->begin + p*psize;
962 if (nbook == -1)
963 continue;
965 assert(rc->type == 0 || rc->type == 2);
966 assert(!(psize % book->ndimensions));
968 if (rc->type == 0) {
969 for (k = 0; k < psize; k += book->ndimensions) {
970 int l;
971 float *a = put_vector(book, pb, &buf[k]);
972 if (!a)
973 return AVERROR(EINVAL);
974 for (l = 0; l < book->ndimensions; l++)
975 buf[k + l] -= a[l];
977 } else {
978 int s = rc->begin + p * psize, a1, b1;
979 a1 = (s % real_ch) * samples;
980 b1 = s / real_ch;
981 s = real_ch * samples;
982 for (k = 0; k < psize; k += book->ndimensions) {
983 int dim, a2 = a1, b2 = b1;
984 float vec[MAX_CODEBOOK_DIM], *pv = vec;
985 for (dim = book->ndimensions; dim--; ) {
986 *pv++ = coeffs[a2 + b2];
987 if ((a2 += samples) == s) {
988 a2 = 0;
989 b2++;
992 pv = put_vector(book, pb, vec);
993 if (!pv)
994 return AVERROR(EINVAL);
995 for (dim = book->ndimensions; dim--; ) {
996 coeffs[a1 + b1] -= *pv++;
997 if ((a1 += samples) == s) {
998 a1 = 0;
999 b1++;
1008 return 0;
1011 static int apply_window_and_mdct(vorbis_enc_context *venc)
1013 int channel;
1014 const float * win = venc->win[1];
1015 int window_len = 1 << (venc->log2_blocksize[1] - 1);
1016 float n = (float)(1 << venc->log2_blocksize[1]) / 4.0;
1017 AVFloatDSPContext *fdsp = venc->fdsp;
1019 for (channel = 0; channel < venc->channels; channel++) {
1020 float *offset = venc->samples + channel * window_len * 2;
1022 fdsp->vector_fmul(offset, offset, win, window_len);
1023 fdsp->vector_fmul_scalar(offset, offset, 1/n, window_len);
1025 offset += window_len;
1027 fdsp->vector_fmul_reverse(offset, offset, win, window_len);
1028 fdsp->vector_fmul_scalar(offset, offset, 1/n, window_len);
1030 venc->mdct_fn[1](venc->mdct[1], venc->coeffs + channel * window_len,
1031 venc->samples + channel * window_len * 2, sizeof(float));
1033 return 1;
1036 /* Used for padding the last encoded packet */
1037 static AVFrame *spawn_empty_frame(AVCodecContext *avctx, int channels)
1039 AVFrame *f = av_frame_alloc();
1040 int ch;
1042 if (!f)
1043 return NULL;
1045 f->format = avctx->sample_fmt;
1046 f->nb_samples = avctx->frame_size;
1047 f->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC;
1048 f->ch_layout.nb_channels = channels;
1050 if (av_frame_get_buffer(f, 4)) {
1051 av_frame_free(&f);
1052 return NULL;
1055 for (ch = 0; ch < channels; ch++) {
1056 size_t bps = av_get_bytes_per_sample(f->format);
1057 memset(f->extended_data[ch], 0, bps * f->nb_samples);
1059 return f;
1062 /* Set up audio samples for psy analysis and window/mdct */
1063 static void move_audio(vorbis_enc_context *venc, int sf_size)
1065 AVFrame *cur = NULL;
1066 int frame_size = 1 << (venc->log2_blocksize[1] - 1);
1067 int subframes = frame_size / sf_size;
1068 int sf, ch;
1070 /* Copy samples from last frame into current frame */
1071 if (venc->have_saved)
1072 for (ch = 0; ch < venc->channels; ch++)
1073 memcpy(venc->samples + 2 * ch * frame_size,
1074 venc->saved + ch * frame_size, sizeof(float) * frame_size);
1075 else
1076 for (ch = 0; ch < venc->channels; ch++)
1077 memset(venc->samples + 2 * ch * frame_size, 0, sizeof(float) * frame_size);
1079 for (sf = 0; sf < subframes; sf++) {
1080 cur = ff_bufqueue_get(&venc->bufqueue);
1082 for (ch = 0; ch < venc->channels; ch++) {
1083 float *offset = venc->samples + 2 * ch * frame_size + frame_size;
1084 float *save = venc->saved + ch * frame_size;
1085 const float *input = (float *) cur->extended_data[ch];
1086 const size_t len = cur->nb_samples * sizeof(float);
1088 memcpy(offset + sf*sf_size, input, len);
1089 memcpy(save + sf*sf_size, input, len); // Move samples for next frame
1091 av_frame_free(&cur);
1093 venc->have_saved = 1;
1094 memcpy(venc->scratch, venc->samples, 2 * venc->channels * frame_size);
1097 static int vorbis_encode_frame(AVCodecContext *avctx, AVPacket *avpkt,
1098 const AVFrame *frame, int *got_packet_ptr)
1100 vorbis_enc_context *venc = avctx->priv_data;
1101 int i, ret, need_more;
1102 int frame_size = 1 << (venc->log2_blocksize[1] - 1);
1103 vorbis_enc_mode *mode;
1104 vorbis_enc_mapping *mapping;
1105 PutBitContext pb;
1107 if (frame) {
1108 AVFrame *clone;
1109 if ((ret = ff_af_queue_add(&venc->afq, frame)) < 0)
1110 return ret;
1111 clone = av_frame_clone(frame);
1112 if (!clone)
1113 return AVERROR(ENOMEM);
1114 ff_bufqueue_add(avctx, &venc->bufqueue, clone);
1115 } else
1116 if (!venc->afq.remaining_samples)
1117 return 0;
1119 need_more = venc->bufqueue.available * avctx->frame_size < frame_size;
1120 need_more = frame && need_more;
1121 if (need_more)
1122 return 0;
1124 /* Pad the bufqueue with empty frames for encoding the last packet. */
1125 if (!frame) {
1126 if (venc->bufqueue.available * avctx->frame_size < frame_size) {
1127 int frames_needed = (frame_size/avctx->frame_size) - venc->bufqueue.available;
1128 int i;
1130 for (i = 0; i < frames_needed; i++) {
1131 AVFrame *empty = spawn_empty_frame(avctx, venc->channels);
1132 if (!empty)
1133 return AVERROR(ENOMEM);
1135 ff_bufqueue_add(avctx, &venc->bufqueue, empty);
1140 move_audio(venc, avctx->frame_size);
1142 if (!apply_window_and_mdct(venc))
1143 return 0;
1145 if ((ret = ff_alloc_packet(avctx, avpkt, 8192)) < 0)
1146 return ret;
1148 init_put_bits(&pb, avpkt->data, avpkt->size);
1150 put_bits(&pb, 1, 0); // magic bit
1152 put_bits(&pb, ilog(venc->nmodes - 1), 1); // Mode for current frame
1154 mode = &venc->modes[1];
1155 mapping = &venc->mappings[mode->mapping];
1156 if (mode->blockflag) {
1157 put_bits(&pb, 1, 1); // Previous windowflag
1158 put_bits(&pb, 1, 1); // Next windowflag
1161 for (i = 0; i < venc->channels; i++) {
1162 vorbis_enc_floor *fc = &venc->floors[mapping->floor[mapping->mux[i]]];
1163 uint16_t posts[MAX_FLOOR_VALUES];
1164 floor_fit(venc, fc, &venc->coeffs[i * frame_size], posts, frame_size);
1165 if (floor_encode(venc, fc, &pb, posts, &venc->floor[i * frame_size], frame_size)) {
1166 av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
1167 return AVERROR(EINVAL);
1171 for (i = 0; i < venc->channels * frame_size; i++)
1172 venc->coeffs[i] /= venc->floor[i];
1174 for (i = 0; i < mapping->coupling_steps; i++) {
1175 float *mag = venc->coeffs + mapping->magnitude[i] * frame_size;
1176 float *ang = venc->coeffs + mapping->angle[i] * frame_size;
1177 int j;
1178 for (j = 0; j < frame_size; j++) {
1179 float a = ang[j];
1180 ang[j] -= mag[j];
1181 if (mag[j] > 0)
1182 ang[j] = -ang[j];
1183 if (ang[j] < 0)
1184 mag[j] = a;
1188 if (residue_encode(venc, &venc->residues[mapping->residue[mapping->mux[0]]],
1189 &pb, venc->coeffs, frame_size, venc->channels)) {
1190 av_log(avctx, AV_LOG_ERROR, "output buffer is too small\n");
1191 return AVERROR(EINVAL);
1194 flush_put_bits(&pb);
1195 avpkt->size = put_bytes_output(&pb);
1197 ff_af_queue_remove(&venc->afq, frame_size, &avpkt->pts, &avpkt->duration);
1199 if (frame_size > avpkt->duration) {
1200 uint8_t *side = av_packet_new_side_data(avpkt, AV_PKT_DATA_SKIP_SAMPLES, 10);
1201 if (!side)
1202 return AVERROR(ENOMEM);
1203 AV_WL32(&side[4], frame_size - avpkt->duration);
1206 *got_packet_ptr = 1;
1207 return 0;
1211 static av_cold int vorbis_encode_close(AVCodecContext *avctx)
1213 vorbis_enc_context *venc = avctx->priv_data;
1214 int i;
1216 if (venc->codebooks)
1217 for (i = 0; i < venc->ncodebooks; i++) {
1218 av_freep(&venc->codebooks[i].lens);
1219 av_freep(&venc->codebooks[i].codewords);
1220 av_freep(&venc->codebooks[i].quantlist);
1221 av_freep(&venc->codebooks[i].dimensions);
1222 av_freep(&venc->codebooks[i].pow2);
1224 av_freep(&venc->codebooks);
1226 if (venc->floors)
1227 for (i = 0; i < venc->nfloors; i++) {
1228 int j;
1229 if (venc->floors[i].classes)
1230 for (j = 0; j < venc->floors[i].nclasses; j++)
1231 av_freep(&venc->floors[i].classes[j].books);
1232 av_freep(&venc->floors[i].classes);
1233 av_freep(&venc->floors[i].partition_to_class);
1234 av_freep(&venc->floors[i].list);
1236 av_freep(&venc->floors);
1238 if (venc->residues)
1239 for (i = 0; i < venc->nresidues; i++) {
1240 av_freep(&venc->residues[i].books);
1241 av_freep(&venc->residues[i].maxes);
1243 av_freep(&venc->residues);
1245 if (venc->mappings)
1246 for (i = 0; i < venc->nmappings; i++) {
1247 av_freep(&venc->mappings[i].mux);
1248 av_freep(&venc->mappings[i].floor);
1249 av_freep(&venc->mappings[i].residue);
1250 av_freep(&venc->mappings[i].magnitude);
1251 av_freep(&venc->mappings[i].angle);
1253 av_freep(&venc->mappings);
1255 av_freep(&venc->modes);
1257 av_freep(&venc->saved);
1258 av_freep(&venc->samples);
1259 av_freep(&venc->floor);
1260 av_freep(&venc->coeffs);
1261 av_freep(&venc->scratch);
1262 av_freep(&venc->fdsp);
1264 av_tx_uninit(&venc->mdct[0]);
1265 av_tx_uninit(&venc->mdct[1]);
1266 ff_af_queue_close(&venc->afq);
1267 ff_bufqueue_discard_all(&venc->bufqueue);
1269 return 0 ;
1272 static av_cold int vorbis_encode_init(AVCodecContext *avctx)
1274 vorbis_enc_context *venc = avctx->priv_data;
1275 int ret;
1277 if (avctx->ch_layout.nb_channels != 2) {
1278 av_log(avctx, AV_LOG_ERROR, "Current FFmpeg Vorbis encoder only supports 2 channels.\n");
1279 return -1;
1282 if ((ret = create_vorbis_context(venc, avctx)) < 0)
1283 return ret;
1285 avctx->bit_rate = 0;
1286 if (avctx->flags & AV_CODEC_FLAG_QSCALE)
1287 venc->quality = avctx->global_quality / (float)FF_QP2LAMBDA;
1288 else
1289 venc->quality = 8;
1290 venc->quality *= venc->quality;
1292 if ((ret = put_main_header(venc, (uint8_t**)&avctx->extradata)) < 0)
1293 return ret;
1294 avctx->extradata_size = ret;
1296 avctx->frame_size = 64;
1297 avctx->initial_padding = 1 << (venc->log2_blocksize[1] - 1);
1299 ff_af_queue_init(avctx, &venc->afq);
1301 return 0;
1304 const FFCodec ff_vorbis_encoder = {
1305 .p.name = "vorbis",
1306 CODEC_LONG_NAME("Vorbis"),
1307 .p.type = AVMEDIA_TYPE_AUDIO,
1308 .p.id = AV_CODEC_ID_VORBIS,
1309 .p.capabilities = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_DELAY |
1310 AV_CODEC_CAP_EXPERIMENTAL,
1311 .priv_data_size = sizeof(vorbis_enc_context),
1312 .init = vorbis_encode_init,
1313 FF_CODEC_ENCODE_CB(vorbis_encode_frame),
1314 .close = vorbis_encode_close,
1315 .p.sample_fmts = (const enum AVSampleFormat[]){ AV_SAMPLE_FMT_FLTP,
1316 AV_SAMPLE_FMT_NONE },
1317 .caps_internal = FF_CODEC_CAP_INIT_CLEANUP,