Statistics
| Branch: | Revision:

ffmpeg / libavcodec / aacenc.c @ 5d6e4c16

History | View | Annotate | Download (22.5 KB)

1
/*
2
 * AAC encoder
3
 * Copyright (C) 2008 Konstantin Shishkov
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
/**
23
 * @file
24
 * AAC encoder
25
 */
26

    
27
/***********************************
28
 *              TODOs:
29
 * add sane pulse detection
30
 * add temporal noise shaping
31
 ***********************************/
32

    
33
#include "avcodec.h"
34
#include "put_bits.h"
35
#include "dsputil.h"
36
#include "mpeg4audio.h"
37

    
38
#include "aac.h"
39
#include "aactab.h"
40
#include "aacenc.h"
41

    
42
#include "psymodel.h"
43

    
44
#define AAC_MAX_CHANNELS 6
45

    
46
static const uint8_t swb_size_1024_96[] = {
47
    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8,
48
    12, 12, 12, 12, 12, 16, 16, 24, 28, 36, 44,
49
    64, 64, 64, 64, 64, 64, 64, 64, 64, 64, 64
50
};
51

    
52
static const uint8_t swb_size_1024_64[] = {
53
    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8,
54
    12, 12, 12, 16, 16, 16, 20, 24, 24, 28, 36,
55
    40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40, 40
56
};
57

    
58
static const uint8_t swb_size_1024_48[] = {
59
    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
60
    12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
61
    32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32,
62
    96
63
};
64

    
65
static const uint8_t swb_size_1024_32[] = {
66
    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8,
67
    12, 12, 12, 12, 16, 16, 20, 20, 24, 24, 28, 28,
68
    32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32, 32
69
};
70

    
71
static const uint8_t swb_size_1024_24[] = {
72
    4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
73
    12, 12, 12, 12, 16, 16, 16, 20, 20, 24, 24, 28, 28,
74
    32, 36, 36, 40, 44, 48, 52, 52, 64, 64, 64, 64, 64
75
};
76

    
77
static const uint8_t swb_size_1024_16[] = {
78
    8, 8, 8, 8, 8, 8, 8, 8, 8, 8, 8,
79
    12, 12, 12, 12, 12, 12, 12, 12, 12, 16, 16, 16, 16, 20, 20, 20, 24, 24, 28, 28,
80
    32, 36, 40, 40, 44, 48, 52, 56, 60, 64, 64, 64
81
};
82

    
83
static const uint8_t swb_size_1024_8[] = {
84
    12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12, 12,
85
    16, 16, 16, 16, 16, 16, 16, 20, 20, 20, 20, 24, 24, 24, 28, 28,
86
    32, 36, 36, 40, 44, 48, 52, 56, 60, 64, 80
87
};
88

    
89
static const uint8_t *swb_size_1024[] = {
90
    swb_size_1024_96, swb_size_1024_96, swb_size_1024_64,
91
    swb_size_1024_48, swb_size_1024_48, swb_size_1024_32,
92
    swb_size_1024_24, swb_size_1024_24, swb_size_1024_16,
93
    swb_size_1024_16, swb_size_1024_16, swb_size_1024_8
94
};
95

    
96
static const uint8_t swb_size_128_96[] = {
97
    4, 4, 4, 4, 4, 4, 8, 8, 8, 16, 28, 36
98
};
99

    
100
static const uint8_t swb_size_128_48[] = {
101
    4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 12, 16, 16, 16
102
};
103

    
104
static const uint8_t swb_size_128_24[] = {
105
    4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 12, 12, 16, 16, 20
106
};
107

    
108
static const uint8_t swb_size_128_16[] = {
109
    4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 12, 12, 16, 20, 20
110
};
111

    
112
static const uint8_t swb_size_128_8[] = {
113
    4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 12, 16, 20, 20
114
};
115

    
116
static const uint8_t *swb_size_128[] = {
117
    /* the last entry on the following row is swb_size_128_64 but is a
118
       duplicate of swb_size_128_96 */
119
    swb_size_128_96, swb_size_128_96, swb_size_128_96,
120
    swb_size_128_48, swb_size_128_48, swb_size_128_48,
121
    swb_size_128_24, swb_size_128_24, swb_size_128_16,
122
    swb_size_128_16, swb_size_128_16, swb_size_128_8
123
};
124

    
125
/** default channel configurations */
126
static const uint8_t aac_chan_configs[6][5] = {
127
 {1, TYPE_SCE},                               // 1 channel  - single channel element
128
 {1, TYPE_CPE},                               // 2 channels - channel pair
129
 {2, TYPE_SCE, TYPE_CPE},                     // 3 channels - center + stereo
130
 {3, TYPE_SCE, TYPE_CPE, TYPE_SCE},           // 4 channels - front center + stereo + back center
131
 {3, TYPE_SCE, TYPE_CPE, TYPE_CPE},           // 5 channels - front center + stereo + back stereo
132
 {4, TYPE_SCE, TYPE_CPE, TYPE_CPE, TYPE_LFE}, // 6 channels - front center + stereo + back stereo + LFE
133
};
134

    
135
/**
136
 * Make AAC audio config object.
137
 * @see 1.6.2.1 "Syntax - AudioSpecificConfig"
138
 */
139
static void put_audio_specific_config(AVCodecContext *avctx)
140
{
141
    PutBitContext pb;
142
    AACEncContext *s = avctx->priv_data;
143

    
144
    init_put_bits(&pb, avctx->extradata, avctx->extradata_size*8);
145
    put_bits(&pb, 5, 2); //object type - AAC-LC
146
    put_bits(&pb, 4, s->samplerate_index); //sample rate index
147
    put_bits(&pb, 4, avctx->channels);
148
    //GASpecificConfig
149
    put_bits(&pb, 1, 0); //frame length - 1024 samples
150
    put_bits(&pb, 1, 0); //does not depend on core coder
151
    put_bits(&pb, 1, 0); //is not extension
152
    flush_put_bits(&pb);
153
}
154

    
155
static av_cold int aac_encode_init(AVCodecContext *avctx)
156
{
157
    AACEncContext *s = avctx->priv_data;
158
    int i;
159
    const uint8_t *sizes[2];
160
    int lengths[2];
161

    
162
    avctx->frame_size = 1024;
163

    
164
    for (i = 0; i < 16; i++)
165
        if (avctx->sample_rate == ff_mpeg4audio_sample_rates[i])
166
            break;
167
    if (i == 16) {
168
        av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
169
        return -1;
170
    }
171
    if (avctx->channels > AAC_MAX_CHANNELS) {
172
        av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
173
        return -1;
174
    }
175
    if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) {
176
        av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile);
177
        return -1;
178
    }
179
    if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) {
180
        av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n");
181
        return -1;
182
    }
183
    s->samplerate_index = i;
184

    
185
    dsputil_init(&s->dsp, avctx);
186
    ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
187
    ff_mdct_init(&s->mdct128,   8, 0, 1.0);
188
    // window init
189
    ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
190
    ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
191
    ff_init_ff_sine_windows(10);
192
    ff_init_ff_sine_windows(7);
193

    
194
    s->samples            = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
195
    s->cpe                = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]);
196
    avctx->extradata      = av_mallocz(2 + FF_INPUT_BUFFER_PADDING_SIZE);
197
    avctx->extradata_size = 2;
198
    put_audio_specific_config(avctx);
199

    
200
    sizes[0]   = swb_size_1024[i];
201
    sizes[1]   = swb_size_128[i];
202
    lengths[0] = ff_aac_num_swb_1024[i];
203
    lengths[1] = ff_aac_num_swb_128[i];
204
    ff_psy_init(&s->psy, avctx, 2, sizes, lengths);
205
    s->psypp = ff_psy_preprocess_init(avctx);
206
    s->coder = &ff_aac_coders[2];
207

    
208
    s->lambda = avctx->global_quality ? avctx->global_quality : 120;
209

    
210
    ff_aac_tableinit();
211

    
212
    return 0;
213
}
214

    
215
static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
216
                                  SingleChannelElement *sce, short *audio)
217
{
218
    int i, k;
219
    const int chans = avctx->channels;
220
    const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
221
    const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
222
    const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
223

    
224
    if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
225
        memcpy(s->output, sce->saved, sizeof(float)*1024);
226
        if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
227
            memset(s->output, 0, sizeof(s->output[0]) * 448);
228
            for (i = 448; i < 576; i++)
229
                s->output[i] = sce->saved[i] * pwindow[i - 448];
230
            for (i = 576; i < 704; i++)
231
                s->output[i] = sce->saved[i];
232
        }
233
        if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
234
            for (i = 0; i < 1024; i++) {
235
                s->output[i+1024]         = audio[i * chans] * lwindow[1024 - i - 1];
236
                sce->saved[i] = audio[i * chans] * lwindow[i];
237
            }
238
        } else {
239
            for (i = 0; i < 448; i++)
240
                s->output[i+1024]         = audio[i * chans];
241
            for (; i < 576; i++)
242
                s->output[i+1024]         = audio[i * chans] * swindow[576 - i - 1];
243
            memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448);
244
            for (i = 0; i < 1024; i++)
245
                sce->saved[i] = audio[i * chans];
246
        }
247
        ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output);
248
    } else {
249
        for (k = 0; k < 1024; k += 128) {
250
            for (i = 448 + k; i < 448 + k + 256; i++)
251
                s->output[i - 448 - k] = (i < 1024)
252
                                         ? sce->saved[i]
253
                                         : audio[(i-1024)*chans];
254
            s->dsp.vector_fmul        (s->output,     k ?  swindow : pwindow, 128);
255
            s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128);
256
            ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output);
257
        }
258
        for (i = 0; i < 1024; i++)
259
            sce->saved[i] = audio[i * chans];
260
    }
261
}
262

    
263
/**
264
 * Encode ics_info element.
265
 * @see Table 4.6 (syntax of ics_info)
266
 */
267
static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
268
{
269
    int w;
270

    
271
    put_bits(&s->pb, 1, 0);                // ics_reserved bit
272
    put_bits(&s->pb, 2, info->window_sequence[0]);
273
    put_bits(&s->pb, 1, info->use_kb_window[0]);
274
    if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
275
        put_bits(&s->pb, 6, info->max_sfb);
276
        put_bits(&s->pb, 1, 0);            // no prediction
277
    } else {
278
        put_bits(&s->pb, 4, info->max_sfb);
279
        for (w = 1; w < 8; w++)
280
            put_bits(&s->pb, 1, !info->group_len[w]);
281
    }
282
}
283

    
284
/**
285
 * Encode MS data.
286
 * @see 4.6.8.1 "Joint Coding - M/S Stereo"
287
 */
288
static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
289
{
290
    int i, w;
291

    
292
    put_bits(pb, 2, cpe->ms_mode);
293
    if (cpe->ms_mode == 1)
294
        for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
295
            for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
296
                put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
297
}
298

    
299
/**
300
 * Produce integer coefficients from scalefactors provided by the model.
301
 */
302
static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans)
303
{
304
    int i, w, w2, g, ch;
305
    int start, maxsfb, cmaxsfb;
306

    
307
    for (ch = 0; ch < chans; ch++) {
308
        IndividualChannelStream *ics = &cpe->ch[ch].ics;
309
        start = 0;
310
        maxsfb = 0;
311
        cpe->ch[ch].pulse.num_pulse = 0;
312
        for (w = 0; w < ics->num_windows*16; w += 16) {
313
            for (g = 0; g < ics->num_swb; g++) {
314
                //apply M/S
315
                if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
316
                    for (i = 0; i < ics->swb_sizes[g]; i++) {
317
                        cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
318
                        cpe->ch[1].coeffs[start+i] =  cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
319
                    }
320
                }
321
                start += ics->swb_sizes[g];
322
            }
323
            for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
324
                ;
325
            maxsfb = FFMAX(maxsfb, cmaxsfb);
326
        }
327
        ics->max_sfb = maxsfb;
328

    
329
        //adjust zero bands for window groups
330
        for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
331
            for (g = 0; g < ics->max_sfb; g++) {
332
                i = 1;
333
                for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
334
                    if (!cpe->ch[ch].zeroes[w2*16 + g]) {
335
                        i = 0;
336
                        break;
337
                    }
338
                }
339
                cpe->ch[ch].zeroes[w*16 + g] = i;
340
            }
341
        }
342
    }
343

    
344
    if (chans > 1 && cpe->common_window) {
345
        IndividualChannelStream *ics0 = &cpe->ch[0].ics;
346
        IndividualChannelStream *ics1 = &cpe->ch[1].ics;
347
        int msc = 0;
348
        ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
349
        ics1->max_sfb = ics0->max_sfb;
350
        for (w = 0; w < ics0->num_windows*16; w += 16)
351
            for (i = 0; i < ics0->max_sfb; i++)
352
                if (cpe->ms_mask[w+i])
353
                    msc++;
354
        if (msc == 0 || ics0->max_sfb == 0)
355
            cpe->ms_mode = 0;
356
        else
357
            cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2;
358
    }
359
}
360

    
361
/**
362
 * Encode scalefactor band coding type.
363
 */
364
static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
365
{
366
    int w;
367

    
368
    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
369
        s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
370
}
371

    
372
/**
373
 * Encode scalefactors.
374
 */
375
static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
376
                                 SingleChannelElement *sce)
377
{
378
    int off = sce->sf_idx[0], diff;
379
    int i, w;
380

    
381
    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
382
        for (i = 0; i < sce->ics.max_sfb; i++) {
383
            if (!sce->zeroes[w*16 + i]) {
384
                diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
385
                if (diff < 0 || diff > 120)
386
                    av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
387
                off = sce->sf_idx[w*16 + i];
388
                put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
389
            }
390
        }
391
    }
392
}
393

    
394
/**
395
 * Encode pulse data.
396
 */
397
static void encode_pulses(AACEncContext *s, Pulse *pulse)
398
{
399
    int i;
400

    
401
    put_bits(&s->pb, 1, !!pulse->num_pulse);
402
    if (!pulse->num_pulse)
403
        return;
404

    
405
    put_bits(&s->pb, 2, pulse->num_pulse - 1);
406
    put_bits(&s->pb, 6, pulse->start);
407
    for (i = 0; i < pulse->num_pulse; i++) {
408
        put_bits(&s->pb, 5, pulse->pos[i]);
409
        put_bits(&s->pb, 4, pulse->amp[i]);
410
    }
411
}
412

    
413
/**
414
 * Encode spectral coefficients processed by psychoacoustic model.
415
 */
416
static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
417
{
418
    int start, i, w, w2;
419

    
420
    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
421
        start = 0;
422
        for (i = 0; i < sce->ics.max_sfb; i++) {
423
            if (sce->zeroes[w*16 + i]) {
424
                start += sce->ics.swb_sizes[i];
425
                continue;
426
            }
427
            for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
428
                s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
429
                                                   sce->ics.swb_sizes[i],
430
                                                   sce->sf_idx[w*16 + i],
431
                                                   sce->band_type[w*16 + i],
432
                                                   s->lambda);
433
            start += sce->ics.swb_sizes[i];
434
        }
435
    }
436
}
437

    
438
/**
439
 * Encode one channel of audio data.
440
 */
441
static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
442
                                     SingleChannelElement *sce,
443
                                     int common_window)
444
{
445
    put_bits(&s->pb, 8, sce->sf_idx[0]);
446
    if (!common_window)
447
        put_ics_info(s, &sce->ics);
448
    encode_band_info(s, sce);
449
    encode_scale_factors(avctx, s, sce);
450
    encode_pulses(s, &sce->pulse);
451
    put_bits(&s->pb, 1, 0); //tns
452
    put_bits(&s->pb, 1, 0); //ssr
453
    encode_spectral_coeffs(s, sce);
454
    return 0;
455
}
456

    
457
/**
458
 * Write some auxiliary information about the created AAC file.
459
 */
460
static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
461
                               const char *name)
462
{
463
    int i, namelen, padbits;
464

    
465
    namelen = strlen(name) + 2;
466
    put_bits(&s->pb, 3, TYPE_FIL);
467
    put_bits(&s->pb, 4, FFMIN(namelen, 15));
468
    if (namelen >= 15)
469
        put_bits(&s->pb, 8, namelen - 16);
470
    put_bits(&s->pb, 4, 0); //extension type - filler
471
    padbits = 8 - (put_bits_count(&s->pb) & 7);
472
    align_put_bits(&s->pb);
473
    for (i = 0; i < namelen - 2; i++)
474
        put_bits(&s->pb, 8, name[i]);
475
    put_bits(&s->pb, 12 - padbits, 0);
476
}
477

    
478
static int aac_encode_frame(AVCodecContext *avctx,
479
                            uint8_t *frame, int buf_size, void *data)
480
{
481
    AACEncContext *s = avctx->priv_data;
482
    int16_t *samples = s->samples, *samples2, *la;
483
    ChannelElement *cpe;
484
    int i, j, chans, tag, start_ch;
485
    const uint8_t *chan_map = aac_chan_configs[avctx->channels-1];
486
    int chan_el_counter[4];
487
    FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
488

    
489
    if (s->last_frame)
490
        return 0;
491
    if (data) {
492
        if (!s->psypp) {
493
            memcpy(s->samples + 1024 * avctx->channels, data,
494
                   1024 * avctx->channels * sizeof(s->samples[0]));
495
        } else {
496
            start_ch = 0;
497
            samples2 = s->samples + 1024 * avctx->channels;
498
            for (i = 0; i < chan_map[0]; i++) {
499
                tag = chan_map[i+1];
500
                chans = tag == TYPE_CPE ? 2 : 1;
501
                ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch,
502
                                  samples2 + start_ch, start_ch, chans);
503
                start_ch += chans;
504
            }
505
        }
506
    }
507
    if (!avctx->frame_number) {
508
        memcpy(s->samples, s->samples + 1024 * avctx->channels,
509
               1024 * avctx->channels * sizeof(s->samples[0]));
510
        return 0;
511
    }
512

    
513
    start_ch = 0;
514
    for (i = 0; i < chan_map[0]; i++) {
515
        FFPsyWindowInfo* wi = windows + start_ch;
516
        tag      = chan_map[i+1];
517
        chans    = tag == TYPE_CPE ? 2 : 1;
518
        cpe      = &s->cpe[i];
519
        for (j = 0; j < chans; j++) {
520
            IndividualChannelStream *ics = &cpe->ch[j].ics;
521
            int k;
522
            int cur_channel = start_ch + j;
523
            samples2 = samples + cur_channel;
524
            la       = samples2 + (448+64) * avctx->channels;
525
            if (!data)
526
                la = NULL;
527
            if (tag == TYPE_LFE) {
528
                wi[j].window_type[0] = ONLY_LONG_SEQUENCE;
529
                wi[j].window_shape   = 0;
530
                wi[j].num_windows    = 1;
531
                wi[j].grouping[0]    = 1;
532
            } else {
533
                wi[j] = ff_psy_suggest_window(&s->psy, samples2, la, cur_channel,
534
                                              ics->window_sequence[0]);
535
            }
536
            ics->window_sequence[1] = ics->window_sequence[0];
537
            ics->window_sequence[0] = wi[j].window_type[0];
538
            ics->use_kb_window[1]   = ics->use_kb_window[0];
539
            ics->use_kb_window[0]   = wi[j].window_shape;
540
            ics->num_windows        = wi[j].num_windows;
541
            ics->swb_sizes          = s->psy.bands    [ics->num_windows == 8];
542
            ics->num_swb            = tag == TYPE_LFE ? 12 : s->psy.num_bands[ics->num_windows == 8];
543
            for (k = 0; k < ics->num_windows; k++)
544
                ics->group_len[k] = wi[j].grouping[k];
545

    
546
            apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2);
547
        }
548
        start_ch += chans;
549
    }
550
    do {
551
        int frame_bits;
552
        init_put_bits(&s->pb, frame, buf_size*8);
553
        if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT))
554
            put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
555
        start_ch = 0;
556
        memset(chan_el_counter, 0, sizeof(chan_el_counter));
557
        for (i = 0; i < chan_map[0]; i++) {
558
            FFPsyWindowInfo* wi = windows + start_ch;
559
            tag      = chan_map[i+1];
560
            chans    = tag == TYPE_CPE ? 2 : 1;
561
            cpe      = &s->cpe[i];
562
            put_bits(&s->pb, 3, tag);
563
            put_bits(&s->pb, 4, chan_el_counter[tag]++);
564
            for (j = 0; j < chans; j++) {
565
                s->cur_channel = start_ch + j;
566
                ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]);
567
                s->coder->search_for_quantizers(avctx, s, &cpe->ch[j], s->lambda);
568
            }
569
            cpe->common_window = 0;
570
            if (chans > 1
571
                && wi[0].window_type[0] == wi[1].window_type[0]
572
                && wi[0].window_shape   == wi[1].window_shape) {
573

    
574
                cpe->common_window = 1;
575
                for (j = 0; j < wi[0].num_windows; j++) {
576
                    if (wi[0].grouping[j] != wi[1].grouping[j]) {
577
                        cpe->common_window = 0;
578
                        break;
579
                    }
580
                }
581
            }
582
            s->cur_channel = start_ch;
583
            if (cpe->common_window && s->coder->search_for_ms)
584
                s->coder->search_for_ms(s, cpe, s->lambda);
585
            adjust_frame_information(s, cpe, chans);
586
            if (chans == 2) {
587
                put_bits(&s->pb, 1, cpe->common_window);
588
                if (cpe->common_window) {
589
                    put_ics_info(s, &cpe->ch[0].ics);
590
                    encode_ms_info(&s->pb, cpe);
591
                }
592
            }
593
            for (j = 0; j < chans; j++) {
594
                s->cur_channel = start_ch + j;
595
                encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window);
596
            }
597
            start_ch += chans;
598
        }
599

    
600
        frame_bits = put_bits_count(&s->pb);
601
        if (frame_bits <= 6144 * avctx->channels - 3)
602
            break;
603

    
604
        s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
605

    
606
    } while (1);
607

    
608
    put_bits(&s->pb, 3, TYPE_END);
609
    flush_put_bits(&s->pb);
610
    avctx->frame_bits = put_bits_count(&s->pb);
611

    
612
    // rate control stuff
613
    if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
614
        float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
615
        s->lambda *= ratio;
616
        s->lambda = FFMIN(s->lambda, 65536.f);
617
    }
618

    
619
    if (!data)
620
        s->last_frame = 1;
621
    memcpy(s->samples, s->samples + 1024 * avctx->channels,
622
           1024 * avctx->channels * sizeof(s->samples[0]));
623
    return put_bits_count(&s->pb)>>3;
624
}
625

    
626
static av_cold int aac_encode_end(AVCodecContext *avctx)
627
{
628
    AACEncContext *s = avctx->priv_data;
629

    
630
    ff_mdct_end(&s->mdct1024);
631
    ff_mdct_end(&s->mdct128);
632
    ff_psy_end(&s->psy);
633
    ff_psy_preprocess_end(s->psypp);
634
    av_freep(&s->samples);
635
    av_freep(&s->cpe);
636
    return 0;
637
}
638

    
639
AVCodec aac_encoder = {
640
    "aac",
641
    AVMEDIA_TYPE_AUDIO,
642
    CODEC_ID_AAC,
643
    sizeof(AACEncContext),
644
    aac_encode_init,
645
    aac_encode_frame,
646
    aac_encode_end,
647
    .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
648
    .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
649
    .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
650
};