Statistics
| Branch: | Revision:

ffmpeg / libavcodec / aacenc.c @ 015f9f1a

History | View | Annotate | Download (22.7 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

    
153
    //Explicitly Mark SBR absent
154
    put_bits(&pb, 11, 0x27b); //sync extension
155
    put_bits(&pb, 5,  AOT_SBR);
156
    put_bits(&pb, 1,  0);
157
    flush_put_bits(&pb);
158
}
159

    
160
static av_cold int aac_encode_init(AVCodecContext *avctx)
161
{
162
    AACEncContext *s = avctx->priv_data;
163
    int i;
164
    const uint8_t *sizes[2];
165
    int lengths[2];
166

    
167
    avctx->frame_size = 1024;
168

    
169
    for (i = 0; i < 16; i++)
170
        if (avctx->sample_rate == ff_mpeg4audio_sample_rates[i])
171
            break;
172
    if (i == 16) {
173
        av_log(avctx, AV_LOG_ERROR, "Unsupported sample rate %d\n", avctx->sample_rate);
174
        return -1;
175
    }
176
    if (avctx->channels > AAC_MAX_CHANNELS) {
177
        av_log(avctx, AV_LOG_ERROR, "Unsupported number of channels: %d\n", avctx->channels);
178
        return -1;
179
    }
180
    if (avctx->profile != FF_PROFILE_UNKNOWN && avctx->profile != FF_PROFILE_AAC_LOW) {
181
        av_log(avctx, AV_LOG_ERROR, "Unsupported profile %d\n", avctx->profile);
182
        return -1;
183
    }
184
    if (1024.0 * avctx->bit_rate / avctx->sample_rate > 6144 * avctx->channels) {
185
        av_log(avctx, AV_LOG_ERROR, "Too many bits per frame requested\n");
186
        return -1;
187
    }
188
    s->samplerate_index = i;
189

    
190
    dsputil_init(&s->dsp, avctx);
191
    ff_mdct_init(&s->mdct1024, 11, 0, 1.0);
192
    ff_mdct_init(&s->mdct128,   8, 0, 1.0);
193
    // window init
194
    ff_kbd_window_init(ff_aac_kbd_long_1024, 4.0, 1024);
195
    ff_kbd_window_init(ff_aac_kbd_short_128, 6.0, 128);
196
    ff_init_ff_sine_windows(10);
197
    ff_init_ff_sine_windows(7);
198

    
199
    s->samples            = av_malloc(2 * 1024 * avctx->channels * sizeof(s->samples[0]));
200
    s->cpe                = av_mallocz(sizeof(ChannelElement) * aac_chan_configs[avctx->channels-1][0]);
201
    avctx->extradata      = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
202
    avctx->extradata_size = 5;
203
    put_audio_specific_config(avctx);
204

    
205
    sizes[0]   = swb_size_1024[i];
206
    sizes[1]   = swb_size_128[i];
207
    lengths[0] = ff_aac_num_swb_1024[i];
208
    lengths[1] = ff_aac_num_swb_128[i];
209
    ff_psy_init(&s->psy, avctx, 2, sizes, lengths);
210
    s->psypp = ff_psy_preprocess_init(avctx);
211
    s->coder = &ff_aac_coders[2];
212

    
213
    s->lambda = avctx->global_quality ? avctx->global_quality : 120;
214

    
215
    ff_aac_tableinit();
216

    
217
    return 0;
218
}
219

    
220
static void apply_window_and_mdct(AVCodecContext *avctx, AACEncContext *s,
221
                                  SingleChannelElement *sce, short *audio)
222
{
223
    int i, k;
224
    const int chans = avctx->channels;
225
    const float * lwindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_long_1024 : ff_sine_1024;
226
    const float * swindow = sce->ics.use_kb_window[0] ? ff_aac_kbd_short_128 : ff_sine_128;
227
    const float * pwindow = sce->ics.use_kb_window[1] ? ff_aac_kbd_short_128 : ff_sine_128;
228

    
229
    if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
230
        memcpy(s->output, sce->saved, sizeof(float)*1024);
231
        if (sce->ics.window_sequence[0] == LONG_STOP_SEQUENCE) {
232
            memset(s->output, 0, sizeof(s->output[0]) * 448);
233
            for (i = 448; i < 576; i++)
234
                s->output[i] = sce->saved[i] * pwindow[i - 448];
235
            for (i = 576; i < 704; i++)
236
                s->output[i] = sce->saved[i];
237
        }
238
        if (sce->ics.window_sequence[0] != LONG_START_SEQUENCE) {
239
            for (i = 0; i < 1024; i++) {
240
                s->output[i+1024]         = audio[i * chans] * lwindow[1024 - i - 1];
241
                sce->saved[i] = audio[i * chans] * lwindow[i];
242
            }
243
        } else {
244
            for (i = 0; i < 448; i++)
245
                s->output[i+1024]         = audio[i * chans];
246
            for (; i < 576; i++)
247
                s->output[i+1024]         = audio[i * chans] * swindow[576 - i - 1];
248
            memset(s->output+1024+576, 0, sizeof(s->output[0]) * 448);
249
            for (i = 0; i < 1024; i++)
250
                sce->saved[i] = audio[i * chans];
251
        }
252
        ff_mdct_calc(&s->mdct1024, sce->coeffs, s->output);
253
    } else {
254
        for (k = 0; k < 1024; k += 128) {
255
            for (i = 448 + k; i < 448 + k + 256; i++)
256
                s->output[i - 448 - k] = (i < 1024)
257
                                         ? sce->saved[i]
258
                                         : audio[(i-1024)*chans];
259
            s->dsp.vector_fmul        (s->output,     s->output, k ?  swindow : pwindow, 128);
260
            s->dsp.vector_fmul_reverse(s->output+128, s->output+128, swindow, 128);
261
            ff_mdct_calc(&s->mdct128, sce->coeffs + k, s->output);
262
        }
263
        for (i = 0; i < 1024; i++)
264
            sce->saved[i] = audio[i * chans];
265
    }
266
}
267

    
268
/**
269
 * Encode ics_info element.
270
 * @see Table 4.6 (syntax of ics_info)
271
 */
272
static void put_ics_info(AACEncContext *s, IndividualChannelStream *info)
273
{
274
    int w;
275

    
276
    put_bits(&s->pb, 1, 0);                // ics_reserved bit
277
    put_bits(&s->pb, 2, info->window_sequence[0]);
278
    put_bits(&s->pb, 1, info->use_kb_window[0]);
279
    if (info->window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
280
        put_bits(&s->pb, 6, info->max_sfb);
281
        put_bits(&s->pb, 1, 0);            // no prediction
282
    } else {
283
        put_bits(&s->pb, 4, info->max_sfb);
284
        for (w = 1; w < 8; w++)
285
            put_bits(&s->pb, 1, !info->group_len[w]);
286
    }
287
}
288

    
289
/**
290
 * Encode MS data.
291
 * @see 4.6.8.1 "Joint Coding - M/S Stereo"
292
 */
293
static void encode_ms_info(PutBitContext *pb, ChannelElement *cpe)
294
{
295
    int i, w;
296

    
297
    put_bits(pb, 2, cpe->ms_mode);
298
    if (cpe->ms_mode == 1)
299
        for (w = 0; w < cpe->ch[0].ics.num_windows; w += cpe->ch[0].ics.group_len[w])
300
            for (i = 0; i < cpe->ch[0].ics.max_sfb; i++)
301
                put_bits(pb, 1, cpe->ms_mask[w*16 + i]);
302
}
303

    
304
/**
305
 * Produce integer coefficients from scalefactors provided by the model.
306
 */
307
static void adjust_frame_information(AACEncContext *apc, ChannelElement *cpe, int chans)
308
{
309
    int i, w, w2, g, ch;
310
    int start, maxsfb, cmaxsfb;
311

    
312
    for (ch = 0; ch < chans; ch++) {
313
        IndividualChannelStream *ics = &cpe->ch[ch].ics;
314
        start = 0;
315
        maxsfb = 0;
316
        cpe->ch[ch].pulse.num_pulse = 0;
317
        for (w = 0; w < ics->num_windows*16; w += 16) {
318
            for (g = 0; g < ics->num_swb; g++) {
319
                //apply M/S
320
                if (cpe->common_window && !ch && cpe->ms_mask[w + g]) {
321
                    for (i = 0; i < ics->swb_sizes[g]; i++) {
322
                        cpe->ch[0].coeffs[start+i] = (cpe->ch[0].coeffs[start+i] + cpe->ch[1].coeffs[start+i]) / 2.0;
323
                        cpe->ch[1].coeffs[start+i] =  cpe->ch[0].coeffs[start+i] - cpe->ch[1].coeffs[start+i];
324
                    }
325
                }
326
                start += ics->swb_sizes[g];
327
            }
328
            for (cmaxsfb = ics->num_swb; cmaxsfb > 0 && cpe->ch[ch].zeroes[w+cmaxsfb-1]; cmaxsfb--)
329
                ;
330
            maxsfb = FFMAX(maxsfb, cmaxsfb);
331
        }
332
        ics->max_sfb = maxsfb;
333

    
334
        //adjust zero bands for window groups
335
        for (w = 0; w < ics->num_windows; w += ics->group_len[w]) {
336
            for (g = 0; g < ics->max_sfb; g++) {
337
                i = 1;
338
                for (w2 = w; w2 < w + ics->group_len[w]; w2++) {
339
                    if (!cpe->ch[ch].zeroes[w2*16 + g]) {
340
                        i = 0;
341
                        break;
342
                    }
343
                }
344
                cpe->ch[ch].zeroes[w*16 + g] = i;
345
            }
346
        }
347
    }
348

    
349
    if (chans > 1 && cpe->common_window) {
350
        IndividualChannelStream *ics0 = &cpe->ch[0].ics;
351
        IndividualChannelStream *ics1 = &cpe->ch[1].ics;
352
        int msc = 0;
353
        ics0->max_sfb = FFMAX(ics0->max_sfb, ics1->max_sfb);
354
        ics1->max_sfb = ics0->max_sfb;
355
        for (w = 0; w < ics0->num_windows*16; w += 16)
356
            for (i = 0; i < ics0->max_sfb; i++)
357
                if (cpe->ms_mask[w+i])
358
                    msc++;
359
        if (msc == 0 || ics0->max_sfb == 0)
360
            cpe->ms_mode = 0;
361
        else
362
            cpe->ms_mode = msc < ics0->max_sfb ? 1 : 2;
363
    }
364
}
365

    
366
/**
367
 * Encode scalefactor band coding type.
368
 */
369
static void encode_band_info(AACEncContext *s, SingleChannelElement *sce)
370
{
371
    int w;
372

    
373
    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w])
374
        s->coder->encode_window_bands_info(s, sce, w, sce->ics.group_len[w], s->lambda);
375
}
376

    
377
/**
378
 * Encode scalefactors.
379
 */
380
static void encode_scale_factors(AVCodecContext *avctx, AACEncContext *s,
381
                                 SingleChannelElement *sce)
382
{
383
    int off = sce->sf_idx[0], diff;
384
    int i, w;
385

    
386
    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
387
        for (i = 0; i < sce->ics.max_sfb; i++) {
388
            if (!sce->zeroes[w*16 + i]) {
389
                diff = sce->sf_idx[w*16 + i] - off + SCALE_DIFF_ZERO;
390
                if (diff < 0 || diff > 120)
391
                    av_log(avctx, AV_LOG_ERROR, "Scalefactor difference is too big to be coded\n");
392
                off = sce->sf_idx[w*16 + i];
393
                put_bits(&s->pb, ff_aac_scalefactor_bits[diff], ff_aac_scalefactor_code[diff]);
394
            }
395
        }
396
    }
397
}
398

    
399
/**
400
 * Encode pulse data.
401
 */
402
static void encode_pulses(AACEncContext *s, Pulse *pulse)
403
{
404
    int i;
405

    
406
    put_bits(&s->pb, 1, !!pulse->num_pulse);
407
    if (!pulse->num_pulse)
408
        return;
409

    
410
    put_bits(&s->pb, 2, pulse->num_pulse - 1);
411
    put_bits(&s->pb, 6, pulse->start);
412
    for (i = 0; i < pulse->num_pulse; i++) {
413
        put_bits(&s->pb, 5, pulse->pos[i]);
414
        put_bits(&s->pb, 4, pulse->amp[i]);
415
    }
416
}
417

    
418
/**
419
 * Encode spectral coefficients processed by psychoacoustic model.
420
 */
421
static void encode_spectral_coeffs(AACEncContext *s, SingleChannelElement *sce)
422
{
423
    int start, i, w, w2;
424

    
425
    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
426
        start = 0;
427
        for (i = 0; i < sce->ics.max_sfb; i++) {
428
            if (sce->zeroes[w*16 + i]) {
429
                start += sce->ics.swb_sizes[i];
430
                continue;
431
            }
432
            for (w2 = w; w2 < w + sce->ics.group_len[w]; w2++)
433
                s->coder->quantize_and_encode_band(s, &s->pb, sce->coeffs + start + w2*128,
434
                                                   sce->ics.swb_sizes[i],
435
                                                   sce->sf_idx[w*16 + i],
436
                                                   sce->band_type[w*16 + i],
437
                                                   s->lambda);
438
            start += sce->ics.swb_sizes[i];
439
        }
440
    }
441
}
442

    
443
/**
444
 * Encode one channel of audio data.
445
 */
446
static int encode_individual_channel(AVCodecContext *avctx, AACEncContext *s,
447
                                     SingleChannelElement *sce,
448
                                     int common_window)
449
{
450
    put_bits(&s->pb, 8, sce->sf_idx[0]);
451
    if (!common_window)
452
        put_ics_info(s, &sce->ics);
453
    encode_band_info(s, sce);
454
    encode_scale_factors(avctx, s, sce);
455
    encode_pulses(s, &sce->pulse);
456
    put_bits(&s->pb, 1, 0); //tns
457
    put_bits(&s->pb, 1, 0); //ssr
458
    encode_spectral_coeffs(s, sce);
459
    return 0;
460
}
461

    
462
/**
463
 * Write some auxiliary information about the created AAC file.
464
 */
465
static void put_bitstream_info(AVCodecContext *avctx, AACEncContext *s,
466
                               const char *name)
467
{
468
    int i, namelen, padbits;
469

    
470
    namelen = strlen(name) + 2;
471
    put_bits(&s->pb, 3, TYPE_FIL);
472
    put_bits(&s->pb, 4, FFMIN(namelen, 15));
473
    if (namelen >= 15)
474
        put_bits(&s->pb, 8, namelen - 16);
475
    put_bits(&s->pb, 4, 0); //extension type - filler
476
    padbits = 8 - (put_bits_count(&s->pb) & 7);
477
    align_put_bits(&s->pb);
478
    for (i = 0; i < namelen - 2; i++)
479
        put_bits(&s->pb, 8, name[i]);
480
    put_bits(&s->pb, 12 - padbits, 0);
481
}
482

    
483
static int aac_encode_frame(AVCodecContext *avctx,
484
                            uint8_t *frame, int buf_size, void *data)
485
{
486
    AACEncContext *s = avctx->priv_data;
487
    int16_t *samples = s->samples, *samples2, *la;
488
    ChannelElement *cpe;
489
    int i, j, chans, tag, start_ch;
490
    const uint8_t *chan_map = aac_chan_configs[avctx->channels-1];
491
    int chan_el_counter[4];
492
    FFPsyWindowInfo windows[AAC_MAX_CHANNELS];
493

    
494
    if (s->last_frame)
495
        return 0;
496
    if (data) {
497
        if (!s->psypp) {
498
            memcpy(s->samples + 1024 * avctx->channels, data,
499
                   1024 * avctx->channels * sizeof(s->samples[0]));
500
        } else {
501
            start_ch = 0;
502
            samples2 = s->samples + 1024 * avctx->channels;
503
            for (i = 0; i < chan_map[0]; i++) {
504
                tag = chan_map[i+1];
505
                chans = tag == TYPE_CPE ? 2 : 1;
506
                ff_psy_preprocess(s->psypp, (uint16_t*)data + start_ch,
507
                                  samples2 + start_ch, start_ch, chans);
508
                start_ch += chans;
509
            }
510
        }
511
    }
512
    if (!avctx->frame_number) {
513
        memcpy(s->samples, s->samples + 1024 * avctx->channels,
514
               1024 * avctx->channels * sizeof(s->samples[0]));
515
        return 0;
516
    }
517

    
518
    start_ch = 0;
519
    for (i = 0; i < chan_map[0]; i++) {
520
        FFPsyWindowInfo* wi = windows + start_ch;
521
        tag      = chan_map[i+1];
522
        chans    = tag == TYPE_CPE ? 2 : 1;
523
        cpe      = &s->cpe[i];
524
        for (j = 0; j < chans; j++) {
525
            IndividualChannelStream *ics = &cpe->ch[j].ics;
526
            int k;
527
            int cur_channel = start_ch + j;
528
            samples2 = samples + cur_channel;
529
            la       = samples2 + (448+64) * avctx->channels;
530
            if (!data)
531
                la = NULL;
532
            if (tag == TYPE_LFE) {
533
                wi[j].window_type[0] = ONLY_LONG_SEQUENCE;
534
                wi[j].window_shape   = 0;
535
                wi[j].num_windows    = 1;
536
                wi[j].grouping[0]    = 1;
537
            } else {
538
                wi[j] = ff_psy_suggest_window(&s->psy, samples2, la, cur_channel,
539
                                              ics->window_sequence[0]);
540
            }
541
            ics->window_sequence[1] = ics->window_sequence[0];
542
            ics->window_sequence[0] = wi[j].window_type[0];
543
            ics->use_kb_window[1]   = ics->use_kb_window[0];
544
            ics->use_kb_window[0]   = wi[j].window_shape;
545
            ics->num_windows        = wi[j].num_windows;
546
            ics->swb_sizes          = s->psy.bands    [ics->num_windows == 8];
547
            ics->num_swb            = tag == TYPE_LFE ? 12 : s->psy.num_bands[ics->num_windows == 8];
548
            for (k = 0; k < ics->num_windows; k++)
549
                ics->group_len[k] = wi[j].grouping[k];
550

    
551
            apply_window_and_mdct(avctx, s, &cpe->ch[j], samples2);
552
        }
553
        start_ch += chans;
554
    }
555
    do {
556
        int frame_bits;
557
        init_put_bits(&s->pb, frame, buf_size*8);
558
        if ((avctx->frame_number & 0xFF)==1 && !(avctx->flags & CODEC_FLAG_BITEXACT))
559
            put_bitstream_info(avctx, s, LIBAVCODEC_IDENT);
560
        start_ch = 0;
561
        memset(chan_el_counter, 0, sizeof(chan_el_counter));
562
        for (i = 0; i < chan_map[0]; i++) {
563
            FFPsyWindowInfo* wi = windows + start_ch;
564
            tag      = chan_map[i+1];
565
            chans    = tag == TYPE_CPE ? 2 : 1;
566
            cpe      = &s->cpe[i];
567
            put_bits(&s->pb, 3, tag);
568
            put_bits(&s->pb, 4, chan_el_counter[tag]++);
569
            for (j = 0; j < chans; j++) {
570
                s->cur_channel = start_ch + j;
571
                ff_psy_set_band_info(&s->psy, s->cur_channel, cpe->ch[j].coeffs, &wi[j]);
572
                s->coder->search_for_quantizers(avctx, s, &cpe->ch[j], s->lambda);
573
            }
574
            cpe->common_window = 0;
575
            if (chans > 1
576
                && wi[0].window_type[0] == wi[1].window_type[0]
577
                && wi[0].window_shape   == wi[1].window_shape) {
578

    
579
                cpe->common_window = 1;
580
                for (j = 0; j < wi[0].num_windows; j++) {
581
                    if (wi[0].grouping[j] != wi[1].grouping[j]) {
582
                        cpe->common_window = 0;
583
                        break;
584
                    }
585
                }
586
            }
587
            s->cur_channel = start_ch;
588
            if (cpe->common_window && s->coder->search_for_ms)
589
                s->coder->search_for_ms(s, cpe, s->lambda);
590
            adjust_frame_information(s, cpe, chans);
591
            if (chans == 2) {
592
                put_bits(&s->pb, 1, cpe->common_window);
593
                if (cpe->common_window) {
594
                    put_ics_info(s, &cpe->ch[0].ics);
595
                    encode_ms_info(&s->pb, cpe);
596
                }
597
            }
598
            for (j = 0; j < chans; j++) {
599
                s->cur_channel = start_ch + j;
600
                encode_individual_channel(avctx, s, &cpe->ch[j], cpe->common_window);
601
            }
602
            start_ch += chans;
603
        }
604

    
605
        frame_bits = put_bits_count(&s->pb);
606
        if (frame_bits <= 6144 * avctx->channels - 3)
607
            break;
608

    
609
        s->lambda *= avctx->bit_rate * 1024.0f / avctx->sample_rate / frame_bits;
610

    
611
    } while (1);
612

    
613
    put_bits(&s->pb, 3, TYPE_END);
614
    flush_put_bits(&s->pb);
615
    avctx->frame_bits = put_bits_count(&s->pb);
616

    
617
    // rate control stuff
618
    if (!(avctx->flags & CODEC_FLAG_QSCALE)) {
619
        float ratio = avctx->bit_rate * 1024.0f / avctx->sample_rate / avctx->frame_bits;
620
        s->lambda *= ratio;
621
        s->lambda = FFMIN(s->lambda, 65536.f);
622
    }
623

    
624
    if (!data)
625
        s->last_frame = 1;
626
    memcpy(s->samples, s->samples + 1024 * avctx->channels,
627
           1024 * avctx->channels * sizeof(s->samples[0]));
628
    return put_bits_count(&s->pb)>>3;
629
}
630

    
631
static av_cold int aac_encode_end(AVCodecContext *avctx)
632
{
633
    AACEncContext *s = avctx->priv_data;
634

    
635
    ff_mdct_end(&s->mdct1024);
636
    ff_mdct_end(&s->mdct128);
637
    ff_psy_end(&s->psy);
638
    ff_psy_preprocess_end(s->psypp);
639
    av_freep(&s->samples);
640
    av_freep(&s->cpe);
641
    return 0;
642
}
643

    
644
AVCodec aac_encoder = {
645
    "aac",
646
    AVMEDIA_TYPE_AUDIO,
647
    CODEC_ID_AAC,
648
    sizeof(AACEncContext),
649
    aac_encode_init,
650
    aac_encode_frame,
651
    aac_encode_end,
652
    .capabilities = CODEC_CAP_SMALL_LAST_FRAME | CODEC_CAP_DELAY | CODEC_CAP_EXPERIMENTAL,
653
    .sample_fmts = (const enum AVSampleFormat[]){AV_SAMPLE_FMT_S16,AV_SAMPLE_FMT_NONE},
654
    .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
655
};