Statistics
| Branch: | Revision:

ffmpeg / libavcodec / tta.c @ 4cf2e30f

History | View | Annotate | Download (14.7 KB)

1
/*
2
 * TTA (The Lossless True Audio) decoder
3
 * Copyright (c) 2006 Alex Beregszaszi
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
 * TTA (The Lossless True Audio) decoder
25
 * (www.true-audio.com or tta.corecodec.org)
26
 * @author Alex Beregszaszi
27
 *
28
 */
29

    
30
#define ALT_BITSTREAM_READER_LE
31
//#define DEBUG
32
#include <limits.h>
33
#include "avcodec.h"
34
#include "get_bits.h"
35

    
36
#define FORMAT_INT 1
37
#define FORMAT_FLOAT 3
38

    
39
#define MAX_ORDER 16
40
typedef struct TTAFilter {
41
    int32_t shift, round, error, mode;
42
    int32_t qm[MAX_ORDER];
43
    int32_t dx[MAX_ORDER];
44
    int32_t dl[MAX_ORDER];
45
} TTAFilter;
46

    
47
typedef struct TTARice {
48
    uint32_t k0, k1, sum0, sum1;
49
} TTARice;
50

    
51
typedef struct TTAChannel {
52
    int32_t predictor;
53
    TTAFilter filter;
54
    TTARice rice;
55
} TTAChannel;
56

    
57
typedef struct TTAContext {
58
    AVCodecContext *avctx;
59
    GetBitContext gb;
60

    
61
    int flags, channels, bps, is_float, data_length;
62
    int frame_length, last_frame_length, total_frames;
63

    
64
    int32_t *decode_buffer;
65

    
66
    TTAChannel *ch_ctx;
67
} TTAContext;
68

    
69
#if 0
70
static inline int shift_1(int i)
71
{
72
    if (i < 32)
73
        return 1 << i;
74
    else
75
        return 0x80000000; // 16 << 31
76
}
77

78
static inline int shift_16(int i)
79
{
80
    if (i < 28)
81
        return 16 << i;
82
    else
83
        return 0x80000000; // 16 << 27
84
}
85
#else
86
static const uint32_t shift_1[] = {
87
    0x00000001, 0x00000002, 0x00000004, 0x00000008,
88
    0x00000010, 0x00000020, 0x00000040, 0x00000080,
89
    0x00000100, 0x00000200, 0x00000400, 0x00000800,
90
    0x00001000, 0x00002000, 0x00004000, 0x00008000,
91
    0x00010000, 0x00020000, 0x00040000, 0x00080000,
92
    0x00100000, 0x00200000, 0x00400000, 0x00800000,
93
    0x01000000, 0x02000000, 0x04000000, 0x08000000,
94
    0x10000000, 0x20000000, 0x40000000, 0x80000000,
95
    0x80000000, 0x80000000, 0x80000000, 0x80000000,
96
    0x80000000, 0x80000000, 0x80000000, 0x80000000
97
};
98

    
99
static const uint32_t * const shift_16 = shift_1 + 4;
100
#endif
101

    
102
static const int32_t ttafilter_configs[4][2] = {
103
    {10, 1},
104
    {9, 1},
105
    {10, 1},
106
    {12, 0}
107
};
108

    
109
static void ttafilter_init(TTAFilter *c, int32_t shift, int32_t mode) {
110
    memset(c, 0, sizeof(TTAFilter));
111
    c->shift = shift;
112
   c->round = shift_1[shift-1];
113
//    c->round = 1 << (shift - 1);
114
    c->mode = mode;
115
}
116

    
117
// FIXME: copy paste from original
118
static inline void memshl(register int32_t *a, register int32_t *b) {
119
    *a++ = *b++;
120
    *a++ = *b++;
121
    *a++ = *b++;
122
    *a++ = *b++;
123
    *a++ = *b++;
124
    *a++ = *b++;
125
    *a++ = *b++;
126
    *a = *b;
127
}
128

    
129
// FIXME: copy paste from original
130
// mode=1 encoder, mode=0 decoder
131
static inline void ttafilter_process(TTAFilter *c, int32_t *in, int32_t mode) {
132
    register int32_t *dl = c->dl, *qm = c->qm, *dx = c->dx, sum = c->round;
133

    
134
    if (!c->error) {
135
        sum += *dl++ * *qm, qm++;
136
        sum += *dl++ * *qm, qm++;
137
        sum += *dl++ * *qm, qm++;
138
        sum += *dl++ * *qm, qm++;
139
        sum += *dl++ * *qm, qm++;
140
        sum += *dl++ * *qm, qm++;
141
        sum += *dl++ * *qm, qm++;
142
        sum += *dl++ * *qm, qm++;
143
        dx += 8;
144
    } else if(c->error < 0) {
145
        sum += *dl++ * (*qm -= *dx++), qm++;
146
        sum += *dl++ * (*qm -= *dx++), qm++;
147
        sum += *dl++ * (*qm -= *dx++), qm++;
148
        sum += *dl++ * (*qm -= *dx++), qm++;
149
        sum += *dl++ * (*qm -= *dx++), qm++;
150
        sum += *dl++ * (*qm -= *dx++), qm++;
151
        sum += *dl++ * (*qm -= *dx++), qm++;
152
        sum += *dl++ * (*qm -= *dx++), qm++;
153
    } else {
154
        sum += *dl++ * (*qm += *dx++), qm++;
155
        sum += *dl++ * (*qm += *dx++), qm++;
156
        sum += *dl++ * (*qm += *dx++), qm++;
157
        sum += *dl++ * (*qm += *dx++), qm++;
158
        sum += *dl++ * (*qm += *dx++), qm++;
159
        sum += *dl++ * (*qm += *dx++), qm++;
160
        sum += *dl++ * (*qm += *dx++), qm++;
161
        sum += *dl++ * (*qm += *dx++), qm++;
162
    }
163

    
164
    *(dx-0) = ((*(dl-1) >> 30) | 1) << 2;
165
    *(dx-1) = ((*(dl-2) >> 30) | 1) << 1;
166
    *(dx-2) = ((*(dl-3) >> 30) | 1) << 1;
167
    *(dx-3) = ((*(dl-4) >> 30) | 1);
168

    
169
    // compress
170
    if (mode) {
171
        *dl = *in;
172
        *in -= (sum >> c->shift);
173
        c->error = *in;
174
    } else {
175
        c->error = *in;
176
        *in += (sum >> c->shift);
177
        *dl = *in;
178
    }
179

    
180
    if (c->mode) {
181
        *(dl-1) = *dl - *(dl-1);
182
        *(dl-2) = *(dl-1) - *(dl-2);
183
        *(dl-3) = *(dl-2) - *(dl-3);
184
    }
185

    
186
    memshl(c->dl, c->dl + 1);
187
    memshl(c->dx, c->dx + 1);
188
}
189

    
190
static void rice_init(TTARice *c, uint32_t k0, uint32_t k1)
191
{
192
    c->k0 = k0;
193
    c->k1 = k1;
194
    c->sum0 = shift_16[k0];
195
    c->sum1 = shift_16[k1];
196
}
197

    
198
static int tta_get_unary(GetBitContext *gb)
199
{
200
    int ret = 0;
201

    
202
    // count ones
203
    while(get_bits1(gb))
204
        ret++;
205
    return ret;
206
}
207

    
208
static const int64_t tta_channel_layouts[7] = {
209
    AV_CH_LAYOUT_STEREO,
210
    AV_CH_LAYOUT_STEREO|AV_CH_LOW_FREQUENCY,
211
    AV_CH_LAYOUT_QUAD,
212
    0,
213
    AV_CH_LAYOUT_5POINT1_BACK,
214
    AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER,
215
    AV_CH_LAYOUT_7POINT1_WIDE
216
};
217

    
218
static av_cold int tta_decode_init(AVCodecContext * avctx)
219
{
220
    TTAContext *s = avctx->priv_data;
221
    int i;
222

    
223
    s->avctx = avctx;
224

    
225
    // 30bytes includes a seektable with one frame
226
    if (avctx->extradata_size < 30)
227
        return -1;
228

    
229
    init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size);
230
    if (show_bits_long(&s->gb, 32) == AV_RL32("TTA1"))
231
    {
232
        /* signature */
233
        skip_bits(&s->gb, 32);
234
//        if (get_bits_long(&s->gb, 32) != av_bswap32(AV_RL32("TTA1"))) {
235
//            av_log(s->avctx, AV_LOG_ERROR, "Missing magic\n");
236
//            return -1;
237
//        }
238

    
239
        s->flags = get_bits(&s->gb, 16);
240
        if (s->flags != 1 && s->flags != 3)
241
        {
242
            av_log(s->avctx, AV_LOG_ERROR, "Invalid flags\n");
243
            return -1;
244
        }
245
        s->is_float = (s->flags == FORMAT_FLOAT);
246
        avctx->channels = s->channels = get_bits(&s->gb, 16);
247
        if (s->channels > 1 && s->channels < 9)
248
            avctx->channel_layout = tta_channel_layouts[s->channels-2];
249
        avctx->bits_per_coded_sample = get_bits(&s->gb, 16);
250
        s->bps = (avctx->bits_per_coded_sample + 7) / 8;
251
        avctx->sample_rate = get_bits_long(&s->gb, 32);
252
        if(avctx->sample_rate > 1000000){ //prevent FRAME_TIME * avctx->sample_rate from overflowing and sanity check
253
            av_log(avctx, AV_LOG_ERROR, "sample_rate too large\n");
254
            return -1;
255
        }
256
        s->data_length = get_bits_long(&s->gb, 32);
257
        skip_bits(&s->gb, 32); // CRC32 of header
258

    
259
        if (s->is_float)
260
        {
261
            avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
262
            av_log_ask_for_sample(s->avctx, "Unsupported sample format.\n");
263
            return -1;
264
        }
265
        else switch(s->bps) {
266
            case 1: avctx->sample_fmt = AV_SAMPLE_FMT_U8; break;
267
            case 2: avctx->sample_fmt = AV_SAMPLE_FMT_S16; break;
268
//            case 3: avctx->sample_fmt = AV_SAMPLE_FMT_S24; break;
269
            case 4: avctx->sample_fmt = AV_SAMPLE_FMT_S32; break;
270
            default:
271
                av_log_ask_for_sample(s->avctx,
272
                                      "Invalid/unsupported sample format.\n");
273
                return -1;
274
        }
275

    
276
        // FIXME: horribly broken, but directly from reference source
277
#define FRAME_TIME 1.04489795918367346939
278
        s->frame_length = (int)(FRAME_TIME * avctx->sample_rate);
279

    
280
        s->last_frame_length = s->data_length % s->frame_length;
281
        s->total_frames = s->data_length / s->frame_length +
282
                        (s->last_frame_length ? 1 : 0);
283

    
284
        av_log(s->avctx, AV_LOG_DEBUG, "flags: %x chans: %d bps: %d rate: %d block: %d\n",
285
            s->flags, avctx->channels, avctx->bits_per_coded_sample, avctx->sample_rate,
286
            avctx->block_align);
287
        av_log(s->avctx, AV_LOG_DEBUG, "data_length: %d frame_length: %d last: %d total: %d\n",
288
            s->data_length, s->frame_length, s->last_frame_length, s->total_frames);
289

    
290
        // FIXME: seek table
291
        for (i = 0; i < s->total_frames; i++)
292
            skip_bits(&s->gb, 32);
293
        skip_bits(&s->gb, 32); // CRC32 of seektable
294

    
295
        if(s->frame_length >= UINT_MAX / (s->channels * sizeof(int32_t))){
296
            av_log(avctx, AV_LOG_ERROR, "frame_length too large\n");
297
            return -1;
298
        }
299

    
300
        s->decode_buffer = av_mallocz(sizeof(int32_t)*s->frame_length*s->channels);
301
        if (!s->decode_buffer)
302
            return AVERROR(ENOMEM);
303
        s->ch_ctx = av_malloc(avctx->channels * sizeof(*s->ch_ctx));
304
        if (!s->ch_ctx)
305
            return AVERROR(ENOMEM);
306
    } else {
307
        av_log(avctx, AV_LOG_ERROR, "Wrong extradata present\n");
308
        return -1;
309
    }
310

    
311
    return 0;
312
}
313

    
314
static int tta_decode_frame(AVCodecContext *avctx,
315
        void *data, int *data_size,
316
        AVPacket *avpkt)
317
{
318
    const uint8_t *buf = avpkt->data;
319
    int buf_size = avpkt->size;
320
    TTAContext *s = avctx->priv_data;
321
    int i;
322

    
323
    init_get_bits(&s->gb, buf, buf_size*8);
324
    {
325
        int cur_chan = 0, framelen = s->frame_length;
326
        int32_t *p;
327

    
328
        if (*data_size < (framelen * s->channels * av_get_bits_per_sample_fmt(avctx->sample_fmt) / 8)) {
329
            av_log(avctx, AV_LOG_ERROR, "Output buffer size is too small.\n");
330
            return -1;
331
        }
332
        // FIXME: seeking
333
        s->total_frames--;
334
        if (!s->total_frames && s->last_frame_length)
335
            framelen = s->last_frame_length;
336

    
337
        // init per channel states
338
        for (i = 0; i < s->channels; i++) {
339
            s->ch_ctx[i].predictor = 0;
340
            ttafilter_init(&s->ch_ctx[i].filter, ttafilter_configs[s->bps-1][0], ttafilter_configs[s->bps-1][1]);
341
            rice_init(&s->ch_ctx[i].rice, 10, 10);
342
        }
343

    
344
        for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
345
            int32_t *predictor = &s->ch_ctx[cur_chan].predictor;
346
            TTAFilter *filter = &s->ch_ctx[cur_chan].filter;
347
            TTARice *rice = &s->ch_ctx[cur_chan].rice;
348
            uint32_t unary, depth, k;
349
            int32_t value;
350

    
351
            unary = tta_get_unary(&s->gb);
352

    
353
            if (unary == 0) {
354
                depth = 0;
355
                k = rice->k0;
356
            } else {
357
                depth = 1;
358
                k = rice->k1;
359
                unary--;
360
            }
361

    
362
            if (get_bits_left(&s->gb) < k)
363
                return -1;
364

    
365
            if (k) {
366
                if (k > MIN_CACHE_BITS)
367
                    return -1;
368
                value = (unary << k) + get_bits(&s->gb, k);
369
            } else
370
                value = unary;
371

    
372
            // FIXME: copy paste from original
373
            switch (depth) {
374
            case 1:
375
                rice->sum1 += value - (rice->sum1 >> 4);
376
                if (rice->k1 > 0 && rice->sum1 < shift_16[rice->k1])
377
                    rice->k1--;
378
                else if(rice->sum1 > shift_16[rice->k1 + 1])
379
                    rice->k1++;
380
                value += shift_1[rice->k0];
381
            default:
382
                rice->sum0 += value - (rice->sum0 >> 4);
383
                if (rice->k0 > 0 && rice->sum0 < shift_16[rice->k0])
384
                    rice->k0--;
385
                else if(rice->sum0 > shift_16[rice->k0 + 1])
386
                    rice->k0++;
387
            }
388

    
389
            // extract coded value
390
#define UNFOLD(x) (((x)&1) ? (++(x)>>1) : (-(x)>>1))
391
            *p = UNFOLD(value);
392

    
393
            // run hybrid filter
394
            ttafilter_process(filter, p, 0);
395

    
396
            // fixed order prediction
397
#define PRED(x, k) (int32_t)((((uint64_t)x << k) - x) >> k)
398
            switch (s->bps) {
399
                case 1: *p += PRED(*predictor, 4); break;
400
                case 2:
401
                case 3: *p += PRED(*predictor, 5); break;
402
                case 4: *p += *predictor; break;
403
            }
404
            *predictor = *p;
405

    
406
#if 0
407
            // extract 32bit float from last two int samples
408
            if (s->is_float && ((p - data) & 1)) {
409
                uint32_t neg = *p & 0x80000000;
410
                uint32_t hi = *(p - 1);
411
                uint32_t lo = abs(*p) - 1;
412

413
                hi += (hi || lo) ? 0x3f80 : 0;
414
                // SWAP16: swap all the 16 bits
415
                *(p - 1) = (hi << 16) | SWAP16(lo) | neg;
416
            }
417
#endif
418

    
419
            /*if ((get_bits_count(&s->gb)+7)/8 > buf_size)
420
            {
421
                av_log(NULL, AV_LOG_INFO, "overread!!\n");
422
                break;
423
            }*/
424

    
425
            // flip channels
426
            if (cur_chan < (s->channels-1))
427
                cur_chan++;
428
            else {
429
                // decorrelate in case of stereo integer
430
                if (!s->is_float && (s->channels > 1)) {
431
                    int32_t *r = p - 1;
432
                    for (*p += *r / 2; r > p - s->channels; r--)
433
                        *r = *(r + 1) - *r;
434
                }
435
                cur_chan = 0;
436
            }
437
        }
438

    
439
        if (get_bits_left(&s->gb) < 32)
440
            return -1;
441
        skip_bits(&s->gb, 32); // frame crc
442

    
443
        // convert to output buffer
444
        switch(s->bps) {
445
            case 1: {
446
                uint8_t *samples = data;
447
                for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
448
                    *samples++ = *p + 0x80;
449
                *data_size = samples - (uint8_t *)data;
450
                break;
451
            }
452
            case 2: {
453
                uint16_t *samples = data;
454
                for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
455
//                    *samples++ = (unsigned char)*p;
456
//                    *samples++ = (unsigned char)(*p >> 8);
457
                    *samples++ = *p;
458
                }
459
                *data_size = (uint8_t *)samples - (uint8_t *)data;
460
                break;
461
            }
462
            default:
463
                av_log(s->avctx, AV_LOG_ERROR, "Error, only 16bit samples supported!\n");
464
        }
465
    }
466

    
467
//    return get_bits_count(&s->gb)+7)/8;
468
    return buf_size;
469
}
470

    
471
static av_cold int tta_decode_close(AVCodecContext *avctx) {
472
    TTAContext *s = avctx->priv_data;
473

    
474
    av_free(s->decode_buffer);
475
    av_freep(&s->ch_ctx);
476

    
477
    return 0;
478
}
479

    
480
AVCodec ff_tta_decoder = {
481
    "tta",
482
    AVMEDIA_TYPE_AUDIO,
483
    CODEC_ID_TTA,
484
    sizeof(TTAContext),
485
    tta_decode_init,
486
    NULL,
487
    tta_decode_close,
488
    tta_decode_frame,
489
    .long_name = NULL_IF_CONFIG_SMALL("True Audio (TTA)"),
490
};