2 * TTA (The Lossless True Audio) decoder
3 * Copyright (c) 2006 Alex Beregszaszi
5 * This file is part of FFmpeg.
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.
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.
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
24 * TTA (The Lossless True Audio) decoder
25 * @see http://www.true-audio.com/
26 * @see http://tta.corecodec.org/
27 * @author Alex Beregszaszi
30 #define ALT_BITSTREAM_READER_LE
37 #define FORMAT_FLOAT 3
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];
47 typedef struct TTARice {
48 uint32_t k0, k1, sum0, sum1;
51 typedef struct TTAChannel {
57 typedef struct TTAContext {
58 AVCodecContext *avctx;
61 int flags, channels, bps, is_float, data_length;
62 int frame_length, last_frame_length, total_frames;
64 int32_t *decode_buffer;
69 static const uint32_t shift_1[] = {
70 0x00000001, 0x00000002, 0x00000004, 0x00000008,
71 0x00000010, 0x00000020, 0x00000040, 0x00000080,
72 0x00000100, 0x00000200, 0x00000400, 0x00000800,
73 0x00001000, 0x00002000, 0x00004000, 0x00008000,
74 0x00010000, 0x00020000, 0x00040000, 0x00080000,
75 0x00100000, 0x00200000, 0x00400000, 0x00800000,
76 0x01000000, 0x02000000, 0x04000000, 0x08000000,
77 0x10000000, 0x20000000, 0x40000000, 0x80000000,
78 0x80000000, 0x80000000, 0x80000000, 0x80000000,
79 0x80000000, 0x80000000, 0x80000000, 0x80000000
82 static const uint32_t * const shift_16 = shift_1 + 4;
84 static const int32_t ttafilter_configs[4][2] = {
91 static void ttafilter_init(TTAFilter *c, int32_t shift, int32_t mode) {
92 memset(c, 0, sizeof(TTAFilter));
94 c->round = shift_1[shift-1];
95 // c->round = 1 << (shift - 1);
99 // FIXME: copy paste from original
100 static inline void memshl(register int32_t *a, register int32_t *b) {
111 // FIXME: copy paste from original
112 // mode=1 encoder, mode=0 decoder
113 static inline void ttafilter_process(TTAFilter *c, int32_t *in, int32_t mode) {
114 register int32_t *dl = c->dl, *qm = c->qm, *dx = c->dx, sum = c->round;
117 sum += *dl++ * *qm, qm++;
118 sum += *dl++ * *qm, qm++;
119 sum += *dl++ * *qm, qm++;
120 sum += *dl++ * *qm, qm++;
121 sum += *dl++ * *qm, qm++;
122 sum += *dl++ * *qm, qm++;
123 sum += *dl++ * *qm, qm++;
124 sum += *dl++ * *qm, qm++;
126 } else if(c->error < 0) {
127 sum += *dl++ * (*qm -= *dx++), qm++;
128 sum += *dl++ * (*qm -= *dx++), qm++;
129 sum += *dl++ * (*qm -= *dx++), qm++;
130 sum += *dl++ * (*qm -= *dx++), qm++;
131 sum += *dl++ * (*qm -= *dx++), qm++;
132 sum += *dl++ * (*qm -= *dx++), qm++;
133 sum += *dl++ * (*qm -= *dx++), qm++;
134 sum += *dl++ * (*qm -= *dx++), qm++;
136 sum += *dl++ * (*qm += *dx++), qm++;
137 sum += *dl++ * (*qm += *dx++), qm++;
138 sum += *dl++ * (*qm += *dx++), qm++;
139 sum += *dl++ * (*qm += *dx++), qm++;
140 sum += *dl++ * (*qm += *dx++), qm++;
141 sum += *dl++ * (*qm += *dx++), qm++;
142 sum += *dl++ * (*qm += *dx++), qm++;
143 sum += *dl++ * (*qm += *dx++), qm++;
146 *(dx-0) = ((*(dl-1) >> 30) | 1) << 2;
147 *(dx-1) = ((*(dl-2) >> 30) | 1) << 1;
148 *(dx-2) = ((*(dl-3) >> 30) | 1) << 1;
149 *(dx-3) = ((*(dl-4) >> 30) | 1);
154 *in -= (sum >> c->shift);
158 *in += (sum >> c->shift);
163 *(dl-1) = *dl - *(dl-1);
164 *(dl-2) = *(dl-1) - *(dl-2);
165 *(dl-3) = *(dl-2) - *(dl-3);
168 memshl(c->dl, c->dl + 1);
169 memshl(c->dx, c->dx + 1);
172 static void rice_init(TTARice *c, uint32_t k0, uint32_t k1)
176 c->sum0 = shift_16[k0];
177 c->sum1 = shift_16[k1];
180 static int tta_get_unary(GetBitContext *gb)
190 static const int64_t tta_channel_layouts[7] = {
192 AV_CH_LAYOUT_STEREO|AV_CH_LOW_FREQUENCY,
195 AV_CH_LAYOUT_5POINT1_BACK,
196 AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER,
197 AV_CH_LAYOUT_7POINT1_WIDE
200 static av_cold int tta_decode_init(AVCodecContext * avctx)
202 TTAContext *s = avctx->priv_data;
207 // 30bytes includes a seektable with one frame
208 if (avctx->extradata_size < 30)
211 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size * 8);
212 if (show_bits_long(&s->gb, 32) == AV_RL32("TTA1"))
215 skip_bits(&s->gb, 32);
216 // if (get_bits_long(&s->gb, 32) != av_bswap32(AV_RL32("TTA1"))) {
217 // av_log(s->avctx, AV_LOG_ERROR, "Missing magic\n");
221 s->flags = get_bits(&s->gb, 16);
222 if (s->flags != 1 && s->flags != 3)
224 av_log(s->avctx, AV_LOG_ERROR, "Invalid flags\n");
227 s->is_float = (s->flags == FORMAT_FLOAT);
228 avctx->channels = s->channels = get_bits(&s->gb, 16);
229 if (s->channels > 1 && s->channels < 9)
230 avctx->channel_layout = tta_channel_layouts[s->channels-2];
231 avctx->bits_per_coded_sample = get_bits(&s->gb, 16);
232 s->bps = (avctx->bits_per_coded_sample + 7) / 8;
233 avctx->sample_rate = get_bits_long(&s->gb, 32);
234 if(avctx->sample_rate > 1000000){ //prevent FRAME_TIME * avctx->sample_rate from overflowing and sanity check
235 av_log(avctx, AV_LOG_ERROR, "sample_rate too large\n");
238 s->data_length = get_bits_long(&s->gb, 32);
239 skip_bits(&s->gb, 32); // CRC32 of header
243 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
244 av_log_ask_for_sample(s->avctx, "Unsupported sample format.\n");
247 else switch(s->bps) {
248 case 1: avctx->sample_fmt = AV_SAMPLE_FMT_U8; break;
249 case 2: avctx->sample_fmt = AV_SAMPLE_FMT_S16; break;
250 case 3: avctx->bits_per_coded_sample = 24;
251 case 4: avctx->sample_fmt = AV_SAMPLE_FMT_S32; break;
253 av_log_ask_for_sample(s->avctx,
254 "Invalid/unsupported sample format.\n");
258 // FIXME: horribly broken, but directly from reference source
259 #define FRAME_TIME 1.04489795918367346939
260 s->frame_length = (int)(FRAME_TIME * avctx->sample_rate);
262 s->last_frame_length = s->data_length % s->frame_length;
263 s->total_frames = s->data_length / s->frame_length +
264 (s->last_frame_length ? 1 : 0);
266 av_log(s->avctx, AV_LOG_DEBUG, "flags: %x chans: %d bps: %d rate: %d block: %d\n",
267 s->flags, avctx->channels, avctx->bits_per_coded_sample, avctx->sample_rate,
269 av_log(s->avctx, AV_LOG_DEBUG, "data_length: %d frame_length: %d last: %d total: %d\n",
270 s->data_length, s->frame_length, s->last_frame_length, s->total_frames);
273 for (i = 0; i < s->total_frames; i++)
274 skip_bits(&s->gb, 32);
275 skip_bits(&s->gb, 32); // CRC32 of seektable
277 if(s->frame_length >= UINT_MAX / (s->channels * sizeof(int32_t))){
278 av_log(avctx, AV_LOG_ERROR, "frame_length too large\n");
282 s->decode_buffer = av_mallocz(sizeof(int32_t)*s->frame_length*s->channels);
283 if (!s->decode_buffer)
284 return AVERROR(ENOMEM);
285 s->ch_ctx = av_malloc(avctx->channels * sizeof(*s->ch_ctx));
287 return AVERROR(ENOMEM);
289 av_log(avctx, AV_LOG_ERROR, "Wrong extradata present\n");
296 static int tta_decode_frame(AVCodecContext *avctx,
297 void *data, int *data_size,
300 const uint8_t *buf = avpkt->data;
301 int buf_size = avpkt->size;
302 TTAContext *s = avctx->priv_data;
305 init_get_bits(&s->gb, buf, buf_size*8);
307 int cur_chan = 0, framelen = s->frame_length;
310 if (*data_size < (framelen * s->channels * av_get_bits_per_sample_fmt(avctx->sample_fmt) / 8)) {
311 av_log(avctx, AV_LOG_ERROR, "Output buffer size is too small.\n");
316 if (!s->total_frames && s->last_frame_length)
317 framelen = s->last_frame_length;
319 // init per channel states
320 for (i = 0; i < s->channels; i++) {
321 s->ch_ctx[i].predictor = 0;
322 ttafilter_init(&s->ch_ctx[i].filter, ttafilter_configs[s->bps-1][0], ttafilter_configs[s->bps-1][1]);
323 rice_init(&s->ch_ctx[i].rice, 10, 10);
326 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
327 int32_t *predictor = &s->ch_ctx[cur_chan].predictor;
328 TTAFilter *filter = &s->ch_ctx[cur_chan].filter;
329 TTARice *rice = &s->ch_ctx[cur_chan].rice;
330 uint32_t unary, depth, k;
333 unary = tta_get_unary(&s->gb);
344 if (get_bits_left(&s->gb) < k)
348 if (k > MIN_CACHE_BITS)
350 value = (unary << k) + get_bits(&s->gb, k);
354 // FIXME: copy paste from original
357 rice->sum1 += value - (rice->sum1 >> 4);
358 if (rice->k1 > 0 && rice->sum1 < shift_16[rice->k1])
360 else if(rice->sum1 > shift_16[rice->k1 + 1])
362 value += shift_1[rice->k0];
364 rice->sum0 += value - (rice->sum0 >> 4);
365 if (rice->k0 > 0 && rice->sum0 < shift_16[rice->k0])
367 else if(rice->sum0 > shift_16[rice->k0 + 1])
371 // extract coded value
372 #define UNFOLD(x) (((x)&1) ? (++(x)>>1) : (-(x)>>1))
376 ttafilter_process(filter, p, 0);
378 // fixed order prediction
379 #define PRED(x, k) (int32_t)((((uint64_t)x << k) - x) >> k)
381 case 1: *p += PRED(*predictor, 4); break;
383 case 3: *p += PRED(*predictor, 5); break;
384 case 4: *p += *predictor; break;
388 /*if ((get_bits_count(&s->gb)+7)/8 > buf_size)
390 av_log(NULL, AV_LOG_INFO, "overread!!\n");
395 if (cur_chan < (s->channels-1))
398 // decorrelate in case of stereo integer
399 if (!s->is_float && (s->channels > 1)) {
401 for (*p += *r / 2; r > p - s->channels; r--)
408 if (get_bits_left(&s->gb) < 32)
410 skip_bits(&s->gb, 32); // frame crc
412 // convert to output buffer
415 uint8_t *samples = data;
416 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
417 *samples++ = *p + 0x80;
418 *data_size = samples - (uint8_t *)data;
422 uint16_t *samples = data;
423 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
424 // *samples++ = (unsigned char)*p;
425 // *samples++ = (unsigned char)(*p >> 8);
428 *data_size = (uint8_t *)samples - (uint8_t *)data;
432 int32_t *samples = data;
433 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
434 *samples++ = AV_RN32(p) << 8;
435 *data_size = (uint8_t *)samples - (uint8_t *)data;
439 av_log(s->avctx, AV_LOG_ERROR, "Error, only 16bit samples supported!\n");
443 // return get_bits_count(&s->gb)+7)/8;
447 static av_cold int tta_decode_close(AVCodecContext *avctx) {
448 TTAContext *s = avctx->priv_data;
450 av_free(s->decode_buffer);
451 av_freep(&s->ch_ctx);
456 AVCodec ff_tta_decoder = {
458 .type = AVMEDIA_TYPE_AUDIO,
460 .priv_data_size = sizeof(TTAContext),
461 .init = tta_decode_init,
462 .close = tta_decode_close,
463 .decode = tta_decode_frame,
464 .long_name = NULL_IF_CONFIG_SMALL("True Audio (TTA)"),