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 * (www.true-audio.com or tta.corecodec.org)
26 * @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;
70 static inline int shift_1(int i)
75 return 0x80000000; // 16 << 31
78 static inline int shift_16(int i)
83 return 0x80000000; // 16 << 27
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
99 static const uint32_t * const shift_16 = shift_1 + 4;
102 static const int32_t ttafilter_configs[4][2] = {
109 static void ttafilter_init(TTAFilter *c, int32_t shift, int32_t mode) {
110 memset(c, 0, sizeof(TTAFilter));
112 c->round = shift_1[shift-1];
113 // c->round = 1 << (shift - 1);
117 // FIXME: copy paste from original
118 static inline void memshl(register int32_t *a, register int32_t *b) {
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;
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++;
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++;
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++;
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);
172 *in -= (sum >> c->shift);
176 *in += (sum >> c->shift);
181 *(dl-1) = *dl - *(dl-1);
182 *(dl-2) = *(dl-1) - *(dl-2);
183 *(dl-3) = *(dl-2) - *(dl-3);
186 memshl(c->dl, c->dl + 1);
187 memshl(c->dx, c->dx + 1);
190 static void rice_init(TTARice *c, uint32_t k0, uint32_t k1)
194 c->sum0 = shift_16[k0];
195 c->sum1 = shift_16[k1];
198 static int tta_get_unary(GetBitContext *gb)
208 static const int64_t tta_channel_layouts[7] = {
210 AV_CH_LAYOUT_STEREO|AV_CH_LOW_FREQUENCY,
213 AV_CH_LAYOUT_5POINT1_BACK,
214 AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER,
215 AV_CH_LAYOUT_7POINT1_WIDE
218 static av_cold int tta_decode_init(AVCodecContext * avctx)
220 TTAContext *s = avctx->priv_data;
225 // 30bytes includes a seektable with one frame
226 if (avctx->extradata_size < 30)
229 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size);
230 if (show_bits_long(&s->gb, 32) == AV_RL32("TTA1"))
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");
239 s->flags = get_bits(&s->gb, 16);
240 if (s->flags != 1 && s->flags != 3)
242 av_log(s->avctx, AV_LOG_ERROR, "Invalid flags\n");
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");
256 s->data_length = get_bits_long(&s->gb, 32);
257 skip_bits(&s->gb, 32); // CRC32 of header
261 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
262 av_log_ask_for_sample(s->avctx, "Unsupported sample format.\n");
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->bits_per_coded_sample = 24;
269 case 4: avctx->sample_fmt = AV_SAMPLE_FMT_S32; break;
271 av_log_ask_for_sample(s->avctx,
272 "Invalid/unsupported sample format.\n");
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);
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);
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,
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);
291 for (i = 0; i < s->total_frames; i++)
292 skip_bits(&s->gb, 32);
293 skip_bits(&s->gb, 32); // CRC32 of seektable
295 if(s->frame_length >= UINT_MAX / (s->channels * sizeof(int32_t))){
296 av_log(avctx, AV_LOG_ERROR, "frame_length too large\n");
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));
305 return AVERROR(ENOMEM);
307 av_log(avctx, AV_LOG_ERROR, "Wrong extradata present\n");
314 static int tta_decode_frame(AVCodecContext *avctx,
315 void *data, int *data_size,
318 const uint8_t *buf = avpkt->data;
319 int buf_size = avpkt->size;
320 TTAContext *s = avctx->priv_data;
323 init_get_bits(&s->gb, buf, buf_size*8);
325 int cur_chan = 0, framelen = s->frame_length;
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");
334 if (!s->total_frames && s->last_frame_length)
335 framelen = s->last_frame_length;
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);
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;
351 unary = tta_get_unary(&s->gb);
362 if (get_bits_left(&s->gb) < k)
366 if (k > MIN_CACHE_BITS)
368 value = (unary << k) + get_bits(&s->gb, k);
372 // FIXME: copy paste from original
375 rice->sum1 += value - (rice->sum1 >> 4);
376 if (rice->k1 > 0 && rice->sum1 < shift_16[rice->k1])
378 else if(rice->sum1 > shift_16[rice->k1 + 1])
380 value += shift_1[rice->k0];
382 rice->sum0 += value - (rice->sum0 >> 4);
383 if (rice->k0 > 0 && rice->sum0 < shift_16[rice->k0])
385 else if(rice->sum0 > shift_16[rice->k0 + 1])
389 // extract coded value
390 #define UNFOLD(x) (((x)&1) ? (++(x)>>1) : (-(x)>>1))
394 ttafilter_process(filter, p, 0);
396 // fixed order prediction
397 #define PRED(x, k) (int32_t)((((uint64_t)x << k) - x) >> k)
399 case 1: *p += PRED(*predictor, 4); break;
401 case 3: *p += PRED(*predictor, 5); break;
402 case 4: *p += *predictor; break;
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;
413 hi += (hi || lo) ? 0x3f80 : 0;
414 // SWAP16: swap all the 16 bits
415 *(p - 1) = (hi << 16) | SWAP16(lo) | neg;
419 /*if ((get_bits_count(&s->gb)+7)/8 > buf_size)
421 av_log(NULL, AV_LOG_INFO, "overread!!\n");
426 if (cur_chan < (s->channels-1))
429 // decorrelate in case of stereo integer
430 if (!s->is_float && (s->channels > 1)) {
432 for (*p += *r / 2; r > p - s->channels; r--)
439 if (get_bits_left(&s->gb) < 32)
441 skip_bits(&s->gb, 32); // frame crc
443 // convert to output buffer
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;
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);
459 *data_size = (uint8_t *)samples - (uint8_t *)data;
463 int32_t *samples = data;
464 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
465 *samples++ = AV_RN32(p) << 8;
466 *data_size = (uint8_t *)samples - (uint8_t *)data;
470 av_log(s->avctx, AV_LOG_ERROR, "Error, only 16bit samples supported!\n");
474 // return get_bits_count(&s->gb)+7)/8;
478 static av_cold int tta_decode_close(AVCodecContext *avctx) {
479 TTAContext *s = avctx->priv_data;
481 av_free(s->decode_buffer);
482 av_freep(&s->ch_ctx);
487 AVCodec ff_tta_decoder = {
496 .long_name = NULL_IF_CONFIG_SMALL("True Audio (TTA)"),