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 BITSTREAM_READER_LE
35 #include "libavutil/crc.h"
37 #define FORMAT_SIMPLE 1
38 #define FORMAT_ENCRYPTED 2
41 typedef struct TTAFilter {
42 int32_t shift, round, error;
43 int32_t qm[MAX_ORDER];
44 int32_t dx[MAX_ORDER];
45 int32_t dl[MAX_ORDER];
48 typedef struct TTARice {
49 uint32_t k0, k1, sum0, sum1;
52 typedef struct TTAChannel {
58 typedef struct TTAContext {
59 AVCodecContext *avctx;
62 const AVCRC *crc_table;
64 int format, channels, bps, data_length;
65 int frame_length, last_frame_length, total_frames;
67 int32_t *decode_buffer;
72 static const uint32_t shift_1[] = {
73 0x00000001, 0x00000002, 0x00000004, 0x00000008,
74 0x00000010, 0x00000020, 0x00000040, 0x00000080,
75 0x00000100, 0x00000200, 0x00000400, 0x00000800,
76 0x00001000, 0x00002000, 0x00004000, 0x00008000,
77 0x00010000, 0x00020000, 0x00040000, 0x00080000,
78 0x00100000, 0x00200000, 0x00400000, 0x00800000,
79 0x01000000, 0x02000000, 0x04000000, 0x08000000,
80 0x10000000, 0x20000000, 0x40000000, 0x80000000,
81 0x80000000, 0x80000000, 0x80000000, 0x80000000,
82 0x80000000, 0x80000000, 0x80000000, 0x80000000
85 static const uint32_t * const shift_16 = shift_1 + 4;
87 static const int32_t ttafilter_configs[4] = {
94 static void ttafilter_init(TTAFilter *c, int32_t shift) {
95 memset(c, 0, sizeof(TTAFilter));
97 c->round = shift_1[shift-1];
98 // c->round = 1 << (shift - 1);
101 static inline void ttafilter_process(TTAFilter *c, int32_t *in)
103 register int32_t *dl = c->dl, *qm = c->qm, *dx = c->dx, sum = c->round;
106 qm[0] -= dx[0]; qm[1] -= dx[1]; qm[2] -= dx[2]; qm[3] -= dx[3];
107 qm[4] -= dx[4]; qm[5] -= dx[5]; qm[6] -= dx[6]; qm[7] -= dx[7];
108 } else if (c->error > 0) {
109 qm[0] += dx[0]; qm[1] += dx[1]; qm[2] += dx[2]; qm[3] += dx[3];
110 qm[4] += dx[4]; qm[5] += dx[5]; qm[6] += dx[6]; qm[7] += dx[7];
113 sum += dl[0] * qm[0] + dl[1] * qm[1] + dl[2] * qm[2] + dl[3] * qm[3] +
114 dl[4] * qm[4] + dl[5] * qm[5] + dl[6] * qm[6] + dl[7] * qm[7];
116 dx[0] = dx[1]; dx[1] = dx[2]; dx[2] = dx[3]; dx[3] = dx[4];
117 dl[0] = dl[1]; dl[1] = dl[2]; dl[2] = dl[3]; dl[3] = dl[4];
119 dx[4] = ((dl[4] >> 30) | 1);
120 dx[5] = ((dl[5] >> 30) | 2) & ~1;
121 dx[6] = ((dl[6] >> 30) | 2) & ~1;
122 dx[7] = ((dl[7] >> 30) | 4) & ~3;
125 *in += (sum >> c->shift);
127 dl[4] = -dl[5]; dl[5] = -dl[6];
128 dl[6] = *in - dl[7]; dl[7] = *in;
129 dl[5] += dl[6]; dl[4] += dl[5];
132 static void rice_init(TTARice *c, uint32_t k0, uint32_t k1)
136 c->sum0 = shift_16[k0];
137 c->sum1 = shift_16[k1];
140 static int tta_get_unary(GetBitContext *gb)
145 while (get_bits_left(gb) > 0 && get_bits1(gb))
150 static const int64_t tta_channel_layouts[7] = {
152 AV_CH_LAYOUT_STEREO|AV_CH_LOW_FREQUENCY,
155 AV_CH_LAYOUT_5POINT1_BACK,
156 AV_CH_LAYOUT_5POINT1_BACK|AV_CH_BACK_CENTER,
157 AV_CH_LAYOUT_7POINT1_WIDE
160 static int tta_check_crc(TTAContext *s, const uint8_t *buf, int buf_size)
164 CRC = AV_RL32(buf + buf_size);
165 crc = av_crc(s->crc_table, 0xFFFFFFFFU, buf, buf_size);
166 if (CRC != (crc ^ 0xFFFFFFFFU)) {
167 av_log(s->avctx, AV_LOG_ERROR, "CRC error\n");
168 return AVERROR_INVALIDDATA;
174 static av_cold int tta_decode_init(AVCodecContext * avctx)
176 TTAContext *s = avctx->priv_data;
180 // 30bytes includes a seektable with one frame
181 if (avctx->extradata_size < 30)
184 init_get_bits(&s->gb, avctx->extradata, avctx->extradata_size * 8);
185 if (show_bits_long(&s->gb, 32) == AV_RL32("TTA1"))
187 if (avctx->err_recognition & AV_EF_CRCCHECK) {
188 s->crc_table = av_crc_get_table(AV_CRC_32_IEEE_LE);
189 tta_check_crc(s, avctx->extradata, 18);
193 skip_bits_long(&s->gb, 32);
195 s->format = get_bits(&s->gb, 16);
197 av_log(s->avctx, AV_LOG_ERROR, "Invalid format\n");
200 if (s->format == FORMAT_ENCRYPTED) {
201 av_log_missing_feature(s->avctx, "Encrypted TTA", 0);
202 return AVERROR(EINVAL);
204 avctx->channels = s->channels = get_bits(&s->gb, 16);
205 if (s->channels > 1 && s->channels < 9)
206 avctx->channel_layout = tta_channel_layouts[s->channels-2];
207 avctx->bits_per_coded_sample = get_bits(&s->gb, 16);
208 s->bps = (avctx->bits_per_coded_sample + 7) / 8;
209 avctx->sample_rate = get_bits_long(&s->gb, 32);
210 s->data_length = get_bits_long(&s->gb, 32);
211 skip_bits_long(&s->gb, 32); // CRC32 of header
213 if (s->channels == 0) {
214 av_log(s->avctx, AV_LOG_ERROR, "Invalid number of channels\n");
215 return AVERROR_INVALIDDATA;
216 } else if (avctx->sample_rate == 0) {
217 av_log(s->avctx, AV_LOG_ERROR, "Invalid samplerate\n");
218 return AVERROR_INVALIDDATA;
222 case 1: avctx->sample_fmt = AV_SAMPLE_FMT_U8; break;
224 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
225 avctx->bits_per_raw_sample = 16;
228 avctx->sample_fmt = AV_SAMPLE_FMT_S32;
229 avctx->bits_per_raw_sample = 24;
231 //case 4: avctx->sample_fmt = AV_SAMPLE_FMT_S32; break;
233 av_log(avctx, AV_LOG_ERROR, "Invalid/unsupported sample format.\n");
234 return AVERROR_INVALIDDATA;
238 if (avctx->sample_rate > 0x7FFFFF) {
239 av_log(avctx, AV_LOG_ERROR, "sample_rate too large\n");
240 return AVERROR(EINVAL);
242 s->frame_length = 256 * avctx->sample_rate / 245;
244 s->last_frame_length = s->data_length % s->frame_length;
245 s->total_frames = s->data_length / s->frame_length +
246 (s->last_frame_length ? 1 : 0);
248 av_log(s->avctx, AV_LOG_DEBUG, "format: %d chans: %d bps: %d rate: %d block: %d\n",
249 s->format, avctx->channels, avctx->bits_per_coded_sample, avctx->sample_rate,
251 av_log(s->avctx, AV_LOG_DEBUG, "data_length: %d frame_length: %d last: %d total: %d\n",
252 s->data_length, s->frame_length, s->last_frame_length, s->total_frames);
254 if (s->total_frames < 0)
255 return AVERROR_INVALIDDATA;
258 if (get_bits_left(&s->gb) < 32 * s->total_frames + 32)
259 av_log(avctx, AV_LOG_WARNING, "Seek table missing or too small\n");
260 else if (avctx->err_recognition & AV_EF_CRCCHECK) {
261 if (avctx->extradata_size < 26 + s->total_frames * 4 || tta_check_crc(s, avctx->extradata + 22, s->total_frames * 4))
262 return AVERROR_INVALIDDATA;
264 skip_bits_long(&s->gb, 32 * s->total_frames);
265 skip_bits_long(&s->gb, 32); // CRC32 of seektable
267 if(s->frame_length >= UINT_MAX / (s->channels * sizeof(int32_t))){
268 av_log(avctx, AV_LOG_ERROR, "frame_length too large\n");
273 s->decode_buffer = av_mallocz(sizeof(int32_t)*s->frame_length*s->channels);
274 if (!s->decode_buffer)
275 return AVERROR(ENOMEM);
277 s->decode_buffer = NULL;
278 s->ch_ctx = av_malloc(avctx->channels * sizeof(*s->ch_ctx));
280 av_freep(&s->decode_buffer);
281 return AVERROR(ENOMEM);
284 av_log(avctx, AV_LOG_ERROR, "Wrong extradata present\n");
288 avcodec_get_frame_defaults(&s->frame);
289 avctx->coded_frame = &s->frame;
294 static int tta_decode_frame(AVCodecContext *avctx, void *data,
295 int *got_frame_ptr, AVPacket *avpkt)
297 const uint8_t *buf = avpkt->data;
298 int buf_size = avpkt->size;
299 TTAContext *s = avctx->priv_data;
301 int cur_chan = 0, framelen = s->frame_length;
304 if (avctx->err_recognition & AV_EF_CRCCHECK) {
305 if (buf_size < 4 || tta_check_crc(s, buf, buf_size - 4))
306 return AVERROR_INVALIDDATA;
309 init_get_bits(&s->gb, buf, buf_size*8);
313 if (!s->total_frames && s->last_frame_length)
314 framelen = s->last_frame_length;
316 /* get output buffer */
317 s->frame.nb_samples = framelen;
318 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
319 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
323 // decode directly to output buffer for 24-bit sample format
325 s->decode_buffer = (int32_t *)s->frame.data[0];
327 // init per channel states
328 for (i = 0; i < s->channels; i++) {
329 s->ch_ctx[i].predictor = 0;
330 ttafilter_init(&s->ch_ctx[i].filter, ttafilter_configs[s->bps-1]);
331 rice_init(&s->ch_ctx[i].rice, 10, 10);
334 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++) {
335 int32_t *predictor = &s->ch_ctx[cur_chan].predictor;
336 TTAFilter *filter = &s->ch_ctx[cur_chan].filter;
337 TTARice *rice = &s->ch_ctx[cur_chan].rice;
338 uint32_t unary, depth, k;
341 unary = tta_get_unary(&s->gb);
352 if (get_bits_left(&s->gb) < k) {
353 ret = AVERROR_INVALIDDATA;
358 if (k > MIN_CACHE_BITS) {
359 ret = AVERROR_INVALIDDATA;
362 value = (unary << k) + get_bits(&s->gb, k);
366 // FIXME: copy paste from original
369 rice->sum1 += value - (rice->sum1 >> 4);
370 if (rice->k1 > 0 && rice->sum1 < shift_16[rice->k1])
372 else if(rice->sum1 > shift_16[rice->k1 + 1])
374 value += shift_1[rice->k0];
376 rice->sum0 += value - (rice->sum0 >> 4);
377 if (rice->k0 > 0 && rice->sum0 < shift_16[rice->k0])
379 else if(rice->sum0 > shift_16[rice->k0 + 1])
383 // extract coded value
384 *p = 1 + ((value >> 1) ^ ((value & 1) - 1));
387 ttafilter_process(filter, p);
389 // fixed order prediction
390 #define PRED(x, k) (int32_t)((((uint64_t)x << k) - x) >> k)
392 case 1: *p += PRED(*predictor, 4); break;
394 case 3: *p += PRED(*predictor, 5); break;
395 case 4: *p += *predictor; break;
400 if (cur_chan < (s->channels-1))
403 // decorrelate in case of stereo integer
404 if (s->channels > 1) {
406 for (*p += *r / 2; r > p - s->channels; r--)
413 if (get_bits_left(&s->gb) < 32) {
414 ret = AVERROR_INVALIDDATA;
417 skip_bits_long(&s->gb, 32); // frame crc
419 // convert to output buffer
422 uint8_t *samples = (uint8_t *)s->frame.data[0];
423 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
424 *samples++ = *p + 0x80;
428 uint16_t *samples = (int16_t *)s->frame.data[0];
429 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
434 // shift samples for 24-bit sample format
435 int32_t *samples = (int32_t *)s->frame.data[0];
436 for (p = s->decode_buffer; p < s->decode_buffer + (framelen * s->channels); p++)
438 // reset decode buffer
439 s->decode_buffer = NULL;
445 *(AVFrame *)data = s->frame;
449 // reset decode buffer
451 s->decode_buffer = NULL;
455 static av_cold int tta_decode_close(AVCodecContext *avctx) {
456 TTAContext *s = avctx->priv_data;
459 av_free(s->decode_buffer);
460 s->decode_buffer = NULL;
461 av_freep(&s->ch_ctx);
466 AVCodec ff_tta_decoder = {
468 .type = AVMEDIA_TYPE_AUDIO,
470 .priv_data_size = sizeof(TTAContext),
471 .init = tta_decode_init,
472 .close = tta_decode_close,
473 .decode = tta_decode_frame,
474 .capabilities = CODEC_CAP_DR1,
475 .long_name = NULL_IF_CONFIG_SMALL("True Audio (TTA)"),