ffmpeg / libavcodec / dca.c @ b2755007
History | View | Annotate | Download (48.2 KB)
1 |
/*
|
---|---|
2 |
* DCA compatible decoder
|
3 |
* Copyright (C) 2004 Gildas Bazin
|
4 |
* Copyright (C) 2004 Benjamin Zores
|
5 |
* Copyright (C) 2006 Benjamin Larsson
|
6 |
* Copyright (C) 2007 Konstantin Shishkov
|
7 |
*
|
8 |
* This file is part of FFmpeg.
|
9 |
*
|
10 |
* FFmpeg is free software; you can redistribute it and/or
|
11 |
* modify it under the terms of the GNU Lesser General Public
|
12 |
* License as published by the Free Software Foundation; either
|
13 |
* version 2.1 of the License, or (at your option) any later version.
|
14 |
*
|
15 |
* FFmpeg is distributed in the hope that it will be useful,
|
16 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
17 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
18 |
* Lesser General Public License for more details.
|
19 |
*
|
20 |
* You should have received a copy of the GNU Lesser General Public
|
21 |
* License along with FFmpeg; if not, write to the Free Software
|
22 |
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
23 |
*/
|
24 |
|
25 |
/**
|
26 |
* @file libavcodec/dca.c
|
27 |
*/
|
28 |
|
29 |
#include <math.h> |
30 |
#include <stddef.h> |
31 |
#include <stdio.h> |
32 |
|
33 |
#include "avcodec.h" |
34 |
#include "dsputil.h" |
35 |
#include "bitstream.h" |
36 |
#include "put_bits.h" |
37 |
#include "dcadata.h" |
38 |
#include "dcahuff.h" |
39 |
#include "dca.h" |
40 |
|
41 |
//#define TRACE
|
42 |
|
43 |
#define DCA_PRIM_CHANNELS_MAX (5) |
44 |
#define DCA_SUBBANDS (32) |
45 |
#define DCA_ABITS_MAX (32) /* Should be 28 */ |
46 |
#define DCA_SUBSUBFAMES_MAX (4) |
47 |
#define DCA_LFE_MAX (3) |
48 |
|
49 |
enum DCAMode {
|
50 |
DCA_MONO = 0,
|
51 |
DCA_CHANNEL, |
52 |
DCA_STEREO, |
53 |
DCA_STEREO_SUMDIFF, |
54 |
DCA_STEREO_TOTAL, |
55 |
DCA_3F, |
56 |
DCA_2F1R, |
57 |
DCA_3F1R, |
58 |
DCA_2F2R, |
59 |
DCA_3F2R, |
60 |
DCA_4F2R |
61 |
}; |
62 |
|
63 |
/* Tables for mapping dts channel configurations to libavcodec multichannel api.
|
64 |
* Some compromises have been made for special configurations. Most configurations
|
65 |
* are never used so complete accuracy is not needed.
|
66 |
*
|
67 |
* L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
|
68 |
* S -> side, when both rear and back are configured move one of them to the side channel
|
69 |
* OV -> center back
|
70 |
* All 2 channel configurations -> CH_LAYOUT_STEREO
|
71 |
*/
|
72 |
|
73 |
static const int64_t dca_core_channel_layout[] = { |
74 |
CH_FRONT_CENTER, ///< 1, A
|
75 |
CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
|
76 |
CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
|
77 |
CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference)
|
78 |
CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total)
|
79 |
CH_LAYOUT_STEREO|CH_FRONT_CENTER, ///< 3, C+L+R
|
80 |
CH_LAYOUT_STEREO|CH_BACK_CENTER, ///< 3, L+R+S
|
81 |
CH_LAYOUT_STEREO|CH_FRONT_CENTER|CH_BACK_CENTER, ///< 4, C + L + R+ S
|
82 |
CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR
|
83 |
CH_LAYOUT_STEREO|CH_FRONT_CENTER|CH_SIDE_LEFT|CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR
|
84 |
CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT|CH_FRONT_LEFT_OF_CENTER|CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
|
85 |
CH_LAYOUT_STEREO|CH_BACK_LEFT|CH_BACK_RIGHT|CH_FRONT_CENTER|CH_BACK_CENTER, ///< 6, C + L + R+ LR + RR + OV
|
86 |
CH_FRONT_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_FRONT_LEFT_OF_CENTER|CH_BACK_CENTER|CH_BACK_LEFT|CH_BACK_RIGHT, ///< 6, CF+ CR+LF+ RF+LR + RR
|
87 |
CH_FRONT_LEFT_OF_CENTER|CH_FRONT_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
|
88 |
CH_FRONT_LEFT_OF_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_SIDE_RIGHT|CH_BACK_LEFT|CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2+ SR1 + SR2
|
89 |
CH_FRONT_LEFT_OF_CENTER|CH_FRONT_CENTER|CH_FRONT_RIGHT_OF_CENTER|CH_LAYOUT_STEREO|CH_SIDE_LEFT|CH_BACK_CENTER|CH_SIDE_RIGHT, ///< 8, CL + C+ CR + L + R + SL + S+ SR
|
90 |
}; |
91 |
|
92 |
static const int8_t dca_lfe_index[] = { |
93 |
1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3 |
94 |
}; |
95 |
|
96 |
static const int8_t dca_channel_reorder_lfe[][8] = { |
97 |
{ 0, -1, -1, -1, -1, -1, -1, -1}, |
98 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
99 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
100 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
101 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
102 |
{ 2, 0, 1, -1, -1, -1, -1, -1}, |
103 |
{ 0, 1, 3, -1, -1, -1, -1, -1}, |
104 |
{ 2, 0, 1, 4, -1, -1, -1, -1}, |
105 |
{ 0, 1, 3, 4, -1, -1, -1, -1}, |
106 |
{ 2, 0, 1, 4, 5, -1, -1, -1}, |
107 |
{ 3, 4, 0, 1, 5, 6, -1, -1}, |
108 |
{ 2, 0, 1, 4, 5, 6, -1, -1}, |
109 |
{ 0, 6, 4, 5, 2, 3, -1, -1}, |
110 |
{ 4, 2, 5, 0, 1, 6, 7, -1}, |
111 |
{ 5, 6, 0, 1, 7, 3, 8, 4}, |
112 |
{ 4, 2, 5, 0, 1, 6, 8, 7}, |
113 |
}; |
114 |
|
115 |
static const int8_t dca_channel_reorder_nolfe[][8] = { |
116 |
{ 0, -1, -1, -1, -1, -1, -1, -1}, |
117 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
118 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
119 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
120 |
{ 0, 1, -1, -1, -1, -1, -1, -1}, |
121 |
{ 2, 0, 1, -1, -1, -1, -1, -1}, |
122 |
{ 0, 1, 2, -1, -1, -1, -1, -1}, |
123 |
{ 2, 0, 1, 3, -1, -1, -1, -1}, |
124 |
{ 0, 1, 2, 3, -1, -1, -1, -1}, |
125 |
{ 2, 0, 1, 3, 4, -1, -1, -1}, |
126 |
{ 2, 3, 0, 1, 4, 5, -1, -1}, |
127 |
{ 2, 0, 1, 3, 4, 5, -1, -1}, |
128 |
{ 0, 5, 3, 4, 1, 2, -1, -1}, |
129 |
{ 3, 2, 4, 0, 1, 5, 6, -1}, |
130 |
{ 4, 5, 0, 1, 6, 2, 7, 3}, |
131 |
{ 3, 2, 4, 0, 1, 5, 7, 6}, |
132 |
}; |
133 |
|
134 |
|
135 |
#define DCA_DOLBY 101 /* FIXME */ |
136 |
|
137 |
#define DCA_CHANNEL_BITS 6 |
138 |
#define DCA_CHANNEL_MASK 0x3F |
139 |
|
140 |
#define DCA_LFE 0x80 |
141 |
|
142 |
#define HEADER_SIZE 14 |
143 |
|
144 |
#define DCA_MAX_FRAME_SIZE 16384 |
145 |
|
146 |
/** Bit allocation */
|
147 |
typedef struct { |
148 |
int offset; ///< code values offset |
149 |
int maxbits[8]; ///< max bits in VLC |
150 |
int wrap; ///< wrap for get_vlc2() |
151 |
VLC vlc[8]; ///< actual codes |
152 |
} BitAlloc; |
153 |
|
154 |
static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select |
155 |
static BitAlloc dca_tmode; ///< transition mode VLCs |
156 |
static BitAlloc dca_scalefactor; ///< scalefactor VLCs |
157 |
static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs |
158 |
|
159 |
static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx) |
160 |
{ |
161 |
return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
|
162 |
} |
163 |
|
164 |
typedef struct { |
165 |
AVCodecContext *avctx; |
166 |
/* Frame header */
|
167 |
int frame_type; ///< type of the current frame |
168 |
int samples_deficit; ///< deficit sample count |
169 |
int crc_present; ///< crc is present in the bitstream |
170 |
int sample_blocks; ///< number of PCM sample blocks |
171 |
int frame_size; ///< primary frame byte size |
172 |
int amode; ///< audio channels arrangement |
173 |
int sample_rate; ///< audio sampling rate |
174 |
int bit_rate; ///< transmission bit rate |
175 |
int bit_rate_index; ///< transmission bit rate index |
176 |
|
177 |
int downmix; ///< embedded downmix enabled |
178 |
int dynrange; ///< embedded dynamic range flag |
179 |
int timestamp; ///< embedded time stamp flag |
180 |
int aux_data; ///< auxiliary data flag |
181 |
int hdcd; ///< source material is mastered in HDCD |
182 |
int ext_descr; ///< extension audio descriptor flag |
183 |
int ext_coding; ///< extended coding flag |
184 |
int aspf; ///< audio sync word insertion flag |
185 |
int lfe; ///< low frequency effects flag |
186 |
int predictor_history; ///< predictor history flag |
187 |
int header_crc; ///< header crc check bytes |
188 |
int multirate_inter; ///< multirate interpolator switch |
189 |
int version; ///< encoder software revision |
190 |
int copy_history; ///< copy history |
191 |
int source_pcm_res; ///< source pcm resolution |
192 |
int front_sum; ///< front sum/difference flag |
193 |
int surround_sum; ///< surround sum/difference flag |
194 |
int dialog_norm; ///< dialog normalisation parameter |
195 |
|
196 |
/* Primary audio coding header */
|
197 |
int subframes; ///< number of subframes |
198 |
int total_channels; ///< number of channels including extensions |
199 |
int prim_channels; ///< number of primary audio channels |
200 |
int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count |
201 |
int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband |
202 |
int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index |
203 |
int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book |
204 |
int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book |
205 |
int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select |
206 |
int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select |
207 |
float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment |
208 |
|
209 |
/* Primary audio coding side information */
|
210 |
int subsubframes; ///< number of subsubframes |
211 |
int partial_samples; ///< partial subsubframe samples count |
212 |
int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not) |
213 |
int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs |
214 |
int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index |
215 |
int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients) |
216 |
int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient) |
217 |
int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook |
218 |
int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors |
219 |
int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients |
220 |
int dynrange_coef; ///< dynamic range coefficient |
221 |
|
222 |
int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands |
223 |
|
224 |
float lfe_data[2 * DCA_SUBSUBFAMES_MAX * DCA_LFE_MAX * |
225 |
2 /*history */ ]; ///< Low frequency effect data |
226 |
int lfe_scale_factor;
|
227 |
|
228 |
/* Subband samples history (for ADPCM) */
|
229 |
float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4]; |
230 |
DECLARE_ALIGNED_16(float, subband_fir_hist[DCA_PRIM_CHANNELS_MAX][512]); |
231 |
float subband_fir_noidea[DCA_PRIM_CHANNELS_MAX][32]; |
232 |
int hist_index[DCA_PRIM_CHANNELS_MAX];
|
233 |
|
234 |
int output; ///< type of output |
235 |
float add_bias; ///< output bias |
236 |
float scale_bias; ///< output scale |
237 |
|
238 |
DECLARE_ALIGNED_16(float, samples[1536]); /* 6 * 256 = 1536, might only need 5 */ |
239 |
const float *samples_chanptr[6]; |
240 |
|
241 |
uint8_t dca_buffer[DCA_MAX_FRAME_SIZE]; |
242 |
int dca_buffer_size; ///< how much data is in the dca_buffer |
243 |
|
244 |
const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe |
245 |
GetBitContext gb; |
246 |
/* Current position in DCA frame */
|
247 |
int current_subframe;
|
248 |
int current_subsubframe;
|
249 |
|
250 |
int debug_flag; ///< used for suppressing repeated error messages output |
251 |
DSPContext dsp; |
252 |
MDCTContext imdct; |
253 |
} DCAContext; |
254 |
|
255 |
static av_cold void dca_init_vlcs(void) |
256 |
{ |
257 |
static int vlcs_initialized = 0; |
258 |
int i, j;
|
259 |
|
260 |
if (vlcs_initialized)
|
261 |
return;
|
262 |
|
263 |
dca_bitalloc_index.offset = 1;
|
264 |
dca_bitalloc_index.wrap = 2;
|
265 |
for (i = 0; i < 5; i++) |
266 |
init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
|
267 |
bitalloc_12_bits[i], 1, 1, |
268 |
bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_STATIC); |
269 |
dca_scalefactor.offset = -64;
|
270 |
dca_scalefactor.wrap = 2;
|
271 |
for (i = 0; i < 5; i++) |
272 |
init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
|
273 |
scales_bits[i], 1, 1, |
274 |
scales_codes[i], 2, 2, INIT_VLC_USE_STATIC); |
275 |
dca_tmode.offset = 0;
|
276 |
dca_tmode.wrap = 1;
|
277 |
for (i = 0; i < 4; i++) |
278 |
init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
|
279 |
tmode_bits[i], 1, 1, |
280 |
tmode_codes[i], 2, 2, INIT_VLC_USE_STATIC); |
281 |
|
282 |
for(i = 0; i < 10; i++) |
283 |
for(j = 0; j < 7; j++){ |
284 |
if(!bitalloc_codes[i][j]) break; |
285 |
dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
|
286 |
dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4); |
287 |
init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
|
288 |
bitalloc_sizes[i], |
289 |
bitalloc_bits[i][j], 1, 1, |
290 |
bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_STATIC); |
291 |
} |
292 |
vlcs_initialized = 1;
|
293 |
} |
294 |
|
295 |
static inline void get_array(GetBitContext *gb, int *dst, int len, int bits) |
296 |
{ |
297 |
while(len--)
|
298 |
*dst++ = get_bits(gb, bits); |
299 |
} |
300 |
|
301 |
static int dca_parse_frame_header(DCAContext * s) |
302 |
{ |
303 |
int i, j;
|
304 |
static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 }; |
305 |
static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 }; |
306 |
static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 }; |
307 |
|
308 |
init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
|
309 |
|
310 |
/* Sync code */
|
311 |
get_bits(&s->gb, 32);
|
312 |
|
313 |
/* Frame header */
|
314 |
s->frame_type = get_bits(&s->gb, 1);
|
315 |
s->samples_deficit = get_bits(&s->gb, 5) + 1; |
316 |
s->crc_present = get_bits(&s->gb, 1);
|
317 |
s->sample_blocks = get_bits(&s->gb, 7) + 1; |
318 |
s->frame_size = get_bits(&s->gb, 14) + 1; |
319 |
if (s->frame_size < 95) |
320 |
return -1; |
321 |
s->amode = get_bits(&s->gb, 6);
|
322 |
s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
|
323 |
if (!s->sample_rate)
|
324 |
return -1; |
325 |
s->bit_rate_index = get_bits(&s->gb, 5);
|
326 |
s->bit_rate = dca_bit_rates[s->bit_rate_index]; |
327 |
if (!s->bit_rate)
|
328 |
return -1; |
329 |
|
330 |
s->downmix = get_bits(&s->gb, 1);
|
331 |
s->dynrange = get_bits(&s->gb, 1);
|
332 |
s->timestamp = get_bits(&s->gb, 1);
|
333 |
s->aux_data = get_bits(&s->gb, 1);
|
334 |
s->hdcd = get_bits(&s->gb, 1);
|
335 |
s->ext_descr = get_bits(&s->gb, 3);
|
336 |
s->ext_coding = get_bits(&s->gb, 1);
|
337 |
s->aspf = get_bits(&s->gb, 1);
|
338 |
s->lfe = get_bits(&s->gb, 2);
|
339 |
s->predictor_history = get_bits(&s->gb, 1);
|
340 |
|
341 |
/* TODO: check CRC */
|
342 |
if (s->crc_present)
|
343 |
s->header_crc = get_bits(&s->gb, 16);
|
344 |
|
345 |
s->multirate_inter = get_bits(&s->gb, 1);
|
346 |
s->version = get_bits(&s->gb, 4);
|
347 |
s->copy_history = get_bits(&s->gb, 2);
|
348 |
s->source_pcm_res = get_bits(&s->gb, 3);
|
349 |
s->front_sum = get_bits(&s->gb, 1);
|
350 |
s->surround_sum = get_bits(&s->gb, 1);
|
351 |
s->dialog_norm = get_bits(&s->gb, 4);
|
352 |
|
353 |
/* FIXME: channels mixing levels */
|
354 |
s->output = s->amode; |
355 |
if(s->lfe) s->output |= DCA_LFE;
|
356 |
|
357 |
#ifdef TRACE
|
358 |
av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
|
359 |
av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
|
360 |
av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
|
361 |
av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
|
362 |
s->sample_blocks, s->sample_blocks * 32);
|
363 |
av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
|
364 |
av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
|
365 |
s->amode, dca_channels[s->amode]); |
366 |
av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
|
367 |
s->sample_rate); |
368 |
av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
|
369 |
s->bit_rate); |
370 |
av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
|
371 |
av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
|
372 |
av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
|
373 |
av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
|
374 |
av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
|
375 |
av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
|
376 |
av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
|
377 |
av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
|
378 |
av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
|
379 |
av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
|
380 |
s->predictor_history); |
381 |
av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
|
382 |
av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
|
383 |
s->multirate_inter); |
384 |
av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
|
385 |
av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
|
386 |
av_log(s->avctx, AV_LOG_DEBUG, |
387 |
"source pcm resolution: %i (%i bits/sample)\n",
|
388 |
s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]); |
389 |
av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
|
390 |
av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
|
391 |
av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
|
392 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
393 |
#endif
|
394 |
|
395 |
/* Primary audio coding header */
|
396 |
s->subframes = get_bits(&s->gb, 4) + 1; |
397 |
s->total_channels = get_bits(&s->gb, 3) + 1; |
398 |
s->prim_channels = s->total_channels; |
399 |
if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
|
400 |
s->prim_channels = DCA_PRIM_CHANNELS_MAX; /* We only support DTS core */
|
401 |
|
402 |
|
403 |
for (i = 0; i < s->prim_channels; i++) { |
404 |
s->subband_activity[i] = get_bits(&s->gb, 5) + 2; |
405 |
if (s->subband_activity[i] > DCA_SUBBANDS)
|
406 |
s->subband_activity[i] = DCA_SUBBANDS; |
407 |
} |
408 |
for (i = 0; i < s->prim_channels; i++) { |
409 |
s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1; |
410 |
if (s->vq_start_subband[i] > DCA_SUBBANDS)
|
411 |
s->vq_start_subband[i] = DCA_SUBBANDS; |
412 |
} |
413 |
get_array(&s->gb, s->joint_intensity, s->prim_channels, 3);
|
414 |
get_array(&s->gb, s->transient_huffman, s->prim_channels, 2);
|
415 |
get_array(&s->gb, s->scalefactor_huffman, s->prim_channels, 3);
|
416 |
get_array(&s->gb, s->bitalloc_huffman, s->prim_channels, 3);
|
417 |
|
418 |
/* Get codebooks quantization indexes */
|
419 |
memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman)); |
420 |
for (j = 1; j < 11; j++) |
421 |
for (i = 0; i < s->prim_channels; i++) |
422 |
s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]); |
423 |
|
424 |
/* Get scale factor adjustment */
|
425 |
for (j = 0; j < 11; j++) |
426 |
for (i = 0; i < s->prim_channels; i++) |
427 |
s->scalefactor_adj[i][j] = 1;
|
428 |
|
429 |
for (j = 1; j < 11; j++) |
430 |
for (i = 0; i < s->prim_channels; i++) |
431 |
if (s->quant_index_huffman[i][j] < thr[j])
|
432 |
s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
|
433 |
|
434 |
if (s->crc_present) {
|
435 |
/* Audio header CRC check */
|
436 |
get_bits(&s->gb, 16);
|
437 |
} |
438 |
|
439 |
s->current_subframe = 0;
|
440 |
s->current_subsubframe = 0;
|
441 |
|
442 |
#ifdef TRACE
|
443 |
av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
|
444 |
av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
|
445 |
for(i = 0; i < s->prim_channels; i++){ |
446 |
av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
|
447 |
av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
|
448 |
av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
|
449 |
av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
|
450 |
av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
|
451 |
av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
|
452 |
av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
|
453 |
for (j = 0; j < 11; j++) |
454 |
av_log(s->avctx, AV_LOG_DEBUG, " %i",
|
455 |
s->quant_index_huffman[i][j]); |
456 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
457 |
av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
|
458 |
for (j = 0; j < 11; j++) |
459 |
av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
|
460 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
461 |
} |
462 |
#endif
|
463 |
|
464 |
return 0; |
465 |
} |
466 |
|
467 |
|
468 |
static inline int get_scale(GetBitContext *gb, int level, int value) |
469 |
{ |
470 |
if (level < 5) { |
471 |
/* huffman encoded */
|
472 |
value += get_bitalloc(gb, &dca_scalefactor, level); |
473 |
} else if(level < 8) |
474 |
value = get_bits(gb, level + 1);
|
475 |
return value;
|
476 |
} |
477 |
|
478 |
static int dca_subframe_header(DCAContext * s) |
479 |
{ |
480 |
/* Primary audio coding side information */
|
481 |
int j, k;
|
482 |
|
483 |
s->subsubframes = get_bits(&s->gb, 2) + 1; |
484 |
s->partial_samples = get_bits(&s->gb, 3);
|
485 |
for (j = 0; j < s->prim_channels; j++) { |
486 |
for (k = 0; k < s->subband_activity[j]; k++) |
487 |
s->prediction_mode[j][k] = get_bits(&s->gb, 1);
|
488 |
} |
489 |
|
490 |
/* Get prediction codebook */
|
491 |
for (j = 0; j < s->prim_channels; j++) { |
492 |
for (k = 0; k < s->subband_activity[j]; k++) { |
493 |
if (s->prediction_mode[j][k] > 0) { |
494 |
/* (Prediction coefficient VQ address) */
|
495 |
s->prediction_vq[j][k] = get_bits(&s->gb, 12);
|
496 |
} |
497 |
} |
498 |
} |
499 |
|
500 |
/* Bit allocation index */
|
501 |
for (j = 0; j < s->prim_channels; j++) { |
502 |
for (k = 0; k < s->vq_start_subband[j]; k++) { |
503 |
if (s->bitalloc_huffman[j] == 6) |
504 |
s->bitalloc[j][k] = get_bits(&s->gb, 5);
|
505 |
else if (s->bitalloc_huffman[j] == 5) |
506 |
s->bitalloc[j][k] = get_bits(&s->gb, 4);
|
507 |
else if (s->bitalloc_huffman[j] == 7) { |
508 |
av_log(s->avctx, AV_LOG_ERROR, |
509 |
"Invalid bit allocation index\n");
|
510 |
return -1; |
511 |
} else {
|
512 |
s->bitalloc[j][k] = |
513 |
get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]); |
514 |
} |
515 |
|
516 |
if (s->bitalloc[j][k] > 26) { |
517 |
// av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
|
518 |
// j, k, s->bitalloc[j][k]);
|
519 |
return -1; |
520 |
} |
521 |
} |
522 |
} |
523 |
|
524 |
/* Transition mode */
|
525 |
for (j = 0; j < s->prim_channels; j++) { |
526 |
for (k = 0; k < s->subband_activity[j]; k++) { |
527 |
s->transition_mode[j][k] = 0;
|
528 |
if (s->subsubframes > 1 && |
529 |
k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
|
530 |
s->transition_mode[j][k] = |
531 |
get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]); |
532 |
} |
533 |
} |
534 |
} |
535 |
|
536 |
for (j = 0; j < s->prim_channels; j++) { |
537 |
const uint32_t *scale_table;
|
538 |
int scale_sum;
|
539 |
|
540 |
memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2); |
541 |
|
542 |
if (s->scalefactor_huffman[j] == 6) |
543 |
scale_table = scale_factor_quant7; |
544 |
else
|
545 |
scale_table = scale_factor_quant6; |
546 |
|
547 |
/* When huffman coded, only the difference is encoded */
|
548 |
scale_sum = 0;
|
549 |
|
550 |
for (k = 0; k < s->subband_activity[j]; k++) { |
551 |
if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) { |
552 |
scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum); |
553 |
s->scale_factor[j][k][0] = scale_table[scale_sum];
|
554 |
} |
555 |
|
556 |
if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
|
557 |
/* Get second scale factor */
|
558 |
scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum); |
559 |
s->scale_factor[j][k][1] = scale_table[scale_sum];
|
560 |
} |
561 |
} |
562 |
} |
563 |
|
564 |
/* Joint subband scale factor codebook select */
|
565 |
for (j = 0; j < s->prim_channels; j++) { |
566 |
/* Transmitted only if joint subband coding enabled */
|
567 |
if (s->joint_intensity[j] > 0) |
568 |
s->joint_huff[j] = get_bits(&s->gb, 3);
|
569 |
} |
570 |
|
571 |
/* Scale factors for joint subband coding */
|
572 |
for (j = 0; j < s->prim_channels; j++) { |
573 |
int source_channel;
|
574 |
|
575 |
/* Transmitted only if joint subband coding enabled */
|
576 |
if (s->joint_intensity[j] > 0) { |
577 |
int scale = 0; |
578 |
source_channel = s->joint_intensity[j] - 1;
|
579 |
|
580 |
/* When huffman coded, only the difference is encoded
|
581 |
* (is this valid as well for joint scales ???) */
|
582 |
|
583 |
for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
|
584 |
scale = get_scale(&s->gb, s->joint_huff[j], 0);
|
585 |
scale += 64; /* bias */ |
586 |
s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
|
587 |
} |
588 |
|
589 |
if (!s->debug_flag & 0x02) { |
590 |
av_log(s->avctx, AV_LOG_DEBUG, |
591 |
"Joint stereo coding not supported\n");
|
592 |
s->debug_flag |= 0x02;
|
593 |
} |
594 |
} |
595 |
} |
596 |
|
597 |
/* Stereo downmix coefficients */
|
598 |
if (s->prim_channels > 2) { |
599 |
if(s->downmix) {
|
600 |
for (j = 0; j < s->prim_channels; j++) { |
601 |
s->downmix_coef[j][0] = get_bits(&s->gb, 7); |
602 |
s->downmix_coef[j][1] = get_bits(&s->gb, 7); |
603 |
} |
604 |
} else {
|
605 |
int am = s->amode & DCA_CHANNEL_MASK;
|
606 |
for (j = 0; j < s->prim_channels; j++) { |
607 |
s->downmix_coef[j][0] = dca_default_coeffs[am][j][0]; |
608 |
s->downmix_coef[j][1] = dca_default_coeffs[am][j][1]; |
609 |
} |
610 |
} |
611 |
} |
612 |
|
613 |
/* Dynamic range coefficient */
|
614 |
if (s->dynrange)
|
615 |
s->dynrange_coef = get_bits(&s->gb, 8);
|
616 |
|
617 |
/* Side information CRC check word */
|
618 |
if (s->crc_present) {
|
619 |
get_bits(&s->gb, 16);
|
620 |
} |
621 |
|
622 |
/*
|
623 |
* Primary audio data arrays
|
624 |
*/
|
625 |
|
626 |
/* VQ encoded high frequency subbands */
|
627 |
for (j = 0; j < s->prim_channels; j++) |
628 |
for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
|
629 |
/* 1 vector -> 32 samples */
|
630 |
s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
|
631 |
|
632 |
/* Low frequency effect data */
|
633 |
if (s->lfe) {
|
634 |
/* LFE samples */
|
635 |
int lfe_samples = 2 * s->lfe * s->subsubframes; |
636 |
float lfe_scale;
|
637 |
|
638 |
for (j = lfe_samples; j < lfe_samples * 2; j++) { |
639 |
/* Signed 8 bits int */
|
640 |
s->lfe_data[j] = get_sbits(&s->gb, 8);
|
641 |
} |
642 |
|
643 |
/* Scale factor index */
|
644 |
s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
|
645 |
|
646 |
/* Quantization step size * scale factor */
|
647 |
lfe_scale = 0.035 * s->lfe_scale_factor; |
648 |
|
649 |
for (j = lfe_samples; j < lfe_samples * 2; j++) |
650 |
s->lfe_data[j] *= lfe_scale; |
651 |
} |
652 |
|
653 |
#ifdef TRACE
|
654 |
av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes);
|
655 |
av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
|
656 |
s->partial_samples); |
657 |
for (j = 0; j < s->prim_channels; j++) { |
658 |
av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
|
659 |
for (k = 0; k < s->subband_activity[j]; k++) |
660 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
|
661 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
662 |
} |
663 |
for (j = 0; j < s->prim_channels; j++) { |
664 |
for (k = 0; k < s->subband_activity[j]; k++) |
665 |
av_log(s->avctx, AV_LOG_DEBUG, |
666 |
"prediction coefs: %f, %f, %f, %f\n",
|
667 |
(float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192, |
668 |
(float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192, |
669 |
(float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192, |
670 |
(float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192); |
671 |
} |
672 |
for (j = 0; j < s->prim_channels; j++) { |
673 |
av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
|
674 |
for (k = 0; k < s->vq_start_subband[j]; k++) |
675 |
av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
|
676 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
677 |
} |
678 |
for (j = 0; j < s->prim_channels; j++) { |
679 |
av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
|
680 |
for (k = 0; k < s->subband_activity[j]; k++) |
681 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
|
682 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
683 |
} |
684 |
for (j = 0; j < s->prim_channels; j++) { |
685 |
av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
|
686 |
for (k = 0; k < s->subband_activity[j]; k++) { |
687 |
if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) |
688 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]); |
689 |
if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
|
690 |
av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]); |
691 |
} |
692 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
693 |
} |
694 |
for (j = 0; j < s->prim_channels; j++) { |
695 |
if (s->joint_intensity[j] > 0) { |
696 |
int source_channel = s->joint_intensity[j] - 1; |
697 |
av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
|
698 |
for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
|
699 |
av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
|
700 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
701 |
} |
702 |
} |
703 |
if (s->prim_channels > 2 && s->downmix) { |
704 |
av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
|
705 |
for (j = 0; j < s->prim_channels; j++) { |
706 |
av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]); |
707 |
av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]); |
708 |
} |
709 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
710 |
} |
711 |
for (j = 0; j < s->prim_channels; j++) |
712 |
for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
|
713 |
av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
|
714 |
if(s->lfe){
|
715 |
int lfe_samples = 2 * s->lfe * s->subsubframes; |
716 |
av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
|
717 |
for (j = lfe_samples; j < lfe_samples * 2; j++) |
718 |
av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
|
719 |
av_log(s->avctx, AV_LOG_DEBUG, "\n");
|
720 |
} |
721 |
#endif
|
722 |
|
723 |
return 0; |
724 |
} |
725 |
|
726 |
static void qmf_32_subbands(DCAContext * s, int chans, |
727 |
float samples_in[32][8], float *samples_out, |
728 |
float scale, float bias) |
729 |
{ |
730 |
const float *prCoeff; |
731 |
int i, j;
|
732 |
DECLARE_ALIGNED_16(float, raXin[32]); |
733 |
|
734 |
int hist_index= s->hist_index[chans];
|
735 |
float *subband_fir_hist2 = s->subband_fir_noidea[chans];
|
736 |
|
737 |
int subindex;
|
738 |
|
739 |
scale *= sqrt(1/8.0); |
740 |
|
741 |
/* Select filter */
|
742 |
if (!s->multirate_inter) /* Non-perfect reconstruction */ |
743 |
prCoeff = fir_32bands_nonperfect; |
744 |
else /* Perfect reconstruction */ |
745 |
prCoeff = fir_32bands_perfect; |
746 |
|
747 |
/* Reconstructed channel sample index */
|
748 |
for (subindex = 0; subindex < 8; subindex++) { |
749 |
float *subband_fir_hist = s->subband_fir_hist[chans] + hist_index;
|
750 |
/* Load in one sample from each subband and clear inactive subbands */
|
751 |
for (i = 0; i < s->subband_activity[chans]; i++){ |
752 |
if((i-1)&2) raXin[i] = -samples_in[i][subindex]; |
753 |
else raXin[i] = samples_in[i][subindex];
|
754 |
} |
755 |
for (; i < 32; i++) |
756 |
raXin[i] = 0.0; |
757 |
|
758 |
ff_imdct_half(&s->imdct, subband_fir_hist, raXin); |
759 |
|
760 |
/* Multiply by filter coefficients */
|
761 |
for (i = 0; i < 16; i++){ |
762 |
float a= subband_fir_hist2[i ];
|
763 |
float b= subband_fir_hist2[i+16]; |
764 |
float c= 0; |
765 |
float d= 0; |
766 |
for (j = 0; j < 512-hist_index; j += 64){ |
767 |
a += prCoeff[i+j ]*(-subband_fir_hist[15-i+j]);
|
768 |
b += prCoeff[i+j+16]*( subband_fir_hist[ i+j]);
|
769 |
c += prCoeff[i+j+32]*( subband_fir_hist[16+i+j]); |
770 |
d += prCoeff[i+j+48]*( subband_fir_hist[31-i+j]); |
771 |
} |
772 |
for ( ; j < 512; j += 64){ |
773 |
a += prCoeff[i+j ]*(-subband_fir_hist[15-i+j-512]); |
774 |
b += prCoeff[i+j+16]*( subband_fir_hist[ i+j-512]); |
775 |
c += prCoeff[i+j+32]*( subband_fir_hist[16+i+j-512]); |
776 |
d += prCoeff[i+j+48]*( subband_fir_hist[31-i+j-512]); |
777 |
} |
778 |
samples_out[i ] = a * scale + bias; |
779 |
samples_out[i+16] = b * scale + bias;
|
780 |
subband_fir_hist2[i ] = c; |
781 |
subband_fir_hist2[i+16] = d;
|
782 |
} |
783 |
samples_out+= 32;
|
784 |
|
785 |
hist_index = (hist_index-32)&511; |
786 |
} |
787 |
s->hist_index[chans]= hist_index; |
788 |
} |
789 |
|
790 |
static void lfe_interpolation_fir(int decimation_select, |
791 |
int num_deci_sample, float *samples_in, |
792 |
float *samples_out, float scale, |
793 |
float bias)
|
794 |
{ |
795 |
/* samples_in: An array holding decimated samples.
|
796 |
* Samples in current subframe starts from samples_in[0],
|
797 |
* while samples_in[-1], samples_in[-2], ..., stores samples
|
798 |
* from last subframe as history.
|
799 |
*
|
800 |
* samples_out: An array holding interpolated samples
|
801 |
*/
|
802 |
|
803 |
int decifactor, k, j;
|
804 |
const float *prCoeff; |
805 |
|
806 |
int interp_index = 0; /* Index to the interpolated samples */ |
807 |
int deciindex;
|
808 |
|
809 |
/* Select decimation filter */
|
810 |
if (decimation_select == 1) { |
811 |
decifactor = 128;
|
812 |
prCoeff = lfe_fir_128; |
813 |
} else {
|
814 |
decifactor = 64;
|
815 |
prCoeff = lfe_fir_64; |
816 |
} |
817 |
/* Interpolation */
|
818 |
for (deciindex = 0; deciindex < num_deci_sample; deciindex++) { |
819 |
/* One decimated sample generates decifactor interpolated ones */
|
820 |
for (k = 0; k < decifactor; k++) { |
821 |
float rTmp = 0.0; |
822 |
//FIXME the coeffs are symetric, fix that
|
823 |
for (j = 0; j < 512 / decifactor; j++) |
824 |
rTmp += samples_in[deciindex - j] * prCoeff[k + j * decifactor]; |
825 |
samples_out[interp_index++] = (rTmp * scale) + bias; |
826 |
} |
827 |
} |
828 |
} |
829 |
|
830 |
/* downmixing routines */
|
831 |
#define MIX_REAR1(samples, si1, rs, coef) \
|
832 |
samples[i] += samples[si1] * coef[rs][0]; \
|
833 |
samples[i+256] += samples[si1] * coef[rs][1]; |
834 |
|
835 |
#define MIX_REAR2(samples, si1, si2, rs, coef) \
|
836 |
samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \ |
837 |
samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1]; |
838 |
|
839 |
#define MIX_FRONT3(samples, coef) \
|
840 |
t = samples[i]; \ |
841 |
samples[i] = t * coef[0][0] + samples[i+256] * coef[1][0] + samples[i+512] * coef[2][0]; \ |
842 |
samples[i+256] = t * coef[0][1] + samples[i+256] * coef[1][1] + samples[i+512] * coef[2][1]; |
843 |
|
844 |
#define DOWNMIX_TO_STEREO(op1, op2) \
|
845 |
for(i = 0; i < 256; i++){ \ |
846 |
op1 \ |
847 |
op2 \ |
848 |
} |
849 |
|
850 |
static void dca_downmix(float *samples, int srcfmt, |
851 |
int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]) |
852 |
{ |
853 |
int i;
|
854 |
float t;
|
855 |
float coef[DCA_PRIM_CHANNELS_MAX][2]; |
856 |
|
857 |
for(i=0; i<DCA_PRIM_CHANNELS_MAX; i++) { |
858 |
coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]]; |
859 |
coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]]; |
860 |
} |
861 |
|
862 |
switch (srcfmt) {
|
863 |
case DCA_MONO:
|
864 |
case DCA_CHANNEL:
|
865 |
case DCA_STEREO_TOTAL:
|
866 |
case DCA_STEREO_SUMDIFF:
|
867 |
case DCA_4F2R:
|
868 |
av_log(NULL, 0, "Not implemented!\n"); |
869 |
break;
|
870 |
case DCA_STEREO:
|
871 |
break;
|
872 |
case DCA_3F:
|
873 |
DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),); |
874 |
break;
|
875 |
case DCA_2F1R:
|
876 |
DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + 512, 2, coef),); |
877 |
break;
|
878 |
case DCA_3F1R:
|
879 |
DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), |
880 |
MIX_REAR1(samples, i + 768, 3, coef)); |
881 |
break;
|
882 |
case DCA_2F2R:
|
883 |
DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + 512, i + 768, 2, coef),); |
884 |
break;
|
885 |
case DCA_3F2R:
|
886 |
DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), |
887 |
MIX_REAR2(samples, i + 768, i + 1024, 3, coef)); |
888 |
break;
|
889 |
} |
890 |
} |
891 |
|
892 |
|
893 |
/* Very compact version of the block code decoder that does not use table
|
894 |
* look-up but is slightly slower */
|
895 |
static int decode_blockcode(int code, int levels, int *values) |
896 |
{ |
897 |
int i;
|
898 |
int offset = (levels - 1) >> 1; |
899 |
|
900 |
for (i = 0; i < 4; i++) { |
901 |
values[i] = (code % levels) - offset; |
902 |
code /= levels; |
903 |
} |
904 |
|
905 |
if (code == 0) |
906 |
return 0; |
907 |
else {
|
908 |
av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n"); |
909 |
return -1; |
910 |
} |
911 |
} |
912 |
|
913 |
static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 }; |
914 |
static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 }; |
915 |
|
916 |
static int dca_subsubframe(DCAContext * s) |
917 |
{ |
918 |
int k, l;
|
919 |
int subsubframe = s->current_subsubframe;
|
920 |
|
921 |
const float *quant_step_table; |
922 |
|
923 |
/* FIXME */
|
924 |
float subband_samples[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8]; |
925 |
|
926 |
/*
|
927 |
* Audio data
|
928 |
*/
|
929 |
|
930 |
/* Select quantization step size table */
|
931 |
if (s->bit_rate_index == 0x1f) |
932 |
quant_step_table = lossless_quant_d; |
933 |
else
|
934 |
quant_step_table = lossy_quant_d; |
935 |
|
936 |
for (k = 0; k < s->prim_channels; k++) { |
937 |
for (l = 0; l < s->vq_start_subband[k]; l++) { |
938 |
int m;
|
939 |
|
940 |
/* Select the mid-tread linear quantizer */
|
941 |
int abits = s->bitalloc[k][l];
|
942 |
|
943 |
float quant_step_size = quant_step_table[abits];
|
944 |
float rscale;
|
945 |
|
946 |
/*
|
947 |
* Determine quantization index code book and its type
|
948 |
*/
|
949 |
|
950 |
/* Select quantization index code book */
|
951 |
int sel = s->quant_index_huffman[k][abits];
|
952 |
|
953 |
/*
|
954 |
* Extract bits from the bit stream
|
955 |
*/
|
956 |
if(!abits){
|
957 |
memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0])); |
958 |
}else if(abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){ |
959 |
if(abits <= 7){ |
960 |
/* Block code */
|
961 |
int block_code1, block_code2, size, levels;
|
962 |
int block[8]; |
963 |
|
964 |
size = abits_sizes[abits-1];
|
965 |
levels = abits_levels[abits-1];
|
966 |
|
967 |
block_code1 = get_bits(&s->gb, size); |
968 |
/* FIXME Should test return value */
|
969 |
decode_blockcode(block_code1, levels, block); |
970 |
block_code2 = get_bits(&s->gb, size); |
971 |
decode_blockcode(block_code2, levels, &block[4]);
|
972 |
for (m = 0; m < 8; m++) |
973 |
subband_samples[k][l][m] = block[m]; |
974 |
}else{
|
975 |
/* no coding */
|
976 |
for (m = 0; m < 8; m++) |
977 |
subband_samples[k][l][m] = get_sbits(&s->gb, abits - 3);
|
978 |
} |
979 |
}else{
|
980 |
/* Huffman coded */
|
981 |
for (m = 0; m < 8; m++) |
982 |
subband_samples[k][l][m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel); |
983 |
} |
984 |
|
985 |
/* Deal with transients */
|
986 |
if (s->transition_mode[k][l] &&
|
987 |
subsubframe >= s->transition_mode[k][l]) |
988 |
rscale = quant_step_size * s->scale_factor[k][l][1];
|
989 |
else
|
990 |
rscale = quant_step_size * s->scale_factor[k][l][0];
|
991 |
|
992 |
rscale *= s->scalefactor_adj[k][sel]; |
993 |
|
994 |
for (m = 0; m < 8; m++) |
995 |
subband_samples[k][l][m] *= rscale; |
996 |
|
997 |
/*
|
998 |
* Inverse ADPCM if in prediction mode
|
999 |
*/
|
1000 |
if (s->prediction_mode[k][l]) {
|
1001 |
int n;
|
1002 |
for (m = 0; m < 8; m++) { |
1003 |
for (n = 1; n <= 4; n++) |
1004 |
if (m >= n)
|
1005 |
subband_samples[k][l][m] += |
1006 |
(adpcm_vb[s->prediction_vq[k][l]][n - 1] *
|
1007 |
subband_samples[k][l][m - n] / 8192);
|
1008 |
else if (s->predictor_history) |
1009 |
subband_samples[k][l][m] += |
1010 |
(adpcm_vb[s->prediction_vq[k][l]][n - 1] *
|
1011 |
s->subband_samples_hist[k][l][m - n + |
1012 |
4] / 8192); |
1013 |
} |
1014 |
} |
1015 |
} |
1016 |
|
1017 |
/*
|
1018 |
* Decode VQ encoded high frequencies
|
1019 |
*/
|
1020 |
for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
|
1021 |
/* 1 vector -> 32 samples but we only need the 8 samples
|
1022 |
* for this subsubframe. */
|
1023 |
int m;
|
1024 |
|
1025 |
if (!s->debug_flag & 0x01) { |
1026 |
av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
|
1027 |
s->debug_flag |= 0x01;
|
1028 |
} |
1029 |
|
1030 |
for (m = 0; m < 8; m++) { |
1031 |
subband_samples[k][l][m] = |
1032 |
high_freq_vq[s->high_freq_vq[k][l]][subsubframe * 8 +
|
1033 |
m] |
1034 |
* (float) s->scale_factor[k][l][0] / 16.0; |
1035 |
} |
1036 |
} |
1037 |
} |
1038 |
|
1039 |
/* Check for DSYNC after subsubframe */
|
1040 |
if (s->aspf || subsubframe == s->subsubframes - 1) { |
1041 |
if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */ |
1042 |
#ifdef TRACE
|
1043 |
av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
|
1044 |
#endif
|
1045 |
} else {
|
1046 |
av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
|
1047 |
} |
1048 |
} |
1049 |
|
1050 |
/* Backup predictor history for adpcm */
|
1051 |
for (k = 0; k < s->prim_channels; k++) |
1052 |
for (l = 0; l < s->vq_start_subband[k]; l++) |
1053 |
memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
|
1054 |
4 * sizeof(subband_samples[0][0][0])); |
1055 |
|
1056 |
/* 32 subbands QMF */
|
1057 |
for (k = 0; k < s->prim_channels; k++) { |
1058 |
/* static float pcm_to_double[8] =
|
1059 |
{32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
|
1060 |
qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
|
1061 |
M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ ,
|
1062 |
s->add_bias ); |
1063 |
} |
1064 |
|
1065 |
/* Down mixing */
|
1066 |
|
1067 |
if (s->prim_channels > dca_channels[s->output & DCA_CHANNEL_MASK]) {
|
1068 |
dca_downmix(s->samples, s->amode, s->downmix_coef); |
1069 |
} |
1070 |
|
1071 |
/* Generate LFE samples for this subsubframe FIXME!!! */
|
1072 |
if (s->output & DCA_LFE) {
|
1073 |
int lfe_samples = 2 * s->lfe * s->subsubframes; |
1074 |
|
1075 |
lfe_interpolation_fir(s->lfe, 2 * s->lfe,
|
1076 |
s->lfe_data + lfe_samples + |
1077 |
2 * s->lfe * subsubframe,
|
1078 |
&s->samples[256 * dca_lfe_index[s->amode]],
|
1079 |
(1.0/256.0)*s->scale_bias, s->add_bias); |
1080 |
/* Outputs 20bits pcm samples */
|
1081 |
} |
1082 |
|
1083 |
return 0; |
1084 |
} |
1085 |
|
1086 |
|
1087 |
static int dca_subframe_footer(DCAContext * s) |
1088 |
{ |
1089 |
int aux_data_count = 0, i; |
1090 |
int lfe_samples;
|
1091 |
|
1092 |
/*
|
1093 |
* Unpack optional information
|
1094 |
*/
|
1095 |
|
1096 |
if (s->timestamp)
|
1097 |
get_bits(&s->gb, 32);
|
1098 |
|
1099 |
if (s->aux_data)
|
1100 |
aux_data_count = get_bits(&s->gb, 6);
|
1101 |
|
1102 |
for (i = 0; i < aux_data_count; i++) |
1103 |
get_bits(&s->gb, 8);
|
1104 |
|
1105 |
if (s->crc_present && (s->downmix || s->dynrange))
|
1106 |
get_bits(&s->gb, 16);
|
1107 |
|
1108 |
lfe_samples = 2 * s->lfe * s->subsubframes;
|
1109 |
for (i = 0; i < lfe_samples; i++) { |
1110 |
s->lfe_data[i] = s->lfe_data[i + lfe_samples]; |
1111 |
} |
1112 |
|
1113 |
return 0; |
1114 |
} |
1115 |
|
1116 |
/**
|
1117 |
* Decode a dca frame block
|
1118 |
*
|
1119 |
* @param s pointer to the DCAContext
|
1120 |
*/
|
1121 |
|
1122 |
static int dca_decode_block(DCAContext * s) |
1123 |
{ |
1124 |
|
1125 |
/* Sanity check */
|
1126 |
if (s->current_subframe >= s->subframes) {
|
1127 |
av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
|
1128 |
s->current_subframe, s->subframes); |
1129 |
return -1; |
1130 |
} |
1131 |
|
1132 |
if (!s->current_subsubframe) {
|
1133 |
#ifdef TRACE
|
1134 |
av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
|
1135 |
#endif
|
1136 |
/* Read subframe header */
|
1137 |
if (dca_subframe_header(s))
|
1138 |
return -1; |
1139 |
} |
1140 |
|
1141 |
/* Read subsubframe */
|
1142 |
#ifdef TRACE
|
1143 |
av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
|
1144 |
#endif
|
1145 |
if (dca_subsubframe(s))
|
1146 |
return -1; |
1147 |
|
1148 |
/* Update state */
|
1149 |
s->current_subsubframe++; |
1150 |
if (s->current_subsubframe >= s->subsubframes) {
|
1151 |
s->current_subsubframe = 0;
|
1152 |
s->current_subframe++; |
1153 |
} |
1154 |
if (s->current_subframe >= s->subframes) {
|
1155 |
#ifdef TRACE
|
1156 |
av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
|
1157 |
#endif
|
1158 |
/* Read subframe footer */
|
1159 |
if (dca_subframe_footer(s))
|
1160 |
return -1; |
1161 |
} |
1162 |
|
1163 |
return 0; |
1164 |
} |
1165 |
|
1166 |
/**
|
1167 |
* Convert bitstream to one representation based on sync marker
|
1168 |
*/
|
1169 |
static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst, |
1170 |
int max_size)
|
1171 |
{ |
1172 |
uint32_t mrk; |
1173 |
int i, tmp;
|
1174 |
const uint16_t *ssrc = (const uint16_t *) src; |
1175 |
uint16_t *sdst = (uint16_t *) dst; |
1176 |
PutBitContext pb; |
1177 |
|
1178 |
if((unsigned)src_size > (unsigned)max_size) { |
1179 |
// av_log(NULL, AV_LOG_ERROR, "Input frame size larger then DCA_MAX_FRAME_SIZE!\n");
|
1180 |
// return -1;
|
1181 |
src_size = max_size; |
1182 |
} |
1183 |
|
1184 |
mrk = AV_RB32(src); |
1185 |
switch (mrk) {
|
1186 |
case DCA_MARKER_RAW_BE:
|
1187 |
memcpy(dst, src, src_size); |
1188 |
return src_size;
|
1189 |
case DCA_MARKER_RAW_LE:
|
1190 |
for (i = 0; i < (src_size + 1) >> 1; i++) |
1191 |
*sdst++ = bswap_16(*ssrc++); |
1192 |
return src_size;
|
1193 |
case DCA_MARKER_14B_BE:
|
1194 |
case DCA_MARKER_14B_LE:
|
1195 |
init_put_bits(&pb, dst, max_size); |
1196 |
for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) { |
1197 |
tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
|
1198 |
put_bits(&pb, 14, tmp);
|
1199 |
} |
1200 |
flush_put_bits(&pb); |
1201 |
return (put_bits_count(&pb) + 7) >> 3; |
1202 |
default:
|
1203 |
return -1; |
1204 |
} |
1205 |
} |
1206 |
|
1207 |
/**
|
1208 |
* Main frame decoding function
|
1209 |
* FIXME add arguments
|
1210 |
*/
|
1211 |
static int dca_decode_frame(AVCodecContext * avctx, |
1212 |
void *data, int *data_size, |
1213 |
AVPacket *avpkt) |
1214 |
{ |
1215 |
const uint8_t *buf = avpkt->data;
|
1216 |
int buf_size = avpkt->size;
|
1217 |
|
1218 |
int i;
|
1219 |
int16_t *samples = data; |
1220 |
DCAContext *s = avctx->priv_data; |
1221 |
int channels;
|
1222 |
|
1223 |
|
1224 |
s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer, DCA_MAX_FRAME_SIZE); |
1225 |
if (s->dca_buffer_size == -1) { |
1226 |
av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
|
1227 |
return -1; |
1228 |
} |
1229 |
|
1230 |
init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
|
1231 |
if (dca_parse_frame_header(s) < 0) { |
1232 |
//seems like the frame is corrupt, try with the next one
|
1233 |
*data_size=0;
|
1234 |
return buf_size;
|
1235 |
} |
1236 |
//set AVCodec values with parsed data
|
1237 |
avctx->sample_rate = s->sample_rate; |
1238 |
avctx->bit_rate = s->bit_rate; |
1239 |
|
1240 |
channels = s->prim_channels + !!s->lfe; |
1241 |
|
1242 |
if (s->amode<16) { |
1243 |
avctx->channel_layout = dca_core_channel_layout[s->amode]; |
1244 |
|
1245 |
if (s->lfe) {
|
1246 |
avctx->channel_layout |= CH_LOW_FREQUENCY; |
1247 |
s->channel_order_tab = dca_channel_reorder_lfe[s->amode]; |
1248 |
} else
|
1249 |
s->channel_order_tab = dca_channel_reorder_nolfe[s->amode]; |
1250 |
|
1251 |
if(avctx->request_channels == 2 && s->prim_channels > 2) { |
1252 |
channels = 2;
|
1253 |
s->output = DCA_STEREO; |
1254 |
avctx->channel_layout = CH_LAYOUT_STEREO; |
1255 |
} |
1256 |
} else {
|
1257 |
av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
|
1258 |
return -1; |
1259 |
} |
1260 |
|
1261 |
|
1262 |
/* There is nothing that prevents a dts frame to change channel configuration
|
1263 |
but FFmpeg doesn't support that so only set the channels if it is previously
|
1264 |
unset. Ideally during the first probe for channels the crc should be checked
|
1265 |
and only set avctx->channels when the crc is ok. Right now the decoder could
|
1266 |
set the channels based on a broken first frame.*/
|
1267 |
if (!avctx->channels)
|
1268 |
avctx->channels = channels; |
1269 |
|
1270 |
if(*data_size < (s->sample_blocks / 8) * 256 * sizeof(int16_t) * channels) |
1271 |
return -1; |
1272 |
*data_size = 256 / 8 * s->sample_blocks * sizeof(int16_t) * channels; |
1273 |
for (i = 0; i < (s->sample_blocks / 8); i++) { |
1274 |
dca_decode_block(s); |
1275 |
s->dsp.float_to_int16_interleave(samples, s->samples_chanptr, 256, channels);
|
1276 |
samples += 256 * channels;
|
1277 |
} |
1278 |
|
1279 |
return buf_size;
|
1280 |
} |
1281 |
|
1282 |
|
1283 |
|
1284 |
/**
|
1285 |
* DCA initialization
|
1286 |
*
|
1287 |
* @param avctx pointer to the AVCodecContext
|
1288 |
*/
|
1289 |
|
1290 |
static av_cold int dca_decode_init(AVCodecContext * avctx) |
1291 |
{ |
1292 |
DCAContext *s = avctx->priv_data; |
1293 |
int i;
|
1294 |
|
1295 |
s->avctx = avctx; |
1296 |
dca_init_vlcs(); |
1297 |
|
1298 |
dsputil_init(&s->dsp, avctx); |
1299 |
ff_mdct_init(&s->imdct, 6, 1); |
1300 |
|
1301 |
for(i = 0; i < 6; i++) |
1302 |
s->samples_chanptr[i] = s->samples + i * 256;
|
1303 |
avctx->sample_fmt = SAMPLE_FMT_S16; |
1304 |
|
1305 |
if(s->dsp.float_to_int16 == ff_float_to_int16_c) {
|
1306 |
s->add_bias = 385.0f; |
1307 |
s->scale_bias = 1.0 / 32768.0; |
1308 |
} else {
|
1309 |
s->add_bias = 0.0f; |
1310 |
s->scale_bias = 1.0; |
1311 |
|
1312 |
/* allow downmixing to stereo */
|
1313 |
if (avctx->channels > 0 && avctx->request_channels < avctx->channels && |
1314 |
avctx->request_channels == 2) {
|
1315 |
avctx->channels = avctx->request_channels; |
1316 |
} |
1317 |
} |
1318 |
|
1319 |
|
1320 |
return 0; |
1321 |
} |
1322 |
|
1323 |
static av_cold int dca_decode_end(AVCodecContext * avctx) |
1324 |
{ |
1325 |
DCAContext *s = avctx->priv_data; |
1326 |
ff_mdct_end(&s->imdct); |
1327 |
return 0; |
1328 |
} |
1329 |
|
1330 |
AVCodec dca_decoder = { |
1331 |
.name = "dca",
|
1332 |
.type = CODEC_TYPE_AUDIO, |
1333 |
.id = CODEC_ID_DTS, |
1334 |
.priv_data_size = sizeof(DCAContext),
|
1335 |
.init = dca_decode_init, |
1336 |
.decode = dca_decode_frame, |
1337 |
.close = dca_decode_end, |
1338 |
.long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
|
1339 |
}; |