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
8 * This file is part of FFmpeg.
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.
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.
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
29 #include "libavutil/common.h"
30 #include "libavutil/intmath.h"
31 #include "libavutil/intreadwrite.h"
32 #include "libavcore/audioconvert.h"
41 #include "synth_filter.h"
46 #define DCA_PRIM_CHANNELS_MAX (7)
47 #define DCA_SUBBANDS (32)
48 #define DCA_ABITS_MAX (32) /* Should be 28 */
49 #define DCA_SUBSUBFRAMES_MAX (4)
50 #define DCA_SUBFRAMES_MAX (16)
51 #define DCA_BLOCKS_MAX (16)
52 #define DCA_LFE_MAX (3)
68 /* these are unconfirmed but should be mostly correct */
69 enum DCAExSSSpeakerMask {
70 DCA_EXSS_FRONT_CENTER = 0x0001,
71 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
72 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
73 DCA_EXSS_LFE = 0x0008,
74 DCA_EXSS_REAR_CENTER = 0x0010,
75 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
76 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
77 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
78 DCA_EXSS_OVERHEAD = 0x0100,
79 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
80 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
81 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
82 DCA_EXSS_LFE2 = 0x1000,
83 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
84 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
85 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
88 enum DCAExtensionMask {
89 DCA_EXT_CORE = 0x001, ///< core in core substream
90 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
91 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
92 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
93 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
94 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
95 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
96 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
97 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
98 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
101 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
102 * Some compromises have been made for special configurations. Most configurations
103 * are never used so complete accuracy is not needed.
105 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
106 * S -> side, when both rear and back are configured move one of them to the side channel
108 * All 2 channel configurations -> CH_LAYOUT_STEREO
111 static const int64_t dca_core_channel_layout[] = {
112 AV_CH_FRONT_CENTER, ///< 1, A
113 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
114 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
115 AV_CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference)
116 AV_CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total)
117 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER, ///< 3, C+L+R
118 AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER, ///< 3, L+R+S
119 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|CH_BACK_CENTER, ///< 4, C + L + R+ S
120 AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR
121 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR
122 AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
123 AV_CH_LAYOUT_STEREO|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 6, C + L + R+ LR + RR + OV
124 AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_BACK_CENTER|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT, ///< 6, CF+ CR+LF+ RF+LR + RR
125 AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
126 AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT|AV_CH_BACK_LEFT|AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2+ SR1 + SR2
127 AV_CH_FRONT_LEFT_OF_CENTER|AV_CH_FRONT_CENTER|AV_CH_FRONT_RIGHT_OF_CENTER|AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_BACK_CENTER|AV_CH_SIDE_RIGHT, ///< 8, CL + C+ CR + L + R + SL + S+ SR
130 static const int8_t dca_lfe_index[] = {
131 1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3
134 static const int8_t dca_channel_reorder_lfe[][9] = {
135 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
136 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
137 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
138 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
139 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
140 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
141 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
142 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
143 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
144 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
145 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
146 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
147 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
148 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
149 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
150 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
153 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
154 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
155 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
156 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
157 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
158 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
159 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
160 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
161 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
162 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
163 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
164 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
165 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
166 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
167 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
168 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
169 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
172 static const int8_t dca_channel_reorder_nolfe[][9] = {
173 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
174 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
175 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
176 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
177 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
178 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
179 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
180 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
181 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
182 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
183 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
184 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
185 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
186 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
187 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
188 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
191 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
192 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
193 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
194 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
195 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
196 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
197 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
198 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
199 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
200 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
201 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
202 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
203 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
204 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
205 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
206 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
207 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
210 #define DCA_DOLBY 101 /* FIXME */
212 #define DCA_CHANNEL_BITS 6
213 #define DCA_CHANNEL_MASK 0x3F
217 #define HEADER_SIZE 14
219 #define DCA_MAX_FRAME_SIZE 16384
220 #define DCA_MAX_EXSS_HEADER_SIZE 4096
222 #define DCA_BUFFER_PADDING_SIZE 1024
224 /** Bit allocation */
226 int offset; ///< code values offset
227 int maxbits[8]; ///< max bits in VLC
228 int wrap; ///< wrap for get_vlc2()
229 VLC vlc[8]; ///< actual codes
232 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
233 static BitAlloc dca_tmode; ///< transition mode VLCs
234 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
235 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
237 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx)
239 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
243 AVCodecContext *avctx;
245 int frame_type; ///< type of the current frame
246 int samples_deficit; ///< deficit sample count
247 int crc_present; ///< crc is present in the bitstream
248 int sample_blocks; ///< number of PCM sample blocks
249 int frame_size; ///< primary frame byte size
250 int amode; ///< audio channels arrangement
251 int sample_rate; ///< audio sampling rate
252 int bit_rate; ///< transmission bit rate
253 int bit_rate_index; ///< transmission bit rate index
255 int downmix; ///< embedded downmix enabled
256 int dynrange; ///< embedded dynamic range flag
257 int timestamp; ///< embedded time stamp flag
258 int aux_data; ///< auxiliary data flag
259 int hdcd; ///< source material is mastered in HDCD
260 int ext_descr; ///< extension audio descriptor flag
261 int ext_coding; ///< extended coding flag
262 int aspf; ///< audio sync word insertion flag
263 int lfe; ///< low frequency effects flag
264 int predictor_history; ///< predictor history flag
265 int header_crc; ///< header crc check bytes
266 int multirate_inter; ///< multirate interpolator switch
267 int version; ///< encoder software revision
268 int copy_history; ///< copy history
269 int source_pcm_res; ///< source pcm resolution
270 int front_sum; ///< front sum/difference flag
271 int surround_sum; ///< surround sum/difference flag
272 int dialog_norm; ///< dialog normalisation parameter
274 /* Primary audio coding header */
275 int subframes; ///< number of subframes
276 int is_channels_set; ///< check for if the channel number is already set
277 int total_channels; ///< number of channels including extensions
278 int prim_channels; ///< number of primary audio channels
279 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
280 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
281 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
282 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
283 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
284 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
285 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
286 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
288 /* Primary audio coding side information */
289 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
290 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
291 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
292 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
293 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
294 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
295 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
296 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
297 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
298 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
299 int dynrange_coef; ///< dynamic range coefficient
301 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
303 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
304 int lfe_scale_factor;
306 /* Subband samples history (for ADPCM) */
307 float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
308 DECLARE_ALIGNED(16, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
309 DECLARE_ALIGNED(16, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
310 int hist_index[DCA_PRIM_CHANNELS_MAX];
311 DECLARE_ALIGNED(16, float, raXin)[32];
313 int output; ///< type of output
314 float add_bias; ///< output bias
315 float scale_bias; ///< output scale
317 DECLARE_ALIGNED(16, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
318 DECLARE_ALIGNED(16, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
319 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
321 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
322 int dca_buffer_size; ///< how much data is in the dca_buffer
324 const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe
326 /* Current position in DCA frame */
327 int current_subframe;
328 int current_subsubframe;
330 /* XCh extension information */
332 int xch_base_channel; ///< index of first (only) channel containing XCH data
334 /* Other detected extensions in the core substream */
338 /* ExSS header parser */
339 int static_fields; ///< static fields present
340 int mix_metadata; ///< mixing metadata present
341 int num_mix_configs; ///< number of mix out configurations
342 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
346 int debug_flag; ///< used for suppressing repeated error messages output
349 SynthFilterContext synth;
350 DCADSPContext dcadsp;
353 static const uint16_t dca_vlc_offs[] = {
354 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
355 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
356 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
357 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
358 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
359 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
362 static av_cold void dca_init_vlcs(void)
364 static int vlcs_initialized = 0;
366 static VLC_TYPE dca_table[23622][2];
368 if (vlcs_initialized)
371 dca_bitalloc_index.offset = 1;
372 dca_bitalloc_index.wrap = 2;
373 for (i = 0; i < 5; i++) {
374 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
375 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
376 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
377 bitalloc_12_bits[i], 1, 1,
378 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
380 dca_scalefactor.offset = -64;
381 dca_scalefactor.wrap = 2;
382 for (i = 0; i < 5; i++) {
383 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
384 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
385 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
386 scales_bits[i], 1, 1,
387 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
389 dca_tmode.offset = 0;
391 for (i = 0; i < 4; i++) {
392 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
393 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
394 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
396 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
399 for (i = 0; i < 10; i++)
400 for (j = 0; j < 7; j++){
401 if (!bitalloc_codes[i][j]) break;
402 dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
403 dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4);
404 dca_smpl_bitalloc[i+1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
405 dca_smpl_bitalloc[i+1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
406 init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
408 bitalloc_bits[i][j], 1, 1,
409 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
412 vlcs_initialized = 1;
415 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
418 *dst++ = get_bits(gb, bits);
421 static int dca_parse_audio_coding_header(DCAContext * s, int base_channel)
424 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
425 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
426 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
428 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
429 s->prim_channels = s->total_channels;
431 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
432 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
435 for (i = base_channel; i < s->prim_channels; i++) {
436 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
437 if (s->subband_activity[i] > DCA_SUBBANDS)
438 s->subband_activity[i] = DCA_SUBBANDS;
440 for (i = base_channel; i < s->prim_channels; i++) {
441 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
442 if (s->vq_start_subband[i] > DCA_SUBBANDS)
443 s->vq_start_subband[i] = DCA_SUBBANDS;
445 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
446 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
447 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
448 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
450 /* Get codebooks quantization indexes */
452 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
453 for (j = 1; j < 11; j++)
454 for (i = base_channel; i < s->prim_channels; i++)
455 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
457 /* Get scale factor adjustment */
458 for (j = 0; j < 11; j++)
459 for (i = base_channel; i < s->prim_channels; i++)
460 s->scalefactor_adj[i][j] = 1;
462 for (j = 1; j < 11; j++)
463 for (i = base_channel; i < s->prim_channels; i++)
464 if (s->quant_index_huffman[i][j] < thr[j])
465 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
467 if (s->crc_present) {
468 /* Audio header CRC check */
469 get_bits(&s->gb, 16);
472 s->current_subframe = 0;
473 s->current_subsubframe = 0;
476 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
477 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
478 for (i = base_channel; i < s->prim_channels; i++){
479 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
480 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
481 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
482 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
483 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
484 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
485 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
486 for (j = 0; j < 11; j++)
487 av_log(s->avctx, AV_LOG_DEBUG, " %i",
488 s->quant_index_huffman[i][j]);
489 av_log(s->avctx, AV_LOG_DEBUG, "\n");
490 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
491 for (j = 0; j < 11; j++)
492 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
493 av_log(s->avctx, AV_LOG_DEBUG, "\n");
500 static int dca_parse_frame_header(DCAContext * s)
502 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
505 get_bits(&s->gb, 32);
508 s->frame_type = get_bits(&s->gb, 1);
509 s->samples_deficit = get_bits(&s->gb, 5) + 1;
510 s->crc_present = get_bits(&s->gb, 1);
511 s->sample_blocks = get_bits(&s->gb, 7) + 1;
512 s->frame_size = get_bits(&s->gb, 14) + 1;
513 if (s->frame_size < 95)
515 s->amode = get_bits(&s->gb, 6);
516 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
519 s->bit_rate_index = get_bits(&s->gb, 5);
520 s->bit_rate = dca_bit_rates[s->bit_rate_index];
524 s->downmix = get_bits(&s->gb, 1);
525 s->dynrange = get_bits(&s->gb, 1);
526 s->timestamp = get_bits(&s->gb, 1);
527 s->aux_data = get_bits(&s->gb, 1);
528 s->hdcd = get_bits(&s->gb, 1);
529 s->ext_descr = get_bits(&s->gb, 3);
530 s->ext_coding = get_bits(&s->gb, 1);
531 s->aspf = get_bits(&s->gb, 1);
532 s->lfe = get_bits(&s->gb, 2);
533 s->predictor_history = get_bits(&s->gb, 1);
535 /* TODO: check CRC */
537 s->header_crc = get_bits(&s->gb, 16);
539 s->multirate_inter = get_bits(&s->gb, 1);
540 s->version = get_bits(&s->gb, 4);
541 s->copy_history = get_bits(&s->gb, 2);
542 s->source_pcm_res = get_bits(&s->gb, 3);
543 s->front_sum = get_bits(&s->gb, 1);
544 s->surround_sum = get_bits(&s->gb, 1);
545 s->dialog_norm = get_bits(&s->gb, 4);
547 /* FIXME: channels mixing levels */
548 s->output = s->amode;
549 if (s->lfe) s->output |= DCA_LFE;
552 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
553 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
554 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
555 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
556 s->sample_blocks, s->sample_blocks * 32);
557 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
558 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
559 s->amode, dca_channels[s->amode]);
560 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
562 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
564 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
565 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
566 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
567 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
568 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
569 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
570 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
571 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
572 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
573 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
574 s->predictor_history);
575 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
576 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
578 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
579 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
580 av_log(s->avctx, AV_LOG_DEBUG,
581 "source pcm resolution: %i (%i bits/sample)\n",
582 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
583 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
584 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
585 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
586 av_log(s->avctx, AV_LOG_DEBUG, "\n");
589 /* Primary audio coding header */
590 s->subframes = get_bits(&s->gb, 4) + 1;
592 return dca_parse_audio_coding_header(s, 0);
596 static inline int get_scale(GetBitContext *gb, int level, int value)
599 /* huffman encoded */
600 value += get_bitalloc(gb, &dca_scalefactor, level);
601 } else if (level < 8)
602 value = get_bits(gb, level + 1);
606 static int dca_subframe_header(DCAContext * s, int base_channel, int block_index)
608 /* Primary audio coding side information */
611 if (get_bits_left(&s->gb) < 0)
615 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
616 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
619 for (j = base_channel; j < s->prim_channels; j++) {
620 for (k = 0; k < s->subband_activity[j]; k++)
621 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
624 /* Get prediction codebook */
625 for (j = base_channel; j < s->prim_channels; j++) {
626 for (k = 0; k < s->subband_activity[j]; k++) {
627 if (s->prediction_mode[j][k] > 0) {
628 /* (Prediction coefficient VQ address) */
629 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
634 /* Bit allocation index */
635 for (j = base_channel; j < s->prim_channels; j++) {
636 for (k = 0; k < s->vq_start_subband[j]; k++) {
637 if (s->bitalloc_huffman[j] == 6)
638 s->bitalloc[j][k] = get_bits(&s->gb, 5);
639 else if (s->bitalloc_huffman[j] == 5)
640 s->bitalloc[j][k] = get_bits(&s->gb, 4);
641 else if (s->bitalloc_huffman[j] == 7) {
642 av_log(s->avctx, AV_LOG_ERROR,
643 "Invalid bit allocation index\n");
647 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
650 if (s->bitalloc[j][k] > 26) {
651 // av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
652 // j, k, s->bitalloc[j][k]);
658 /* Transition mode */
659 for (j = base_channel; j < s->prim_channels; j++) {
660 for (k = 0; k < s->subband_activity[j]; k++) {
661 s->transition_mode[j][k] = 0;
662 if (s->subsubframes[s->current_subframe] > 1 &&
663 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
664 s->transition_mode[j][k] =
665 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
670 if (get_bits_left(&s->gb) < 0)
673 for (j = base_channel; j < s->prim_channels; j++) {
674 const uint32_t *scale_table;
677 memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
679 if (s->scalefactor_huffman[j] == 6)
680 scale_table = scale_factor_quant7;
682 scale_table = scale_factor_quant6;
684 /* When huffman coded, only the difference is encoded */
687 for (k = 0; k < s->subband_activity[j]; k++) {
688 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
689 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
690 s->scale_factor[j][k][0] = scale_table[scale_sum];
693 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
694 /* Get second scale factor */
695 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
696 s->scale_factor[j][k][1] = scale_table[scale_sum];
701 /* Joint subband scale factor codebook select */
702 for (j = base_channel; j < s->prim_channels; j++) {
703 /* Transmitted only if joint subband coding enabled */
704 if (s->joint_intensity[j] > 0)
705 s->joint_huff[j] = get_bits(&s->gb, 3);
708 if (get_bits_left(&s->gb) < 0)
711 /* Scale factors for joint subband coding */
712 for (j = base_channel; j < s->prim_channels; j++) {
715 /* Transmitted only if joint subband coding enabled */
716 if (s->joint_intensity[j] > 0) {
718 source_channel = s->joint_intensity[j] - 1;
720 /* When huffman coded, only the difference is encoded
721 * (is this valid as well for joint scales ???) */
723 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
724 scale = get_scale(&s->gb, s->joint_huff[j], 0);
725 scale += 64; /* bias */
726 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
729 if (!(s->debug_flag & 0x02)) {
730 av_log(s->avctx, AV_LOG_DEBUG,
731 "Joint stereo coding not supported\n");
732 s->debug_flag |= 0x02;
737 /* Stereo downmix coefficients */
738 if (!base_channel && s->prim_channels > 2) {
740 for (j = base_channel; j < s->prim_channels; j++) {
741 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
742 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
745 int am = s->amode & DCA_CHANNEL_MASK;
746 for (j = base_channel; j < s->prim_channels; j++) {
747 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
748 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
753 /* Dynamic range coefficient */
754 if (!base_channel && s->dynrange)
755 s->dynrange_coef = get_bits(&s->gb, 8);
757 /* Side information CRC check word */
758 if (s->crc_present) {
759 get_bits(&s->gb, 16);
763 * Primary audio data arrays
766 /* VQ encoded high frequency subbands */
767 for (j = base_channel; j < s->prim_channels; j++)
768 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
769 /* 1 vector -> 32 samples */
770 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
772 /* Low frequency effect data */
773 if (!base_channel && s->lfe) {
775 int lfe_samples = 2 * s->lfe * (4 + block_index);
776 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
779 for (j = lfe_samples; j < lfe_end_sample; j++) {
780 /* Signed 8 bits int */
781 s->lfe_data[j] = get_sbits(&s->gb, 8);
784 /* Scale factor index */
785 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
787 /* Quantization step size * scale factor */
788 lfe_scale = 0.035 * s->lfe_scale_factor;
790 for (j = lfe_samples; j < lfe_end_sample; j++)
791 s->lfe_data[j] *= lfe_scale;
795 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes[s->current_subframe]);
796 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
797 s->partial_samples[s->current_subframe]);
798 for (j = base_channel; j < s->prim_channels; j++) {
799 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
800 for (k = 0; k < s->subband_activity[j]; k++)
801 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
802 av_log(s->avctx, AV_LOG_DEBUG, "\n");
804 for (j = base_channel; j < s->prim_channels; j++) {
805 for (k = 0; k < s->subband_activity[j]; k++)
806 av_log(s->avctx, AV_LOG_DEBUG,
807 "prediction coefs: %f, %f, %f, %f\n",
808 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
809 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
810 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
811 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
813 for (j = base_channel; j < s->prim_channels; j++) {
814 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
815 for (k = 0; k < s->vq_start_subband[j]; k++)
816 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
817 av_log(s->avctx, AV_LOG_DEBUG, "\n");
819 for (j = base_channel; j < s->prim_channels; j++) {
820 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
821 for (k = 0; k < s->subband_activity[j]; k++)
822 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
823 av_log(s->avctx, AV_LOG_DEBUG, "\n");
825 for (j = base_channel; j < s->prim_channels; j++) {
826 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
827 for (k = 0; k < s->subband_activity[j]; k++) {
828 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
829 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
830 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
831 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
833 av_log(s->avctx, AV_LOG_DEBUG, "\n");
835 for (j = base_channel; j < s->prim_channels; j++) {
836 if (s->joint_intensity[j] > 0) {
837 int source_channel = s->joint_intensity[j] - 1;
838 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
839 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
840 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
841 av_log(s->avctx, AV_LOG_DEBUG, "\n");
844 if (!base_channel && s->prim_channels > 2 && s->downmix) {
845 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
846 for (j = 0; j < s->prim_channels; j++) {
847 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]);
848 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]);
850 av_log(s->avctx, AV_LOG_DEBUG, "\n");
852 for (j = base_channel; j < s->prim_channels; j++)
853 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
854 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
855 if (!base_channel && s->lfe) {
856 int lfe_samples = 2 * s->lfe * (4 + block_index);
857 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
859 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
860 for (j = lfe_samples; j < lfe_end_sample; j++)
861 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
862 av_log(s->avctx, AV_LOG_DEBUG, "\n");
869 static void qmf_32_subbands(DCAContext * s, int chans,
870 float samples_in[32][8], float *samples_out,
871 float scale, float bias)
873 const float *prCoeff;
876 int sb_act = s->subband_activity[chans];
879 scale *= sqrt(1/8.0);
882 if (!s->multirate_inter) /* Non-perfect reconstruction */
883 prCoeff = fir_32bands_nonperfect;
884 else /* Perfect reconstruction */
885 prCoeff = fir_32bands_perfect;
887 /* Reconstructed channel sample index */
888 for (subindex = 0; subindex < 8; subindex++) {
889 /* Load in one sample from each subband and clear inactive subbands */
890 for (i = 0; i < sb_act; i++){
891 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ ((i-1)&2)<<30;
892 AV_WN32A(&s->raXin[i], v);
897 s->synth.synth_filter_float(&s->imdct,
898 s->subband_fir_hist[chans], &s->hist_index[chans],
899 s->subband_fir_noidea[chans], prCoeff,
900 samples_out, s->raXin, scale, bias);
906 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
907 int num_deci_sample, float *samples_in,
908 float *samples_out, float scale,
911 /* samples_in: An array holding decimated samples.
912 * Samples in current subframe starts from samples_in[0],
913 * while samples_in[-1], samples_in[-2], ..., stores samples
914 * from last subframe as history.
916 * samples_out: An array holding interpolated samples
920 const float *prCoeff;
923 /* Select decimation filter */
924 if (decimation_select == 1) {
926 prCoeff = lfe_fir_128;
929 prCoeff = lfe_fir_64;
932 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
933 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
936 samples_out += 2 * decifactor;
940 /* downmixing routines */
941 #define MIX_REAR1(samples, si1, rs, coef) \
942 samples[i] += (samples[si1] - add_bias) * coef[rs][0]; \
943 samples[i+256] += (samples[si1] - add_bias) * coef[rs][1];
945 #define MIX_REAR2(samples, si1, si2, rs, coef) \
946 samples[i] += (samples[si1] - add_bias) * coef[rs][0] + (samples[si2] - add_bias) * coef[rs+1][0]; \
947 samples[i+256] += (samples[si1] - add_bias) * coef[rs][1] + (samples[si2] - add_bias) * coef[rs+1][1];
949 #define MIX_FRONT3(samples, coef) \
950 t = samples[i+c] - add_bias; \
951 u = samples[i+l] - add_bias; \
952 v = samples[i+r] - add_bias; \
953 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0] + add_bias; \
954 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1] + add_bias;
956 #define DOWNMIX_TO_STEREO(op1, op2) \
957 for (i = 0; i < 256; i++){ \
962 static void dca_downmix(float *samples, int srcfmt,
963 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
964 const int8_t *channel_mapping, float add_bias)
969 float coef[DCA_PRIM_CHANNELS_MAX][2];
971 for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) {
972 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
973 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
979 case DCA_STEREO_TOTAL:
980 case DCA_STEREO_SUMDIFF:
982 av_log(NULL, 0, "Not implemented!\n");
987 c = channel_mapping[0] * 256;
988 l = channel_mapping[1] * 256;
989 r = channel_mapping[2] * 256;
990 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),);
993 s = channel_mapping[2] * 256;
994 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
997 c = channel_mapping[0] * 256;
998 l = channel_mapping[1] * 256;
999 r = channel_mapping[2] * 256;
1000 s = channel_mapping[3] * 256;
1001 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1002 MIX_REAR1(samples, i + s, 3, coef));
1005 sl = channel_mapping[2] * 256;
1006 sr = channel_mapping[3] * 256;
1007 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
1010 c = channel_mapping[0] * 256;
1011 l = channel_mapping[1] * 256;
1012 r = channel_mapping[2] * 256;
1013 sl = channel_mapping[3] * 256;
1014 sr = channel_mapping[4] * 256;
1015 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1016 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1022 /* Very compact version of the block code decoder that does not use table
1023 * look-up but is slightly slower */
1024 static int decode_blockcode(int code, int levels, int *values)
1027 int offset = (levels - 1) >> 1;
1029 for (i = 0; i < 4; i++) {
1030 int div = FASTDIV(code, levels);
1031 values[i] = code - offset - div*levels;
1038 av_log(NULL, AV_LOG_ERROR, "ERROR: block code look-up failed\n");
1043 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1044 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1046 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1049 int subsubframe = s->current_subsubframe;
1051 const float *quant_step_table;
1054 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1055 LOCAL_ALIGNED_16(int, block, [8]);
1061 /* Select quantization step size table */
1062 if (s->bit_rate_index == 0x1f)
1063 quant_step_table = lossless_quant_d;
1065 quant_step_table = lossy_quant_d;
1067 for (k = base_channel; k < s->prim_channels; k++) {
1068 if (get_bits_left(&s->gb) < 0)
1071 for (l = 0; l < s->vq_start_subband[k]; l++) {
1074 /* Select the mid-tread linear quantizer */
1075 int abits = s->bitalloc[k][l];
1077 float quant_step_size = quant_step_table[abits];
1080 * Determine quantization index code book and its type
1083 /* Select quantization index code book */
1084 int sel = s->quant_index_huffman[k][abits];
1087 * Extract bits from the bit stream
1090 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1092 /* Deal with transients */
1093 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1094 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1096 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1099 int block_code1, block_code2, size, levels;
1101 size = abits_sizes[abits-1];
1102 levels = abits_levels[abits-1];
1104 block_code1 = get_bits(&s->gb, size);
1105 /* FIXME Should test return value */
1106 decode_blockcode(block_code1, levels, block);
1107 block_code2 = get_bits(&s->gb, size);
1108 decode_blockcode(block_code2, levels, &block[4]);
1111 for (m = 0; m < 8; m++)
1112 block[m] = get_sbits(&s->gb, abits - 3);
1116 for (m = 0; m < 8; m++)
1117 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1120 s->dsp.int32_to_float_fmul_scalar(subband_samples[k][l],
1125 * Inverse ADPCM if in prediction mode
1127 if (s->prediction_mode[k][l]) {
1129 for (m = 0; m < 8; m++) {
1130 for (n = 1; n <= 4; n++)
1132 subband_samples[k][l][m] +=
1133 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1134 subband_samples[k][l][m - n] / 8192);
1135 else if (s->predictor_history)
1136 subband_samples[k][l][m] +=
1137 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1138 s->subband_samples_hist[k][l][m - n +
1145 * Decode VQ encoded high frequencies
1147 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1148 /* 1 vector -> 32 samples but we only need the 8 samples
1149 * for this subsubframe. */
1152 if (!s->debug_flag & 0x01) {
1153 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1154 s->debug_flag |= 0x01;
1157 for (m = 0; m < 8; m++) {
1158 subband_samples[k][l][m] =
1159 high_freq_vq[s->high_freq_vq[k][l]][subsubframe * 8 +
1161 * (float) s->scale_factor[k][l][0] / 16.0;
1166 /* Check for DSYNC after subsubframe */
1167 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1168 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1170 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1173 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1177 /* Backup predictor history for adpcm */
1178 for (k = base_channel; k < s->prim_channels; k++)
1179 for (l = 0; l < s->vq_start_subband[k]; l++)
1180 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1181 4 * sizeof(subband_samples[0][0][0]));
1186 static int dca_filter_channels(DCAContext * s, int block_index)
1188 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1191 /* 32 subbands QMF */
1192 for (k = 0; k < s->prim_channels; k++) {
1193 /* static float pcm_to_double[8] =
1194 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1195 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1196 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ ,
1201 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1202 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab, s->add_bias);
1205 /* Generate LFE samples for this subsubframe FIXME!!! */
1206 if (s->output & DCA_LFE) {
1207 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1208 s->lfe_data + 2 * s->lfe * (block_index + 4),
1209 &s->samples[256 * dca_lfe_index[s->amode]],
1210 (1.0/256.0)*s->scale_bias, s->add_bias);
1211 /* Outputs 20bits pcm samples */
1218 static int dca_subframe_footer(DCAContext * s, int base_channel)
1220 int aux_data_count = 0, i;
1223 * Unpack optional information
1226 /* presumably optional information only appears in the core? */
1227 if (!base_channel) {
1229 get_bits(&s->gb, 32);
1232 aux_data_count = get_bits(&s->gb, 6);
1234 for (i = 0; i < aux_data_count; i++)
1235 get_bits(&s->gb, 8);
1237 if (s->crc_present && (s->downmix || s->dynrange))
1238 get_bits(&s->gb, 16);
1245 * Decode a dca frame block
1247 * @param s pointer to the DCAContext
1250 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1254 if (s->current_subframe >= s->subframes) {
1255 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1256 s->current_subframe, s->subframes);
1260 if (!s->current_subsubframe) {
1262 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1264 /* Read subframe header */
1265 if (dca_subframe_header(s, base_channel, block_index))
1269 /* Read subsubframe */
1271 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1273 if (dca_subsubframe(s, base_channel, block_index))
1277 s->current_subsubframe++;
1278 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1279 s->current_subsubframe = 0;
1280 s->current_subframe++;
1282 if (s->current_subframe >= s->subframes) {
1284 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1286 /* Read subframe footer */
1287 if (dca_subframe_footer(s, base_channel))
1295 * Convert bitstream to one representation based on sync marker
1297 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1302 const uint16_t *ssrc = (const uint16_t *) src;
1303 uint16_t *sdst = (uint16_t *) dst;
1306 if ((unsigned)src_size > (unsigned)max_size) {
1307 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger then DCA_MAX_FRAME_SIZE!\n");
1309 src_size = max_size;
1314 case DCA_MARKER_RAW_BE:
1315 memcpy(dst, src, src_size);
1317 case DCA_MARKER_RAW_LE:
1318 for (i = 0; i < (src_size + 1) >> 1; i++)
1319 *sdst++ = av_bswap16(*ssrc++);
1321 case DCA_MARKER_14B_BE:
1322 case DCA_MARKER_14B_LE:
1323 init_put_bits(&pb, dst, max_size);
1324 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1325 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1326 put_bits(&pb, 14, tmp);
1328 flush_put_bits(&pb);
1329 return (put_bits_count(&pb) + 7) >> 3;
1336 * Return the number of channels in an ExSS speaker mask (HD)
1338 static int dca_exss_mask2count(int mask)
1340 /* count bits that mean speaker pairs twice */
1341 return av_popcount(mask)
1342 + av_popcount(mask & (
1343 DCA_EXSS_CENTER_LEFT_RIGHT
1344 | DCA_EXSS_FRONT_LEFT_RIGHT
1345 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1346 | DCA_EXSS_WIDE_LEFT_RIGHT
1347 | DCA_EXSS_SIDE_LEFT_RIGHT
1348 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1349 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1350 | DCA_EXSS_REAR_LEFT_RIGHT
1351 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1356 * Skip mixing coefficients of a single mix out configuration (HD)
1358 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1360 for (int i = 0; i < channels; i++) {
1361 int mix_map_mask = get_bits(gb, out_ch);
1362 int num_coeffs = av_popcount(mix_map_mask);
1363 skip_bits_long(gb, num_coeffs * 6);
1368 * Parse extension substream asset header (HD)
1370 static int dca_exss_parse_asset_header(DCAContext *s)
1372 int header_pos = get_bits_count(&s->gb);
1375 int embedded_stereo = 0;
1376 int embedded_6ch = 0;
1377 int drc_code_present;
1378 int extensions_mask;
1381 if (get_bits_left(&s->gb) < 16)
1384 /* We will parse just enough to get to the extensions bitmask with which
1385 * we can set the profile value. */
1387 header_size = get_bits(&s->gb, 9) + 1;
1388 skip_bits(&s->gb, 3); // asset index
1390 if (s->static_fields) {
1391 if (get_bits1(&s->gb))
1392 skip_bits(&s->gb, 4); // asset type descriptor
1393 if (get_bits1(&s->gb))
1394 skip_bits_long(&s->gb, 24); // language descriptor
1396 if (get_bits1(&s->gb)) {
1397 /* How can one fit 1024 bytes of text here if the maximum value
1398 * for the asset header size field above was 512 bytes? */
1399 int text_length = get_bits(&s->gb, 10) + 1;
1400 if (get_bits_left(&s->gb) < text_length * 8)
1402 skip_bits_long(&s->gb, text_length * 8); // info text
1405 skip_bits(&s->gb, 5); // bit resolution - 1
1406 skip_bits(&s->gb, 4); // max sample rate code
1407 channels = get_bits(&s->gb, 8) + 1;
1409 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1410 int spkr_remap_sets;
1411 int spkr_mask_size = 16;
1415 embedded_stereo = get_bits1(&s->gb);
1417 embedded_6ch = get_bits1(&s->gb);
1419 if (get_bits1(&s->gb)) {
1420 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1421 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1424 spkr_remap_sets = get_bits(&s->gb, 3);
1426 for (i = 0; i < spkr_remap_sets; i++) {
1427 /* std layout mask for each remap set */
1428 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1431 for (i = 0; i < spkr_remap_sets; i++) {
1432 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1433 if (get_bits_left(&s->gb) < 0)
1436 for (j = 0; j < num_spkrs[i]; j++) {
1437 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1438 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1439 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1444 skip_bits(&s->gb, 3); // representation type
1448 drc_code_present = get_bits1(&s->gb);
1449 if (drc_code_present)
1450 get_bits(&s->gb, 8); // drc code
1452 if (get_bits1(&s->gb))
1453 skip_bits(&s->gb, 5); // dialog normalization code
1455 if (drc_code_present && embedded_stereo)
1456 get_bits(&s->gb, 8); // drc stereo code
1458 if (s->mix_metadata && get_bits1(&s->gb)) {
1459 skip_bits(&s->gb, 1); // external mix
1460 skip_bits(&s->gb, 6); // post mix gain code
1462 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1463 skip_bits(&s->gb, 3); // drc limit
1465 skip_bits(&s->gb, 8); // custom drc code
1467 if (get_bits1(&s->gb)) // channel specific scaling
1468 for (i = 0; i < s->num_mix_configs; i++)
1469 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1471 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1473 for (i = 0; i < s->num_mix_configs; i++) {
1474 if (get_bits_left(&s->gb) < 0)
1476 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1478 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1479 if (embedded_stereo)
1480 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1484 switch (get_bits(&s->gb, 2)) {
1485 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1486 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1487 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1488 case 3: extensions_mask = 0; /* aux coding */ break;
1491 /* not parsed further, we were only interested in the extensions mask */
1493 if (get_bits_left(&s->gb) < 0)
1496 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1497 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1500 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1502 if (extensions_mask & DCA_EXT_EXSS_XLL)
1503 s->profile = FF_PROFILE_DTS_HD_MA;
1504 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1506 s->profile = FF_PROFILE_DTS_HD_HRA;
1508 if (!(extensions_mask & DCA_EXT_CORE))
1509 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1510 if (!!(extensions_mask & DCA_EXT_XCH) != s->xch_present)
1511 av_log(s->avctx, AV_LOG_WARNING, "DTS XCh detection mismatch.\n");
1512 if (!!(extensions_mask & DCA_EXT_XXCH) != s->xxch_present)
1513 av_log(s->avctx, AV_LOG_WARNING, "DTS XXCh detection mismatch.\n");
1514 if (!!(extensions_mask & DCA_EXT_X96) != s->x96_present)
1515 av_log(s->avctx, AV_LOG_WARNING, "DTS X96 detection mismatch.\n");
1521 * Parse extension substream header (HD)
1523 static void dca_exss_parse_header(DCAContext *s)
1531 int active_ss_mask[8];
1534 if (get_bits_left(&s->gb) < 52)
1537 skip_bits(&s->gb, 8); // user data
1538 ss_index = get_bits(&s->gb, 2);
1540 blownup = get_bits1(&s->gb);
1541 header_size = get_bits(&s->gb, 8 + 4 * blownup) + 1;
1542 hd_size = get_bits_long(&s->gb, 16 + 4 * blownup) + 1;
1544 s->static_fields = get_bits1(&s->gb);
1545 if (s->static_fields) {
1546 skip_bits(&s->gb, 2); // reference clock code
1547 skip_bits(&s->gb, 3); // frame duration code
1549 if (get_bits1(&s->gb))
1550 skip_bits_long(&s->gb, 36); // timestamp
1552 /* a single stream can contain multiple audio assets that can be
1553 * combined to form multiple audio presentations */
1555 num_audiop = get_bits(&s->gb, 3) + 1;
1556 if (num_audiop > 1) {
1557 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1558 /* ignore such streams for now */
1562 num_assets = get_bits(&s->gb, 3) + 1;
1563 if (num_assets > 1) {
1564 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1565 /* ignore such streams for now */
1569 for (i = 0; i < num_audiop; i++)
1570 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1572 for (i = 0; i < num_audiop; i++)
1573 for (j = 0; j <= ss_index; j++)
1574 if (active_ss_mask[i] & (1 << j))
1575 skip_bits(&s->gb, 8); // active asset mask
1577 s->mix_metadata = get_bits1(&s->gb);
1578 if (s->mix_metadata) {
1579 int mix_out_mask_size;
1581 skip_bits(&s->gb, 2); // adjustment level
1582 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1583 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1585 for (i = 0; i < s->num_mix_configs; i++) {
1586 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1587 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1592 for (i = 0; i < num_assets; i++)
1593 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1595 for (i = 0; i < num_assets; i++) {
1596 if (dca_exss_parse_asset_header(s))
1600 /* not parsed further, we were only interested in the extensions mask
1601 * from the asset header */
1605 * Main frame decoding function
1606 * FIXME add arguments
1608 static int dca_decode_frame(AVCodecContext * avctx,
1609 void *data, int *data_size,
1612 const uint8_t *buf = avpkt->data;
1613 int buf_size = avpkt->size;
1616 int num_core_channels = 0;
1618 int16_t *samples = data;
1619 DCAContext *s = avctx->priv_data;
1626 s->xxch_present = 0;
1628 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1629 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1630 if (s->dca_buffer_size == -1) {
1631 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1635 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1636 if (dca_parse_frame_header(s) < 0) {
1637 //seems like the frame is corrupt, try with the next one
1641 //set AVCodec values with parsed data
1642 avctx->sample_rate = s->sample_rate;
1643 avctx->bit_rate = s->bit_rate;
1645 s->profile = FF_PROFILE_DTS;
1647 for (i = 0; i < (s->sample_blocks / 8); i++) {
1648 dca_decode_block(s, 0, i);
1651 /* record number of core channels incase less than max channels are requested */
1652 num_core_channels = s->prim_channels;
1654 /* extensions start at 32-bit boundaries into bitstream */
1655 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1657 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1659 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1660 uint32_t bits = get_bits_long(&s->gb, 32);
1664 int ext_amode, xch_fsize;
1666 s->xch_base_channel = s->prim_channels;
1668 /* validate sync word using XCHFSIZE field */
1669 xch_fsize = show_bits(&s->gb, 10);
1670 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1671 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1674 /* skip length-to-end-of-frame field for the moment */
1675 skip_bits(&s->gb, 10);
1677 s->profile = FFMAX(s->profile, FF_PROFILE_DTS_ES);
1679 /* extension amode should == 1, number of channels in extension */
1680 /* AFAIK XCh is not used for more channels */
1681 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1682 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1683 " supported!\n",ext_amode);
1687 /* much like core primary audio coding header */
1688 dca_parse_audio_coding_header(s, s->xch_base_channel);
1690 for (i = 0; i < (s->sample_blocks / 8); i++) {
1691 dca_decode_block(s, s->xch_base_channel, i);
1698 /* XXCh: extended channels */
1699 /* usually found either in core or HD part in DTS-HD HRA streams,
1700 * but not in DTS-ES which contains XCh extensions instead */
1701 s->xxch_present = 1;
1702 s->profile = FFMAX(s->profile, FF_PROFILE_DTS_ES);
1706 int fsize96 = show_bits(&s->gb, 12) + 1;
1707 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1710 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1711 skip_bits(&s->gb, 12);
1712 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1713 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1716 s->profile = FFMAX(s->profile, FF_PROFILE_DTS_96_24);
1721 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1724 /* check for ExSS (HD part) */
1725 if (s->dca_buffer_size - s->frame_size > 32
1726 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1727 dca_exss_parse_header(s);
1729 avctx->profile = s->profile;
1731 channels = s->prim_channels + !!s->lfe;
1734 avctx->channel_layout = dca_core_channel_layout[s->amode];
1736 if (s->xch_present && (!avctx->request_channels ||
1737 avctx->request_channels > num_core_channels + !!s->lfe)) {
1738 avctx->channel_layout |= AV_CH_BACK_CENTER;
1740 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1741 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1743 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1746 channels = num_core_channels + !!s->lfe;
1747 s->xch_present = 0; /* disable further xch processing */
1749 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1750 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1752 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1755 if (channels > !!s->lfe &&
1756 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1759 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1761 s->output = DCA_STEREO;
1762 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1765 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1770 /* There is nothing that prevents a dts frame to change channel configuration
1771 but FFmpeg doesn't support that so only set the channels if it is previously
1772 unset. Ideally during the first probe for channels the crc should be checked
1773 and only set avctx->channels when the crc is ok. Right now the decoder could
1774 set the channels based on a broken first frame.*/
1775 if (s->is_channels_set == 0) {
1776 s->is_channels_set = 1;
1777 avctx->channels = channels;
1779 if (avctx->channels != channels) {
1780 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1781 "channels changing in stream. Skipping frame.\n");
1785 if (*data_size < (s->sample_blocks / 8) * 256 * sizeof(int16_t) * channels)
1787 *data_size = 256 / 8 * s->sample_blocks * sizeof(int16_t) * channels;
1789 /* filter to get final output */
1790 for (i = 0; i < (s->sample_blocks / 8); i++) {
1791 dca_filter_channels(s, i);
1793 /* If this was marked as a DTS-ES stream we need to subtract back- */
1794 /* channel from SL & SR to remove matrixed back-channel signal */
1795 if((s->source_pcm_res & 1) && s->xch_present) {
1796 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1797 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1798 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1800 for(j = 0; j < 256; ++j) {
1801 lt_chan[j] -= (back_chan[j] - s->add_bias) * M_SQRT1_2;
1802 rt_chan[j] -= (back_chan[j] - s->add_bias) * M_SQRT1_2;
1806 s->dsp.float_to_int16_interleave(samples, s->samples_chanptr, 256, channels);
1807 samples += 256 * channels;
1810 /* update lfe history */
1811 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1812 for (i = 0; i < 2 * s->lfe * 4; i++) {
1813 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1822 * DCA initialization
1824 * @param avctx pointer to the AVCodecContext
1827 static av_cold int dca_decode_init(AVCodecContext * avctx)
1829 DCAContext *s = avctx->priv_data;
1835 dsputil_init(&s->dsp, avctx);
1836 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1837 ff_synth_filter_init(&s->synth);
1838 ff_dcadsp_init(&s->dcadsp);
1840 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1841 s->samples_chanptr[i] = s->samples + i * 256;
1842 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1844 if (s->dsp.float_to_int16_interleave == ff_float_to_int16_interleave_c) {
1845 s->add_bias = 385.0f;
1846 s->scale_bias = 1.0 / 32768.0;
1849 s->scale_bias = 1.0;
1851 /* allow downmixing to stereo */
1852 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1853 avctx->request_channels == 2) {
1854 avctx->channels = avctx->request_channels;
1862 static av_cold int dca_decode_end(AVCodecContext * avctx)
1864 DCAContext *s = avctx->priv_data;
1865 ff_mdct_end(&s->imdct);
1869 static const AVProfile profiles[] = {
1870 { FF_PROFILE_DTS, "DTS" },
1871 { FF_PROFILE_DTS_ES, "DTS-ES" },
1872 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1873 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1874 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1875 { FF_PROFILE_UNKNOWN },
1878 AVCodec dca_decoder = {
1880 .type = AVMEDIA_TYPE_AUDIO,
1882 .priv_data_size = sizeof(DCAContext),
1883 .init = dca_decode_init,
1884 .decode = dca_decode_frame,
1885 .close = dca_decode_end,
1886 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1887 .capabilities = CODEC_CAP_CHANNEL_CONF,
1888 .profiles = NULL_IF_CONFIG_SMALL(profiles),