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 Libav.
10 * Libav 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 * Libav 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 Libav; 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 "libavutil/audioconvert.h"
41 #include "synth_filter.h"
43 #include "fmtconvert.h"
51 #define DCA_PRIM_CHANNELS_MAX (7)
52 #define DCA_SUBBANDS (32)
53 #define DCA_ABITS_MAX (32) /* Should be 28 */
54 #define DCA_SUBSUBFRAMES_MAX (4)
55 #define DCA_SUBFRAMES_MAX (16)
56 #define DCA_BLOCKS_MAX (16)
57 #define DCA_LFE_MAX (3)
73 /* these are unconfirmed but should be mostly correct */
74 enum DCAExSSSpeakerMask {
75 DCA_EXSS_FRONT_CENTER = 0x0001,
76 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
77 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
78 DCA_EXSS_LFE = 0x0008,
79 DCA_EXSS_REAR_CENTER = 0x0010,
80 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
81 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
82 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
83 DCA_EXSS_OVERHEAD = 0x0100,
84 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
85 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
86 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
87 DCA_EXSS_LFE2 = 0x1000,
88 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
89 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
90 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
93 enum DCAExtensionMask {
94 DCA_EXT_CORE = 0x001, ///< core in core substream
95 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
96 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
97 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
98 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
99 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
100 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
101 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
102 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
103 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
106 /* -1 are reserved or unknown */
107 static const int dca_ext_audio_descr_mask[] = {
111 DCA_EXT_XCH | DCA_EXT_X96,
118 /* extensions that reside in core substream */
119 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
121 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
122 * Some compromises have been made for special configurations. Most configurations
123 * are never used so complete accuracy is not needed.
125 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
126 * S -> side, when both rear and back are configured move one of them to the side channel
128 * All 2 channel configurations -> AV_CH_LAYOUT_STEREO
131 static const uint64_t dca_core_channel_layout[] = {
132 AV_CH_FRONT_CENTER, ///< 1, A
133 AV_CH_LAYOUT_STEREO, ///< 2, A + B (dual mono)
134 AV_CH_LAYOUT_STEREO, ///< 2, L + R (stereo)
135 AV_CH_LAYOUT_STEREO, ///< 2, (L+R) + (L-R) (sum-difference)
136 AV_CH_LAYOUT_STEREO, ///< 2, LT +RT (left and right total)
137 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER, ///< 3, C+L+R
138 AV_CH_LAYOUT_STEREO|AV_CH_BACK_CENTER, ///< 3, L+R+S
139 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_BACK_CENTER, ///< 4, C + L + R+ S
140 AV_CH_LAYOUT_STEREO|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 4, L + R +SL+ SR
141 AV_CH_LAYOUT_STEREO|AV_CH_FRONT_CENTER|AV_CH_SIDE_LEFT|AV_CH_SIDE_RIGHT, ///< 5, C + L + R+ SL+SR
142 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
143 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
144 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
145 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
146 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
147 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
150 static const int8_t dca_lfe_index[] = {
151 1,2,2,2,2,3,2,3,2,3,2,3,1,3,2,3
154 static const int8_t dca_channel_reorder_lfe[][9] = {
155 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
156 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
157 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
158 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
159 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
160 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
161 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
162 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
163 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
164 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
165 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
166 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
167 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
168 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
169 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
170 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
173 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
174 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
175 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
176 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
177 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
178 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
179 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
180 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
181 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
182 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
183 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
184 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
185 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
186 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
187 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
188 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
189 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
192 static const int8_t dca_channel_reorder_nolfe[][9] = {
193 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
194 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
195 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
196 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
197 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
198 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
199 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
200 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
201 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
202 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
203 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
204 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
205 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
206 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
207 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
208 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
211 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
212 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
213 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
214 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
215 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
216 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
217 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
218 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
219 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
220 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
221 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
222 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
223 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
224 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
225 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
226 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
227 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
230 #define DCA_DOLBY 101 /* FIXME */
232 #define DCA_CHANNEL_BITS 6
233 #define DCA_CHANNEL_MASK 0x3F
237 #define HEADER_SIZE 14
239 #define DCA_MAX_FRAME_SIZE 16384
240 #define DCA_MAX_EXSS_HEADER_SIZE 4096
242 #define DCA_BUFFER_PADDING_SIZE 1024
244 /** Bit allocation */
246 int offset; ///< code values offset
247 int maxbits[8]; ///< max bits in VLC
248 int wrap; ///< wrap for get_vlc2()
249 VLC vlc[8]; ///< actual codes
252 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
253 static BitAlloc dca_tmode; ///< transition mode VLCs
254 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
255 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
257 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba, int idx)
259 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) + ba->offset;
263 AVCodecContext *avctx;
266 int frame_type; ///< type of the current frame
267 int samples_deficit; ///< deficit sample count
268 int crc_present; ///< crc is present in the bitstream
269 int sample_blocks; ///< number of PCM sample blocks
270 int frame_size; ///< primary frame byte size
271 int amode; ///< audio channels arrangement
272 int sample_rate; ///< audio sampling rate
273 int bit_rate; ///< transmission bit rate
274 int bit_rate_index; ///< transmission bit rate index
276 int downmix; ///< embedded downmix enabled
277 int dynrange; ///< embedded dynamic range flag
278 int timestamp; ///< embedded time stamp flag
279 int aux_data; ///< auxiliary data flag
280 int hdcd; ///< source material is mastered in HDCD
281 int ext_descr; ///< extension audio descriptor flag
282 int ext_coding; ///< extended coding flag
283 int aspf; ///< audio sync word insertion flag
284 int lfe; ///< low frequency effects flag
285 int predictor_history; ///< predictor history flag
286 int header_crc; ///< header crc check bytes
287 int multirate_inter; ///< multirate interpolator switch
288 int version; ///< encoder software revision
289 int copy_history; ///< copy history
290 int source_pcm_res; ///< source pcm resolution
291 int front_sum; ///< front sum/difference flag
292 int surround_sum; ///< surround sum/difference flag
293 int dialog_norm; ///< dialog normalisation parameter
295 /* Primary audio coding header */
296 int subframes; ///< number of subframes
297 int is_channels_set; ///< check for if the channel number is already set
298 int total_channels; ///< number of channels including extensions
299 int prim_channels; ///< number of primary audio channels
300 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
301 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
302 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
303 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
304 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
305 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
306 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
307 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
309 /* Primary audio coding side information */
310 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
311 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
312 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
313 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
314 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
315 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
316 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
317 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
318 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
319 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
320 int dynrange_coef; ///< dynamic range coefficient
322 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
324 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
325 int lfe_scale_factor;
327 /* Subband samples history (for ADPCM) */
328 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
329 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
330 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
331 int hist_index[DCA_PRIM_CHANNELS_MAX];
332 DECLARE_ALIGNED(32, float, raXin)[32];
334 int output; ///< type of output
335 float scale_bias; ///< output scale
337 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
338 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
339 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
341 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
342 int dca_buffer_size; ///< how much data is in the dca_buffer
344 const int8_t* channel_order_tab; ///< channel reordering table, lfe and non lfe
346 /* Current position in DCA frame */
347 int current_subframe;
348 int current_subsubframe;
350 int core_ext_mask; ///< present extensions in the core substream
352 /* XCh extension information */
353 int xch_present; ///< XCh extension present and valid
354 int xch_base_channel; ///< index of first (only) channel containing XCH data
356 /* ExSS header parser */
357 int static_fields; ///< static fields present
358 int mix_metadata; ///< mixing metadata present
359 int num_mix_configs; ///< number of mix out configurations
360 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
364 int debug_flag; ///< used for suppressing repeated error messages output
367 SynthFilterContext synth;
368 DCADSPContext dcadsp;
369 FmtConvertContext fmt_conv;
372 static const uint16_t dca_vlc_offs[] = {
373 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
374 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
375 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
376 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
377 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
378 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
381 static av_cold void dca_init_vlcs(void)
383 static int vlcs_initialized = 0;
385 static VLC_TYPE dca_table[23622][2];
387 if (vlcs_initialized)
390 dca_bitalloc_index.offset = 1;
391 dca_bitalloc_index.wrap = 2;
392 for (i = 0; i < 5; i++) {
393 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
394 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
395 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
396 bitalloc_12_bits[i], 1, 1,
397 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
399 dca_scalefactor.offset = -64;
400 dca_scalefactor.wrap = 2;
401 for (i = 0; i < 5; i++) {
402 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
403 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
404 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
405 scales_bits[i], 1, 1,
406 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
408 dca_tmode.offset = 0;
410 for (i = 0; i < 4; i++) {
411 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
412 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
413 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
415 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
418 for (i = 0; i < 10; i++)
419 for (j = 0; j < 7; j++){
420 if (!bitalloc_codes[i][j]) break;
421 dca_smpl_bitalloc[i+1].offset = bitalloc_offsets[i];
422 dca_smpl_bitalloc[i+1].wrap = 1 + (j > 4);
423 dca_smpl_bitalloc[i+1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
424 dca_smpl_bitalloc[i+1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
425 init_vlc(&dca_smpl_bitalloc[i+1].vlc[j], bitalloc_maxbits[i][j],
427 bitalloc_bits[i][j], 1, 1,
428 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
431 vlcs_initialized = 1;
434 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
437 *dst++ = get_bits(gb, bits);
440 static int dca_parse_audio_coding_header(DCAContext * s, int base_channel)
443 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
444 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
445 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
447 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
448 s->prim_channels = s->total_channels;
450 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
451 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
454 for (i = base_channel; i < s->prim_channels; i++) {
455 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
456 if (s->subband_activity[i] > DCA_SUBBANDS)
457 s->subband_activity[i] = DCA_SUBBANDS;
459 for (i = base_channel; i < s->prim_channels; i++) {
460 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
461 if (s->vq_start_subband[i] > DCA_SUBBANDS)
462 s->vq_start_subband[i] = DCA_SUBBANDS;
464 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
465 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
466 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
467 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
469 /* Get codebooks quantization indexes */
471 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
472 for (j = 1; j < 11; j++)
473 for (i = base_channel; i < s->prim_channels; i++)
474 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
476 /* Get scale factor adjustment */
477 for (j = 0; j < 11; j++)
478 for (i = base_channel; i < s->prim_channels; i++)
479 s->scalefactor_adj[i][j] = 1;
481 for (j = 1; j < 11; j++)
482 for (i = base_channel; i < s->prim_channels; i++)
483 if (s->quant_index_huffman[i][j] < thr[j])
484 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
486 if (s->crc_present) {
487 /* Audio header CRC check */
488 get_bits(&s->gb, 16);
491 s->current_subframe = 0;
492 s->current_subsubframe = 0;
495 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
496 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
497 for (i = base_channel; i < s->prim_channels; i++){
498 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n", s->subband_activity[i]);
499 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n", s->vq_start_subband[i]);
500 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n", s->joint_intensity[i]);
501 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n", s->transient_huffman[i]);
502 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n", s->scalefactor_huffman[i]);
503 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n", s->bitalloc_huffman[i]);
504 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
505 for (j = 0; j < 11; j++)
506 av_log(s->avctx, AV_LOG_DEBUG, " %i",
507 s->quant_index_huffman[i][j]);
508 av_log(s->avctx, AV_LOG_DEBUG, "\n");
509 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
510 for (j = 0; j < 11; j++)
511 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
512 av_log(s->avctx, AV_LOG_DEBUG, "\n");
519 static int dca_parse_frame_header(DCAContext * s)
521 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
524 skip_bits_long(&s->gb, 32);
527 s->frame_type = get_bits(&s->gb, 1);
528 s->samples_deficit = get_bits(&s->gb, 5) + 1;
529 s->crc_present = get_bits(&s->gb, 1);
530 s->sample_blocks = get_bits(&s->gb, 7) + 1;
531 s->frame_size = get_bits(&s->gb, 14) + 1;
532 if (s->frame_size < 95)
533 return AVERROR_INVALIDDATA;
534 s->amode = get_bits(&s->gb, 6);
535 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
537 return AVERROR_INVALIDDATA;
538 s->bit_rate_index = get_bits(&s->gb, 5);
539 s->bit_rate = dca_bit_rates[s->bit_rate_index];
541 return AVERROR_INVALIDDATA;
543 s->downmix = get_bits(&s->gb, 1);
544 s->dynrange = get_bits(&s->gb, 1);
545 s->timestamp = get_bits(&s->gb, 1);
546 s->aux_data = get_bits(&s->gb, 1);
547 s->hdcd = get_bits(&s->gb, 1);
548 s->ext_descr = get_bits(&s->gb, 3);
549 s->ext_coding = get_bits(&s->gb, 1);
550 s->aspf = get_bits(&s->gb, 1);
551 s->lfe = get_bits(&s->gb, 2);
552 s->predictor_history = get_bits(&s->gb, 1);
554 /* TODO: check CRC */
556 s->header_crc = get_bits(&s->gb, 16);
558 s->multirate_inter = get_bits(&s->gb, 1);
559 s->version = get_bits(&s->gb, 4);
560 s->copy_history = get_bits(&s->gb, 2);
561 s->source_pcm_res = get_bits(&s->gb, 3);
562 s->front_sum = get_bits(&s->gb, 1);
563 s->surround_sum = get_bits(&s->gb, 1);
564 s->dialog_norm = get_bits(&s->gb, 4);
566 /* FIXME: channels mixing levels */
567 s->output = s->amode;
568 if (s->lfe) s->output |= DCA_LFE;
571 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
572 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
573 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
574 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
575 s->sample_blocks, s->sample_blocks * 32);
576 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
577 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
578 s->amode, dca_channels[s->amode]);
579 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
581 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
583 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
584 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
585 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
586 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
587 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
588 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
589 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
590 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
591 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
592 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
593 s->predictor_history);
594 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
595 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
597 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
598 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
599 av_log(s->avctx, AV_LOG_DEBUG,
600 "source pcm resolution: %i (%i bits/sample)\n",
601 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
602 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
603 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
604 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
605 av_log(s->avctx, AV_LOG_DEBUG, "\n");
608 /* Primary audio coding header */
609 s->subframes = get_bits(&s->gb, 4) + 1;
611 return dca_parse_audio_coding_header(s, 0);
615 static inline int get_scale(GetBitContext *gb, int level, int value)
618 /* huffman encoded */
619 value += get_bitalloc(gb, &dca_scalefactor, level);
620 } else if (level < 8)
621 value = get_bits(gb, level + 1);
625 static int dca_subframe_header(DCAContext * s, int base_channel, int block_index)
627 /* Primary audio coding side information */
630 if (get_bits_left(&s->gb) < 0)
631 return AVERROR_INVALIDDATA;
634 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
635 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
638 for (j = base_channel; j < s->prim_channels; j++) {
639 for (k = 0; k < s->subband_activity[j]; k++)
640 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
643 /* Get prediction codebook */
644 for (j = base_channel; j < s->prim_channels; j++) {
645 for (k = 0; k < s->subband_activity[j]; k++) {
646 if (s->prediction_mode[j][k] > 0) {
647 /* (Prediction coefficient VQ address) */
648 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
653 /* Bit allocation index */
654 for (j = base_channel; j < s->prim_channels; j++) {
655 for (k = 0; k < s->vq_start_subband[j]; k++) {
656 if (s->bitalloc_huffman[j] == 6)
657 s->bitalloc[j][k] = get_bits(&s->gb, 5);
658 else if (s->bitalloc_huffman[j] == 5)
659 s->bitalloc[j][k] = get_bits(&s->gb, 4);
660 else if (s->bitalloc_huffman[j] == 7) {
661 av_log(s->avctx, AV_LOG_ERROR,
662 "Invalid bit allocation index\n");
663 return AVERROR_INVALIDDATA;
666 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
669 if (s->bitalloc[j][k] > 26) {
670 // av_log(s->avctx,AV_LOG_DEBUG,"bitalloc index [%i][%i] too big (%i)\n",
671 // j, k, s->bitalloc[j][k]);
672 return AVERROR_INVALIDDATA;
677 /* Transition mode */
678 for (j = base_channel; j < s->prim_channels; j++) {
679 for (k = 0; k < s->subband_activity[j]; k++) {
680 s->transition_mode[j][k] = 0;
681 if (s->subsubframes[s->current_subframe] > 1 &&
682 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
683 s->transition_mode[j][k] =
684 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
689 if (get_bits_left(&s->gb) < 0)
690 return AVERROR_INVALIDDATA;
692 for (j = base_channel; j < s->prim_channels; j++) {
693 const uint32_t *scale_table;
696 memset(s->scale_factor[j], 0, s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
698 if (s->scalefactor_huffman[j] == 6)
699 scale_table = scale_factor_quant7;
701 scale_table = scale_factor_quant6;
703 /* When huffman coded, only the difference is encoded */
706 for (k = 0; k < s->subband_activity[j]; k++) {
707 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
708 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
709 s->scale_factor[j][k][0] = scale_table[scale_sum];
712 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
713 /* Get second scale factor */
714 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
715 s->scale_factor[j][k][1] = scale_table[scale_sum];
720 /* Joint subband scale factor codebook select */
721 for (j = base_channel; j < s->prim_channels; j++) {
722 /* Transmitted only if joint subband coding enabled */
723 if (s->joint_intensity[j] > 0)
724 s->joint_huff[j] = get_bits(&s->gb, 3);
727 if (get_bits_left(&s->gb) < 0)
728 return AVERROR_INVALIDDATA;
730 /* Scale factors for joint subband coding */
731 for (j = base_channel; j < s->prim_channels; j++) {
734 /* Transmitted only if joint subband coding enabled */
735 if (s->joint_intensity[j] > 0) {
737 source_channel = s->joint_intensity[j] - 1;
739 /* When huffman coded, only the difference is encoded
740 * (is this valid as well for joint scales ???) */
742 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
743 scale = get_scale(&s->gb, s->joint_huff[j], 0);
744 scale += 64; /* bias */
745 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
748 if (!(s->debug_flag & 0x02)) {
749 av_log(s->avctx, AV_LOG_DEBUG,
750 "Joint stereo coding not supported\n");
751 s->debug_flag |= 0x02;
756 /* Stereo downmix coefficients */
757 if (!base_channel && s->prim_channels > 2) {
759 for (j = base_channel; j < s->prim_channels; j++) {
760 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
761 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
764 int am = s->amode & DCA_CHANNEL_MASK;
765 for (j = base_channel; j < s->prim_channels; j++) {
766 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
767 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
772 /* Dynamic range coefficient */
773 if (!base_channel && s->dynrange)
774 s->dynrange_coef = get_bits(&s->gb, 8);
776 /* Side information CRC check word */
777 if (s->crc_present) {
778 get_bits(&s->gb, 16);
782 * Primary audio data arrays
785 /* VQ encoded high frequency subbands */
786 for (j = base_channel; j < s->prim_channels; j++)
787 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
788 /* 1 vector -> 32 samples */
789 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
791 /* Low frequency effect data */
792 if (!base_channel && s->lfe) {
794 int lfe_samples = 2 * s->lfe * (4 + block_index);
795 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
798 for (j = lfe_samples; j < lfe_end_sample; j++) {
799 /* Signed 8 bits int */
800 s->lfe_data[j] = get_sbits(&s->gb, 8);
803 /* Scale factor index */
804 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
806 /* Quantization step size * scale factor */
807 lfe_scale = 0.035 * s->lfe_scale_factor;
809 for (j = lfe_samples; j < lfe_end_sample; j++)
810 s->lfe_data[j] *= lfe_scale;
814 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n", s->subsubframes[s->current_subframe]);
815 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
816 s->partial_samples[s->current_subframe]);
817 for (j = base_channel; j < s->prim_channels; j++) {
818 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
819 for (k = 0; k < s->subband_activity[j]; k++)
820 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
821 av_log(s->avctx, AV_LOG_DEBUG, "\n");
823 for (j = base_channel; j < s->prim_channels; j++) {
824 for (k = 0; k < s->subband_activity[j]; k++)
825 av_log(s->avctx, AV_LOG_DEBUG,
826 "prediction coefs: %f, %f, %f, %f\n",
827 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
828 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
829 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
830 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
832 for (j = base_channel; j < s->prim_channels; j++) {
833 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
834 for (k = 0; k < s->vq_start_subband[j]; k++)
835 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
836 av_log(s->avctx, AV_LOG_DEBUG, "\n");
838 for (j = base_channel; j < s->prim_channels; j++) {
839 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
840 for (k = 0; k < s->subband_activity[j]; k++)
841 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
842 av_log(s->avctx, AV_LOG_DEBUG, "\n");
844 for (j = base_channel; j < s->prim_channels; j++) {
845 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
846 for (k = 0; k < s->subband_activity[j]; k++) {
847 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
848 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
849 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
850 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
852 av_log(s->avctx, AV_LOG_DEBUG, "\n");
854 for (j = base_channel; j < s->prim_channels; j++) {
855 if (s->joint_intensity[j] > 0) {
856 int source_channel = s->joint_intensity[j] - 1;
857 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
858 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
859 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
860 av_log(s->avctx, AV_LOG_DEBUG, "\n");
863 if (!base_channel && s->prim_channels > 2 && s->downmix) {
864 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
865 for (j = 0; j < s->prim_channels; j++) {
866 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][0]]);
867 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1,%d = %f\n", j, dca_downmix_coeffs[s->downmix_coef[j][1]]);
869 av_log(s->avctx, AV_LOG_DEBUG, "\n");
871 for (j = base_channel; j < s->prim_channels; j++)
872 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
873 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
874 if (!base_channel && s->lfe) {
875 int lfe_samples = 2 * s->lfe * (4 + block_index);
876 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
878 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
879 for (j = lfe_samples; j < lfe_end_sample; j++)
880 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
881 av_log(s->avctx, AV_LOG_DEBUG, "\n");
888 static void qmf_32_subbands(DCAContext * s, int chans,
889 float samples_in[32][8], float *samples_out,
892 const float *prCoeff;
895 int sb_act = s->subband_activity[chans];
898 scale *= sqrt(1/8.0);
901 if (!s->multirate_inter) /* Non-perfect reconstruction */
902 prCoeff = fir_32bands_nonperfect;
903 else /* Perfect reconstruction */
904 prCoeff = fir_32bands_perfect;
906 for (i = sb_act; i < 32; i++)
909 /* Reconstructed channel sample index */
910 for (subindex = 0; subindex < 8; subindex++) {
911 /* Load in one sample from each subband and clear inactive subbands */
912 for (i = 0; i < sb_act; i++){
913 unsigned sign = (i - 1) & 2;
914 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
915 AV_WN32A(&s->raXin[i], v);
918 s->synth.synth_filter_float(&s->imdct,
919 s->subband_fir_hist[chans], &s->hist_index[chans],
920 s->subband_fir_noidea[chans], prCoeff,
921 samples_out, s->raXin, scale);
927 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
928 int num_deci_sample, float *samples_in,
929 float *samples_out, float scale)
931 /* samples_in: An array holding decimated samples.
932 * Samples in current subframe starts from samples_in[0],
933 * while samples_in[-1], samples_in[-2], ..., stores samples
934 * from last subframe as history.
936 * samples_out: An array holding interpolated samples
940 const float *prCoeff;
943 /* Select decimation filter */
944 if (decimation_select == 1) {
946 prCoeff = lfe_fir_128;
949 prCoeff = lfe_fir_64;
952 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
953 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
956 samples_out += 2 * decifactor;
960 /* downmixing routines */
961 #define MIX_REAR1(samples, si1, rs, coef) \
962 samples[i] += samples[si1] * coef[rs][0]; \
963 samples[i+256] += samples[si1] * coef[rs][1];
965 #define MIX_REAR2(samples, si1, si2, rs, coef) \
966 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs+1][0]; \
967 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs+1][1];
969 #define MIX_FRONT3(samples, coef) \
973 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
974 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
976 #define DOWNMIX_TO_STEREO(op1, op2) \
977 for (i = 0; i < 256; i++){ \
982 static void dca_downmix(float *samples, int srcfmt,
983 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
984 const int8_t *channel_mapping)
989 float coef[DCA_PRIM_CHANNELS_MAX][2];
991 for (i=0; i<DCA_PRIM_CHANNELS_MAX; i++) {
992 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
993 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
999 case DCA_STEREO_TOTAL:
1000 case DCA_STEREO_SUMDIFF:
1002 av_log(NULL, 0, "Not implemented!\n");
1007 c = channel_mapping[0] * 256;
1008 l = channel_mapping[1] * 256;
1009 r = channel_mapping[2] * 256;
1010 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),);
1013 s = channel_mapping[2] * 256;
1014 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef),);
1017 c = channel_mapping[0] * 256;
1018 l = channel_mapping[1] * 256;
1019 r = channel_mapping[2] * 256;
1020 s = channel_mapping[3] * 256;
1021 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1022 MIX_REAR1(samples, i + s, 3, coef));
1025 sl = channel_mapping[2] * 256;
1026 sr = channel_mapping[3] * 256;
1027 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef),);
1030 c = channel_mapping[0] * 256;
1031 l = channel_mapping[1] * 256;
1032 r = channel_mapping[2] * 256;
1033 sl = channel_mapping[3] * 256;
1034 sr = channel_mapping[4] * 256;
1035 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1036 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1042 #ifndef decode_blockcodes
1043 /* Very compact version of the block code decoder that does not use table
1044 * look-up but is slightly slower */
1045 static int decode_blockcode(int code, int levels, int *values)
1048 int offset = (levels - 1) >> 1;
1050 for (i = 0; i < 4; i++) {
1051 int div = FASTDIV(code, levels);
1052 values[i] = code - offset - div*levels;
1059 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1061 return decode_blockcode(code1, levels, values) |
1062 decode_blockcode(code2, levels, values + 4);
1066 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1067 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1069 #ifndef int8x8_fmul_int32
1070 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1072 float fscale = scale / 16.0;
1074 for (i = 0; i < 8; i++)
1075 dst[i] = src[i] * fscale;
1079 static int dca_subsubframe(DCAContext * s, int base_channel, int block_index)
1082 int subsubframe = s->current_subsubframe;
1084 const float *quant_step_table;
1087 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1088 LOCAL_ALIGNED_16(int, block, [8]);
1094 /* Select quantization step size table */
1095 if (s->bit_rate_index == 0x1f)
1096 quant_step_table = lossless_quant_d;
1098 quant_step_table = lossy_quant_d;
1100 for (k = base_channel; k < s->prim_channels; k++) {
1101 if (get_bits_left(&s->gb) < 0)
1102 return AVERROR_INVALIDDATA;
1104 for (l = 0; l < s->vq_start_subband[k]; l++) {
1107 /* Select the mid-tread linear quantizer */
1108 int abits = s->bitalloc[k][l];
1110 float quant_step_size = quant_step_table[abits];
1113 * Determine quantization index code book and its type
1116 /* Select quantization index code book */
1117 int sel = s->quant_index_huffman[k][abits];
1120 * Extract bits from the bit stream
1123 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1125 /* Deal with transients */
1126 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1127 float rscale = quant_step_size * s->scale_factor[k][l][sfi] * s->scalefactor_adj[k][sel];
1129 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table){
1132 int block_code1, block_code2, size, levels, err;
1134 size = abits_sizes[abits-1];
1135 levels = abits_levels[abits-1];
1137 block_code1 = get_bits(&s->gb, size);
1138 block_code2 = get_bits(&s->gb, size);
1139 err = decode_blockcodes(block_code1, block_code2,
1142 av_log(s->avctx, AV_LOG_ERROR,
1143 "ERROR: block code look-up failed\n");
1144 return AVERROR_INVALIDDATA;
1148 for (m = 0; m < 8; m++)
1149 block[m] = get_sbits(&s->gb, abits - 3);
1153 for (m = 0; m < 8; m++)
1154 block[m] = get_bitalloc(&s->gb, &dca_smpl_bitalloc[abits], sel);
1157 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1162 * Inverse ADPCM if in prediction mode
1164 if (s->prediction_mode[k][l]) {
1166 for (m = 0; m < 8; m++) {
1167 for (n = 1; n <= 4; n++)
1169 subband_samples[k][l][m] +=
1170 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1171 subband_samples[k][l][m - n] / 8192);
1172 else if (s->predictor_history)
1173 subband_samples[k][l][m] +=
1174 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1175 s->subband_samples_hist[k][l][m - n +
1182 * Decode VQ encoded high frequencies
1184 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1185 /* 1 vector -> 32 samples but we only need the 8 samples
1186 * for this subsubframe. */
1187 int hfvq = s->high_freq_vq[k][l];
1189 if (!s->debug_flag & 0x01) {
1190 av_log(s->avctx, AV_LOG_DEBUG, "Stream with high frequencies VQ coding\n");
1191 s->debug_flag |= 0x01;
1194 int8x8_fmul_int32(subband_samples[k][l],
1195 &high_freq_vq[hfvq][subsubframe * 8],
1196 s->scale_factor[k][l][0]);
1200 /* Check for DSYNC after subsubframe */
1201 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1202 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1204 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1207 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1211 /* Backup predictor history for adpcm */
1212 for (k = base_channel; k < s->prim_channels; k++)
1213 for (l = 0; l < s->vq_start_subband[k]; l++)
1214 memcpy(s->subband_samples_hist[k][l], &subband_samples[k][l][4],
1215 4 * sizeof(subband_samples[0][0][0]));
1220 static int dca_filter_channels(DCAContext * s, int block_index)
1222 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1225 /* 32 subbands QMF */
1226 for (k = 0; k < s->prim_channels; k++) {
1227 /* static float pcm_to_double[8] =
1228 {32768.0, 32768.0, 524288.0, 524288.0, 0, 8388608.0, 8388608.0};*/
1229 qmf_32_subbands(s, k, subband_samples[k], &s->samples[256 * s->channel_order_tab[k]],
1230 M_SQRT1_2*s->scale_bias /*pcm_to_double[s->source_pcm_res] */ );
1234 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1235 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1238 /* Generate LFE samples for this subsubframe FIXME!!! */
1239 if (s->output & DCA_LFE) {
1240 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1241 s->lfe_data + 2 * s->lfe * (block_index + 4),
1242 &s->samples[256 * dca_lfe_index[s->amode]],
1243 (1.0/256.0)*s->scale_bias);
1244 /* Outputs 20bits pcm samples */
1251 static int dca_subframe_footer(DCAContext * s, int base_channel)
1253 int aux_data_count = 0, i;
1256 * Unpack optional information
1259 /* presumably optional information only appears in the core? */
1260 if (!base_channel) {
1262 skip_bits_long(&s->gb, 32);
1265 aux_data_count = get_bits(&s->gb, 6);
1267 for (i = 0; i < aux_data_count; i++)
1268 get_bits(&s->gb, 8);
1270 if (s->crc_present && (s->downmix || s->dynrange))
1271 get_bits(&s->gb, 16);
1278 * Decode a dca frame block
1280 * @param s pointer to the DCAContext
1283 static int dca_decode_block(DCAContext * s, int base_channel, int block_index)
1288 if (s->current_subframe >= s->subframes) {
1289 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1290 s->current_subframe, s->subframes);
1291 return AVERROR_INVALIDDATA;
1294 if (!s->current_subsubframe) {
1296 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1298 /* Read subframe header */
1299 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1303 /* Read subsubframe */
1305 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1307 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1311 s->current_subsubframe++;
1312 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1313 s->current_subsubframe = 0;
1314 s->current_subframe++;
1316 if (s->current_subframe >= s->subframes) {
1318 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1320 /* Read subframe footer */
1321 if ((ret = dca_subframe_footer(s, base_channel)))
1329 * Convert bitstream to one representation based on sync marker
1331 static int dca_convert_bitstream(const uint8_t * src, int src_size, uint8_t * dst,
1336 const uint16_t *ssrc = (const uint16_t *) src;
1337 uint16_t *sdst = (uint16_t *) dst;
1340 if ((unsigned)src_size > (unsigned)max_size) {
1341 // av_log(NULL, AV_LOG_ERROR, "Input frame size larger than DCA_MAX_FRAME_SIZE!\n");
1343 src_size = max_size;
1348 case DCA_MARKER_RAW_BE:
1349 memcpy(dst, src, src_size);
1351 case DCA_MARKER_RAW_LE:
1352 for (i = 0; i < (src_size + 1) >> 1; i++)
1353 *sdst++ = av_bswap16(*ssrc++);
1355 case DCA_MARKER_14B_BE:
1356 case DCA_MARKER_14B_LE:
1357 init_put_bits(&pb, dst, max_size);
1358 for (i = 0; i < (src_size + 1) >> 1; i++, src += 2) {
1359 tmp = ((mrk == DCA_MARKER_14B_BE) ? AV_RB16(src) : AV_RL16(src)) & 0x3FFF;
1360 put_bits(&pb, 14, tmp);
1362 flush_put_bits(&pb);
1363 return (put_bits_count(&pb) + 7) >> 3;
1365 return AVERROR_INVALIDDATA;
1370 * Return the number of channels in an ExSS speaker mask (HD)
1372 static int dca_exss_mask2count(int mask)
1374 /* count bits that mean speaker pairs twice */
1375 return av_popcount(mask)
1376 + av_popcount(mask & (
1377 DCA_EXSS_CENTER_LEFT_RIGHT
1378 | DCA_EXSS_FRONT_LEFT_RIGHT
1379 | DCA_EXSS_FRONT_HIGH_LEFT_RIGHT
1380 | DCA_EXSS_WIDE_LEFT_RIGHT
1381 | DCA_EXSS_SIDE_LEFT_RIGHT
1382 | DCA_EXSS_SIDE_HIGH_LEFT_RIGHT
1383 | DCA_EXSS_SIDE_REAR_LEFT_RIGHT
1384 | DCA_EXSS_REAR_LEFT_RIGHT
1385 | DCA_EXSS_REAR_HIGH_LEFT_RIGHT
1390 * Skip mixing coefficients of a single mix out configuration (HD)
1392 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1396 for (i = 0; i < channels; i++) {
1397 int mix_map_mask = get_bits(gb, out_ch);
1398 int num_coeffs = av_popcount(mix_map_mask);
1399 skip_bits_long(gb, num_coeffs * 6);
1404 * Parse extension substream asset header (HD)
1406 static int dca_exss_parse_asset_header(DCAContext *s)
1408 int header_pos = get_bits_count(&s->gb);
1411 int embedded_stereo = 0;
1412 int embedded_6ch = 0;
1413 int drc_code_present;
1414 int extensions_mask;
1417 if (get_bits_left(&s->gb) < 16)
1420 /* We will parse just enough to get to the extensions bitmask with which
1421 * we can set the profile value. */
1423 header_size = get_bits(&s->gb, 9) + 1;
1424 skip_bits(&s->gb, 3); // asset index
1426 if (s->static_fields) {
1427 if (get_bits1(&s->gb))
1428 skip_bits(&s->gb, 4); // asset type descriptor
1429 if (get_bits1(&s->gb))
1430 skip_bits_long(&s->gb, 24); // language descriptor
1432 if (get_bits1(&s->gb)) {
1433 /* How can one fit 1024 bytes of text here if the maximum value
1434 * for the asset header size field above was 512 bytes? */
1435 int text_length = get_bits(&s->gb, 10) + 1;
1436 if (get_bits_left(&s->gb) < text_length * 8)
1438 skip_bits_long(&s->gb, text_length * 8); // info text
1441 skip_bits(&s->gb, 5); // bit resolution - 1
1442 skip_bits(&s->gb, 4); // max sample rate code
1443 channels = get_bits(&s->gb, 8) + 1;
1445 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1446 int spkr_remap_sets;
1447 int spkr_mask_size = 16;
1451 embedded_stereo = get_bits1(&s->gb);
1453 embedded_6ch = get_bits1(&s->gb);
1455 if (get_bits1(&s->gb)) {
1456 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1457 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1460 spkr_remap_sets = get_bits(&s->gb, 3);
1462 for (i = 0; i < spkr_remap_sets; i++) {
1463 /* std layout mask for each remap set */
1464 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1467 for (i = 0; i < spkr_remap_sets; i++) {
1468 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1469 if (get_bits_left(&s->gb) < 0)
1472 for (j = 0; j < num_spkrs[i]; j++) {
1473 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1474 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1475 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1480 skip_bits(&s->gb, 3); // representation type
1484 drc_code_present = get_bits1(&s->gb);
1485 if (drc_code_present)
1486 get_bits(&s->gb, 8); // drc code
1488 if (get_bits1(&s->gb))
1489 skip_bits(&s->gb, 5); // dialog normalization code
1491 if (drc_code_present && embedded_stereo)
1492 get_bits(&s->gb, 8); // drc stereo code
1494 if (s->mix_metadata && get_bits1(&s->gb)) {
1495 skip_bits(&s->gb, 1); // external mix
1496 skip_bits(&s->gb, 6); // post mix gain code
1498 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1499 skip_bits(&s->gb, 3); // drc limit
1501 skip_bits(&s->gb, 8); // custom drc code
1503 if (get_bits1(&s->gb)) // channel specific scaling
1504 for (i = 0; i < s->num_mix_configs; i++)
1505 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1507 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1509 for (i = 0; i < s->num_mix_configs; i++) {
1510 if (get_bits_left(&s->gb) < 0)
1512 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1514 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1515 if (embedded_stereo)
1516 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1520 switch (get_bits(&s->gb, 2)) {
1521 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1522 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1523 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1524 case 3: extensions_mask = 0; /* aux coding */ break;
1527 /* not parsed further, we were only interested in the extensions mask */
1529 if (get_bits_left(&s->gb) < 0)
1532 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1533 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1536 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1538 if (extensions_mask & DCA_EXT_EXSS_XLL)
1539 s->profile = FF_PROFILE_DTS_HD_MA;
1540 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1542 s->profile = FF_PROFILE_DTS_HD_HRA;
1544 if (!(extensions_mask & DCA_EXT_CORE))
1545 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1546 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1547 av_log(s->avctx, AV_LOG_WARNING, "DTS extensions detection mismatch (%d, %d)\n",
1548 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1554 * Parse extension substream header (HD)
1556 static void dca_exss_parse_header(DCAContext *s)
1562 int active_ss_mask[8];
1565 if (get_bits_left(&s->gb) < 52)
1568 skip_bits(&s->gb, 8); // user data
1569 ss_index = get_bits(&s->gb, 2);
1571 blownup = get_bits1(&s->gb);
1572 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1573 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1575 s->static_fields = get_bits1(&s->gb);
1576 if (s->static_fields) {
1577 skip_bits(&s->gb, 2); // reference clock code
1578 skip_bits(&s->gb, 3); // frame duration code
1580 if (get_bits1(&s->gb))
1581 skip_bits_long(&s->gb, 36); // timestamp
1583 /* a single stream can contain multiple audio assets that can be
1584 * combined to form multiple audio presentations */
1586 num_audiop = get_bits(&s->gb, 3) + 1;
1587 if (num_audiop > 1) {
1588 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1589 /* ignore such streams for now */
1593 num_assets = get_bits(&s->gb, 3) + 1;
1594 if (num_assets > 1) {
1595 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1596 /* ignore such streams for now */
1600 for (i = 0; i < num_audiop; i++)
1601 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1603 for (i = 0; i < num_audiop; i++)
1604 for (j = 0; j <= ss_index; j++)
1605 if (active_ss_mask[i] & (1 << j))
1606 skip_bits(&s->gb, 8); // active asset mask
1608 s->mix_metadata = get_bits1(&s->gb);
1609 if (s->mix_metadata) {
1610 int mix_out_mask_size;
1612 skip_bits(&s->gb, 2); // adjustment level
1613 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1614 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1616 for (i = 0; i < s->num_mix_configs; i++) {
1617 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1618 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1623 for (i = 0; i < num_assets; i++)
1624 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1626 for (i = 0; i < num_assets; i++) {
1627 if (dca_exss_parse_asset_header(s))
1631 /* not parsed further, we were only interested in the extensions mask
1632 * from the asset header */
1636 * Main frame decoding function
1637 * FIXME add arguments
1639 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1640 int *got_frame_ptr, AVPacket *avpkt)
1642 const uint8_t *buf = avpkt->data;
1643 int buf_size = avpkt->size;
1646 int num_core_channels = 0;
1649 int16_t *samples_s16;
1650 DCAContext *s = avctx->priv_data;
1657 s->dca_buffer_size = dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1658 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1659 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1660 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1661 return AVERROR_INVALIDDATA;
1664 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1665 if ((ret = dca_parse_frame_header(s)) < 0) {
1666 //seems like the frame is corrupt, try with the next one
1669 //set AVCodec values with parsed data
1670 avctx->sample_rate = s->sample_rate;
1671 avctx->bit_rate = s->bit_rate;
1672 avctx->frame_size = s->sample_blocks * 32;
1674 s->profile = FF_PROFILE_DTS;
1676 for (i = 0; i < (s->sample_blocks / 8); i++) {
1677 if ((ret = dca_decode_block(s, 0, i))) {
1678 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1683 /* record number of core channels incase less than max channels are requested */
1684 num_core_channels = s->prim_channels;
1687 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1689 s->core_ext_mask = 0;
1691 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1693 /* only scan for extensions if ext_descr was unknown or indicated a
1694 * supported XCh extension */
1695 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1697 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1698 * extensions scan can fill it up */
1699 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1701 /* extensions start at 32-bit boundaries into bitstream */
1702 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1704 while(core_ss_end - get_bits_count(&s->gb) >= 32) {
1705 uint32_t bits = get_bits_long(&s->gb, 32);
1709 int ext_amode, xch_fsize;
1711 s->xch_base_channel = s->prim_channels;
1713 /* validate sync word using XCHFSIZE field */
1714 xch_fsize = show_bits(&s->gb, 10);
1715 if((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1716 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1719 /* skip length-to-end-of-frame field for the moment */
1720 skip_bits(&s->gb, 10);
1722 s->core_ext_mask |= DCA_EXT_XCH;
1724 /* extension amode should == 1, number of channels in extension */
1725 /* AFAIK XCh is not used for more channels */
1726 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1727 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1728 " supported!\n",ext_amode);
1732 /* much like core primary audio coding header */
1733 dca_parse_audio_coding_header(s, s->xch_base_channel);
1735 for (i = 0; i < (s->sample_blocks / 8); i++) {
1736 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1737 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1746 /* XXCh: extended channels */
1747 /* usually found either in core or HD part in DTS-HD HRA streams,
1748 * but not in DTS-ES which contains XCh extensions instead */
1749 s->core_ext_mask |= DCA_EXT_XXCH;
1753 int fsize96 = show_bits(&s->gb, 12) + 1;
1754 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1757 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n", get_bits_count(&s->gb));
1758 skip_bits(&s->gb, 12);
1759 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1760 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1762 s->core_ext_mask |= DCA_EXT_X96;
1767 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1771 /* no supported extensions, skip the rest of the core substream */
1772 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1775 if (s->core_ext_mask & DCA_EXT_X96)
1776 s->profile = FF_PROFILE_DTS_96_24;
1777 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1778 s->profile = FF_PROFILE_DTS_ES;
1780 /* check for ExSS (HD part) */
1781 if (s->dca_buffer_size - s->frame_size > 32
1782 && get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1783 dca_exss_parse_header(s);
1785 avctx->profile = s->profile;
1787 channels = s->prim_channels + !!s->lfe;
1790 avctx->channel_layout = dca_core_channel_layout[s->amode];
1792 if (s->xch_present && (!avctx->request_channels ||
1793 avctx->request_channels > num_core_channels + !!s->lfe)) {
1794 avctx->channel_layout |= AV_CH_BACK_CENTER;
1796 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1797 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1799 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1802 channels = num_core_channels + !!s->lfe;
1803 s->xch_present = 0; /* disable further xch processing */
1805 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1806 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1808 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1811 if (channels > !!s->lfe &&
1812 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1813 return AVERROR_INVALIDDATA;
1815 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1817 s->output = DCA_STEREO;
1818 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1821 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n",s->amode);
1822 return AVERROR_INVALIDDATA;
1826 /* There is nothing that prevents a dts frame to change channel configuration
1827 but Libav doesn't support that so only set the channels if it is previously
1828 unset. Ideally during the first probe for channels the crc should be checked
1829 and only set avctx->channels when the crc is ok. Right now the decoder could
1830 set the channels based on a broken first frame.*/
1831 if (s->is_channels_set == 0) {
1832 s->is_channels_set = 1;
1833 avctx->channels = channels;
1835 if (avctx->channels != channels) {
1836 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1837 "channels changing in stream. Skipping frame.\n");
1838 return AVERROR_PATCHWELCOME;
1841 /* get output buffer */
1842 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
1843 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
1844 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1847 samples_flt = (float *)s->frame.data[0];
1848 samples_s16 = (int16_t *)s->frame.data[0];
1850 /* filter to get final output */
1851 for (i = 0; i < (s->sample_blocks / 8); i++) {
1852 dca_filter_channels(s, i);
1854 /* If this was marked as a DTS-ES stream we need to subtract back- */
1855 /* channel from SL & SR to remove matrixed back-channel signal */
1856 if((s->source_pcm_res & 1) && s->xch_present) {
1857 float* back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1858 float* lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1859 float* rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1860 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1861 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1864 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1865 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1867 samples_flt += 256 * channels;
1869 s->fmt_conv.float_to_int16_interleave(samples_s16,
1870 s->samples_chanptr, 256,
1872 samples_s16 += 256 * channels;
1876 /* update lfe history */
1877 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1878 for (i = 0; i < 2 * s->lfe * 4; i++) {
1879 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1883 *(AVFrame *)data = s->frame;
1891 * DCA initialization
1893 * @param avctx pointer to the AVCodecContext
1896 static av_cold int dca_decode_init(AVCodecContext * avctx)
1898 DCAContext *s = avctx->priv_data;
1904 dsputil_init(&s->dsp, avctx);
1905 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1906 ff_synth_filter_init(&s->synth);
1907 ff_dcadsp_init(&s->dcadsp);
1908 ff_fmt_convert_init(&s->fmt_conv, avctx);
1910 for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
1911 s->samples_chanptr[i] = s->samples + i * 256;
1913 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1914 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1915 s->scale_bias = 1.0 / 32768.0;
1917 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1918 s->scale_bias = 1.0;
1921 /* allow downmixing to stereo */
1922 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1923 avctx->request_channels == 2) {
1924 avctx->channels = avctx->request_channels;
1927 avcodec_get_frame_defaults(&s->frame);
1928 avctx->coded_frame = &s->frame;
1933 static av_cold int dca_decode_end(AVCodecContext * avctx)
1935 DCAContext *s = avctx->priv_data;
1936 ff_mdct_end(&s->imdct);
1940 static const AVProfile profiles[] = {
1941 { FF_PROFILE_DTS, "DTS" },
1942 { FF_PROFILE_DTS_ES, "DTS-ES" },
1943 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1944 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1945 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1946 { FF_PROFILE_UNKNOWN },
1949 AVCodec ff_dca_decoder = {
1951 .type = AVMEDIA_TYPE_AUDIO,
1953 .priv_data_size = sizeof(DCAContext),
1954 .init = dca_decode_init,
1955 .decode = dca_decode_frame,
1956 .close = dca_decode_end,
1957 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1958 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1959 .sample_fmts = (const enum AVSampleFormat[]) {
1960 AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
1962 .profiles = NULL_IF_CONFIG_SMALL(profiles),