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 "dca_parser.h"
42 #include "synth_filter.h"
44 #include "fmtconvert.h"
52 #define DCA_PRIM_CHANNELS_MAX (7)
53 #define DCA_SUBBANDS (32)
54 #define DCA_ABITS_MAX (32) /* Should be 28 */
55 #define DCA_SUBSUBFRAMES_MAX (4)
56 #define DCA_SUBFRAMES_MAX (16)
57 #define DCA_BLOCKS_MAX (16)
58 #define DCA_LFE_MAX (3)
74 /* these are unconfirmed but should be mostly correct */
75 enum DCAExSSSpeakerMask {
76 DCA_EXSS_FRONT_CENTER = 0x0001,
77 DCA_EXSS_FRONT_LEFT_RIGHT = 0x0002,
78 DCA_EXSS_SIDE_REAR_LEFT_RIGHT = 0x0004,
79 DCA_EXSS_LFE = 0x0008,
80 DCA_EXSS_REAR_CENTER = 0x0010,
81 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT = 0x0020,
82 DCA_EXSS_REAR_LEFT_RIGHT = 0x0040,
83 DCA_EXSS_FRONT_HIGH_CENTER = 0x0080,
84 DCA_EXSS_OVERHEAD = 0x0100,
85 DCA_EXSS_CENTER_LEFT_RIGHT = 0x0200,
86 DCA_EXSS_WIDE_LEFT_RIGHT = 0x0400,
87 DCA_EXSS_SIDE_LEFT_RIGHT = 0x0800,
88 DCA_EXSS_LFE2 = 0x1000,
89 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT = 0x2000,
90 DCA_EXSS_REAR_HIGH_CENTER = 0x4000,
91 DCA_EXSS_REAR_HIGH_LEFT_RIGHT = 0x8000,
94 enum DCAExtensionMask {
95 DCA_EXT_CORE = 0x001, ///< core in core substream
96 DCA_EXT_XXCH = 0x002, ///< XXCh channels extension in core substream
97 DCA_EXT_X96 = 0x004, ///< 96/24 extension in core substream
98 DCA_EXT_XCH = 0x008, ///< XCh channel extension in core substream
99 DCA_EXT_EXSS_CORE = 0x010, ///< core in ExSS (extension substream)
100 DCA_EXT_EXSS_XBR = 0x020, ///< extended bitrate extension in ExSS
101 DCA_EXT_EXSS_XXCH = 0x040, ///< XXCh channels extension in ExSS
102 DCA_EXT_EXSS_X96 = 0x080, ///< 96/24 extension in ExSS
103 DCA_EXT_EXSS_LBR = 0x100, ///< low bitrate component in ExSS
104 DCA_EXT_EXSS_XLL = 0x200, ///< lossless extension in ExSS
107 /* -1 are reserved or unknown */
108 static const int dca_ext_audio_descr_mask[] = {
112 DCA_EXT_XCH | DCA_EXT_X96,
119 /* extensions that reside in core substream */
120 #define DCA_CORE_EXTS (DCA_EXT_XCH | DCA_EXT_XXCH | DCA_EXT_X96)
122 /* Tables for mapping dts channel configurations to libavcodec multichannel api.
123 * Some compromises have been made for special configurations. Most configurations
124 * are never used so complete accuracy is not needed.
126 * L = left, R = right, C = center, S = surround, F = front, R = rear, T = total, OV = overhead.
127 * S -> side, when both rear and back are configured move one of them to the side channel
129 * 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
142 AV_CH_LAYOUT_STEREO | AV_CH_FRONT_CENTER | AV_CH_SIDE_LEFT |
143 AV_CH_SIDE_RIGHT, ///< 5, C + L + R + SL + SR
145 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
146 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER, ///< 6, CL + CR + L + R + SL + SR
148 AV_CH_LAYOUT_STEREO | AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT |
149 AV_CH_FRONT_CENTER | AV_CH_BACK_CENTER, ///< 6, C + L + R + LR + RR + OV
151 AV_CH_FRONT_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
152 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_BACK_CENTER |
153 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 6, CF + CR + LF + RF + LR + RR
155 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
156 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
157 AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT, ///< 7, CL + C + CR + L + R + SL + SR
159 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_RIGHT_OF_CENTER |
160 AV_CH_LAYOUT_STEREO | AV_CH_SIDE_LEFT | AV_CH_SIDE_RIGHT |
161 AV_CH_BACK_LEFT | AV_CH_BACK_RIGHT, ///< 8, CL + CR + L + R + SL1 + SL2 + SR1 + SR2
163 AV_CH_FRONT_LEFT_OF_CENTER | AV_CH_FRONT_CENTER |
164 AV_CH_FRONT_RIGHT_OF_CENTER | AV_CH_LAYOUT_STEREO |
165 AV_CH_SIDE_LEFT | AV_CH_BACK_CENTER | AV_CH_SIDE_RIGHT, ///< 8, CL + C + CR + L + R + SL + S + SR
168 static const int8_t dca_lfe_index[] = {
169 1, 2, 2, 2, 2, 3, 2, 3, 2, 3, 2, 3, 1, 3, 2, 3
172 static const int8_t dca_channel_reorder_lfe[][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, 3, -1, -1, -1, -1, -1, -1},
180 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
181 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
182 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
183 { 3, 4, 0, 1, 5, 6, -1, -1, -1},
184 { 2, 0, 1, 4, 5, 6, -1, -1, -1},
185 { 0, 6, 4, 5, 2, 3, -1, -1, -1},
186 { 4, 2, 5, 0, 1, 6, 7, -1, -1},
187 { 5, 6, 0, 1, 7, 3, 8, 4, -1},
188 { 4, 2, 5, 0, 1, 6, 8, 7, -1},
191 static const int8_t dca_channel_reorder_lfe_xch[][9] = {
192 { 0, 2, -1, -1, -1, -1, -1, -1, -1},
193 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
194 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
195 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
196 { 0, 1, 3, -1, -1, -1, -1, -1, -1},
197 { 2, 0, 1, 4, -1, -1, -1, -1, -1},
198 { 0, 1, 3, 4, -1, -1, -1, -1, -1},
199 { 2, 0, 1, 4, 5, -1, -1, -1, -1},
200 { 0, 1, 4, 5, 3, -1, -1, -1, -1},
201 { 2, 0, 1, 5, 6, 4, -1, -1, -1},
202 { 3, 4, 0, 1, 6, 7, 5, -1, -1},
203 { 2, 0, 1, 4, 5, 6, 7, -1, -1},
204 { 0, 6, 4, 5, 2, 3, 7, -1, -1},
205 { 4, 2, 5, 0, 1, 7, 8, 6, -1},
206 { 5, 6, 0, 1, 8, 3, 9, 4, 7},
207 { 4, 2, 5, 0, 1, 6, 9, 8, 7},
210 static const int8_t dca_channel_reorder_nolfe[][9] = {
211 { 0, -1, -1, -1, -1, -1, -1, -1, -1},
212 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
213 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
214 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
215 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
216 { 2, 0, 1, -1, -1, -1, -1, -1, -1},
217 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
218 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
219 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
220 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
221 { 2, 3, 0, 1, 4, 5, -1, -1, -1},
222 { 2, 0, 1, 3, 4, 5, -1, -1, -1},
223 { 0, 5, 3, 4, 1, 2, -1, -1, -1},
224 { 3, 2, 4, 0, 1, 5, 6, -1, -1},
225 { 4, 5, 0, 1, 6, 2, 7, 3, -1},
226 { 3, 2, 4, 0, 1, 5, 7, 6, -1},
229 static const int8_t dca_channel_reorder_nolfe_xch[][9] = {
230 { 0, 1, -1, -1, -1, -1, -1, -1, -1},
231 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
232 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
233 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
234 { 0, 1, 2, -1, -1, -1, -1, -1, -1},
235 { 2, 0, 1, 3, -1, -1, -1, -1, -1},
236 { 0, 1, 2, 3, -1, -1, -1, -1, -1},
237 { 2, 0, 1, 3, 4, -1, -1, -1, -1},
238 { 0, 1, 3, 4, 2, -1, -1, -1, -1},
239 { 2, 0, 1, 4, 5, 3, -1, -1, -1},
240 { 2, 3, 0, 1, 5, 6, 4, -1, -1},
241 { 2, 0, 1, 3, 4, 5, 6, -1, -1},
242 { 0, 5, 3, 4, 1, 2, 6, -1, -1},
243 { 3, 2, 4, 0, 1, 6, 7, 5, -1},
244 { 4, 5, 0, 1, 7, 2, 8, 3, 6},
245 { 3, 2, 4, 0, 1, 5, 8, 7, 6},
248 #define DCA_DOLBY 101 /* FIXME */
250 #define DCA_CHANNEL_BITS 6
251 #define DCA_CHANNEL_MASK 0x3F
255 #define HEADER_SIZE 14
257 #define DCA_MAX_FRAME_SIZE 16384
258 #define DCA_MAX_EXSS_HEADER_SIZE 4096
260 #define DCA_BUFFER_PADDING_SIZE 1024
262 /** Bit allocation */
264 int offset; ///< code values offset
265 int maxbits[8]; ///< max bits in VLC
266 int wrap; ///< wrap for get_vlc2()
267 VLC vlc[8]; ///< actual codes
270 static BitAlloc dca_bitalloc_index; ///< indexes for samples VLC select
271 static BitAlloc dca_tmode; ///< transition mode VLCs
272 static BitAlloc dca_scalefactor; ///< scalefactor VLCs
273 static BitAlloc dca_smpl_bitalloc[11]; ///< samples VLCs
275 static av_always_inline int get_bitalloc(GetBitContext *gb, BitAlloc *ba,
278 return get_vlc2(gb, ba->vlc[idx].table, ba->vlc[idx].bits, ba->wrap) +
283 AVCodecContext *avctx;
286 int frame_type; ///< type of the current frame
287 int samples_deficit; ///< deficit sample count
288 int crc_present; ///< crc is present in the bitstream
289 int sample_blocks; ///< number of PCM sample blocks
290 int frame_size; ///< primary frame byte size
291 int amode; ///< audio channels arrangement
292 int sample_rate; ///< audio sampling rate
293 int bit_rate; ///< transmission bit rate
294 int bit_rate_index; ///< transmission bit rate index
296 int downmix; ///< embedded downmix enabled
297 int dynrange; ///< embedded dynamic range flag
298 int timestamp; ///< embedded time stamp flag
299 int aux_data; ///< auxiliary data flag
300 int hdcd; ///< source material is mastered in HDCD
301 int ext_descr; ///< extension audio descriptor flag
302 int ext_coding; ///< extended coding flag
303 int aspf; ///< audio sync word insertion flag
304 int lfe; ///< low frequency effects flag
305 int predictor_history; ///< predictor history flag
306 int header_crc; ///< header crc check bytes
307 int multirate_inter; ///< multirate interpolator switch
308 int version; ///< encoder software revision
309 int copy_history; ///< copy history
310 int source_pcm_res; ///< source pcm resolution
311 int front_sum; ///< front sum/difference flag
312 int surround_sum; ///< surround sum/difference flag
313 int dialog_norm; ///< dialog normalisation parameter
315 /* Primary audio coding header */
316 int subframes; ///< number of subframes
317 int is_channels_set; ///< check for if the channel number is already set
318 int total_channels; ///< number of channels including extensions
319 int prim_channels; ///< number of primary audio channels
320 int subband_activity[DCA_PRIM_CHANNELS_MAX]; ///< subband activity count
321 int vq_start_subband[DCA_PRIM_CHANNELS_MAX]; ///< high frequency vq start subband
322 int joint_intensity[DCA_PRIM_CHANNELS_MAX]; ///< joint intensity coding index
323 int transient_huffman[DCA_PRIM_CHANNELS_MAX]; ///< transient mode code book
324 int scalefactor_huffman[DCA_PRIM_CHANNELS_MAX]; ///< scale factor code book
325 int bitalloc_huffman[DCA_PRIM_CHANNELS_MAX]; ///< bit allocation quantizer select
326 int quant_index_huffman[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< quantization index codebook select
327 float scalefactor_adj[DCA_PRIM_CHANNELS_MAX][DCA_ABITS_MAX]; ///< scale factor adjustment
329 /* Primary audio coding side information */
330 int subsubframes[DCA_SUBFRAMES_MAX]; ///< number of subsubframes
331 int partial_samples[DCA_SUBFRAMES_MAX]; ///< partial subsubframe samples count
332 int prediction_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction mode (ADPCM used or not)
333 int prediction_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< prediction VQ coefs
334 int bitalloc[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< bit allocation index
335 int transition_mode[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< transition mode (transients)
336 int scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][2]; ///< scale factors (2 if transient)
337 int joint_huff[DCA_PRIM_CHANNELS_MAX]; ///< joint subband scale factors codebook
338 int joint_scale_factor[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< joint subband scale factors
339 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2]; ///< stereo downmix coefficients
340 int dynrange_coef; ///< dynamic range coefficient
342 int high_freq_vq[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS]; ///< VQ encoded high frequency subbands
344 float lfe_data[2 * DCA_LFE_MAX * (DCA_BLOCKS_MAX + 4)]; ///< Low frequency effect data
345 int lfe_scale_factor;
347 /* Subband samples history (for ADPCM) */
348 DECLARE_ALIGNED(16, float, subband_samples_hist)[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
349 DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
350 DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
351 int hist_index[DCA_PRIM_CHANNELS_MAX];
352 DECLARE_ALIGNED(32, float, raXin)[32];
354 int output; ///< type of output
355 float scale_bias; ///< output scale
357 DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
358 DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX + 1) * 256];
359 const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX + 1];
361 uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
362 int dca_buffer_size; ///< how much data is in the dca_buffer
364 const int8_t *channel_order_tab; ///< channel reordering table, lfe and non lfe
366 /* Current position in DCA frame */
367 int current_subframe;
368 int current_subsubframe;
370 int core_ext_mask; ///< present extensions in the core substream
372 /* XCh extension information */
373 int xch_present; ///< XCh extension present and valid
374 int xch_base_channel; ///< index of first (only) channel containing XCH data
376 /* ExSS header parser */
377 int static_fields; ///< static fields present
378 int mix_metadata; ///< mixing metadata present
379 int num_mix_configs; ///< number of mix out configurations
380 int mix_config_num_ch[4]; ///< number of channels in each mix out configuration
384 int debug_flag; ///< used for suppressing repeated error messages output
387 SynthFilterContext synth;
388 DCADSPContext dcadsp;
389 FmtConvertContext fmt_conv;
392 static const uint16_t dca_vlc_offs[] = {
393 0, 512, 640, 768, 1282, 1794, 2436, 3080, 3770, 4454, 5364,
394 5372, 5380, 5388, 5392, 5396, 5412, 5420, 5428, 5460, 5492, 5508,
395 5572, 5604, 5668, 5796, 5860, 5892, 6412, 6668, 6796, 7308, 7564,
396 7820, 8076, 8620, 9132, 9388, 9910, 10166, 10680, 11196, 11726, 12240,
397 12752, 13298, 13810, 14326, 14840, 15500, 16022, 16540, 17158, 17678, 18264,
398 18796, 19352, 19926, 20468, 21472, 22398, 23014, 23622,
401 static av_cold void dca_init_vlcs(void)
403 static int vlcs_initialized = 0;
405 static VLC_TYPE dca_table[23622][2];
407 if (vlcs_initialized)
410 dca_bitalloc_index.offset = 1;
411 dca_bitalloc_index.wrap = 2;
412 for (i = 0; i < 5; i++) {
413 dca_bitalloc_index.vlc[i].table = &dca_table[dca_vlc_offs[i]];
414 dca_bitalloc_index.vlc[i].table_allocated = dca_vlc_offs[i + 1] - dca_vlc_offs[i];
415 init_vlc(&dca_bitalloc_index.vlc[i], bitalloc_12_vlc_bits[i], 12,
416 bitalloc_12_bits[i], 1, 1,
417 bitalloc_12_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
419 dca_scalefactor.offset = -64;
420 dca_scalefactor.wrap = 2;
421 for (i = 0; i < 5; i++) {
422 dca_scalefactor.vlc[i].table = &dca_table[dca_vlc_offs[i + 5]];
423 dca_scalefactor.vlc[i].table_allocated = dca_vlc_offs[i + 6] - dca_vlc_offs[i + 5];
424 init_vlc(&dca_scalefactor.vlc[i], SCALES_VLC_BITS, 129,
425 scales_bits[i], 1, 1,
426 scales_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
428 dca_tmode.offset = 0;
430 for (i = 0; i < 4; i++) {
431 dca_tmode.vlc[i].table = &dca_table[dca_vlc_offs[i + 10]];
432 dca_tmode.vlc[i].table_allocated = dca_vlc_offs[i + 11] - dca_vlc_offs[i + 10];
433 init_vlc(&dca_tmode.vlc[i], tmode_vlc_bits[i], 4,
435 tmode_codes[i], 2, 2, INIT_VLC_USE_NEW_STATIC);
438 for (i = 0; i < 10; i++)
439 for (j = 0; j < 7; j++) {
440 if (!bitalloc_codes[i][j])
442 dca_smpl_bitalloc[i + 1].offset = bitalloc_offsets[i];
443 dca_smpl_bitalloc[i + 1].wrap = 1 + (j > 4);
444 dca_smpl_bitalloc[i + 1].vlc[j].table = &dca_table[dca_vlc_offs[c]];
445 dca_smpl_bitalloc[i + 1].vlc[j].table_allocated = dca_vlc_offs[c + 1] - dca_vlc_offs[c];
447 init_vlc(&dca_smpl_bitalloc[i + 1].vlc[j], bitalloc_maxbits[i][j],
449 bitalloc_bits[i][j], 1, 1,
450 bitalloc_codes[i][j], 2, 2, INIT_VLC_USE_NEW_STATIC);
453 vlcs_initialized = 1;
456 static inline void get_array(GetBitContext *gb, int *dst, int len, int bits)
459 *dst++ = get_bits(gb, bits);
462 static int dca_parse_audio_coding_header(DCAContext *s, int base_channel)
465 static const float adj_table[4] = { 1.0, 1.1250, 1.2500, 1.4375 };
466 static const int bitlen[11] = { 0, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3 };
467 static const int thr[11] = { 0, 1, 3, 3, 3, 3, 7, 7, 7, 7, 7 };
469 s->total_channels = get_bits(&s->gb, 3) + 1 + base_channel;
470 s->prim_channels = s->total_channels;
472 if (s->prim_channels > DCA_PRIM_CHANNELS_MAX)
473 s->prim_channels = DCA_PRIM_CHANNELS_MAX;
476 for (i = base_channel; i < s->prim_channels; i++) {
477 s->subband_activity[i] = get_bits(&s->gb, 5) + 2;
478 if (s->subband_activity[i] > DCA_SUBBANDS)
479 s->subband_activity[i] = DCA_SUBBANDS;
481 for (i = base_channel; i < s->prim_channels; i++) {
482 s->vq_start_subband[i] = get_bits(&s->gb, 5) + 1;
483 if (s->vq_start_subband[i] > DCA_SUBBANDS)
484 s->vq_start_subband[i] = DCA_SUBBANDS;
486 get_array(&s->gb, s->joint_intensity + base_channel, s->prim_channels - base_channel, 3);
487 get_array(&s->gb, s->transient_huffman + base_channel, s->prim_channels - base_channel, 2);
488 get_array(&s->gb, s->scalefactor_huffman + base_channel, s->prim_channels - base_channel, 3);
489 get_array(&s->gb, s->bitalloc_huffman + base_channel, s->prim_channels - base_channel, 3);
491 /* Get codebooks quantization indexes */
493 memset(s->quant_index_huffman, 0, sizeof(s->quant_index_huffman));
494 for (j = 1; j < 11; j++)
495 for (i = base_channel; i < s->prim_channels; i++)
496 s->quant_index_huffman[i][j] = get_bits(&s->gb, bitlen[j]);
498 /* Get scale factor adjustment */
499 for (j = 0; j < 11; j++)
500 for (i = base_channel; i < s->prim_channels; i++)
501 s->scalefactor_adj[i][j] = 1;
503 for (j = 1; j < 11; j++)
504 for (i = base_channel; i < s->prim_channels; i++)
505 if (s->quant_index_huffman[i][j] < thr[j])
506 s->scalefactor_adj[i][j] = adj_table[get_bits(&s->gb, 2)];
508 if (s->crc_present) {
509 /* Audio header CRC check */
510 get_bits(&s->gb, 16);
513 s->current_subframe = 0;
514 s->current_subsubframe = 0;
517 av_log(s->avctx, AV_LOG_DEBUG, "subframes: %i\n", s->subframes);
518 av_log(s->avctx, AV_LOG_DEBUG, "prim channels: %i\n", s->prim_channels);
519 for (i = base_channel; i < s->prim_channels; i++) {
520 av_log(s->avctx, AV_LOG_DEBUG, "subband activity: %i\n",
521 s->subband_activity[i]);
522 av_log(s->avctx, AV_LOG_DEBUG, "vq start subband: %i\n",
523 s->vq_start_subband[i]);
524 av_log(s->avctx, AV_LOG_DEBUG, "joint intensity: %i\n",
525 s->joint_intensity[i]);
526 av_log(s->avctx, AV_LOG_DEBUG, "transient mode codebook: %i\n",
527 s->transient_huffman[i]);
528 av_log(s->avctx, AV_LOG_DEBUG, "scale factor codebook: %i\n",
529 s->scalefactor_huffman[i]);
530 av_log(s->avctx, AV_LOG_DEBUG, "bit allocation quantizer: %i\n",
531 s->bitalloc_huffman[i]);
532 av_log(s->avctx, AV_LOG_DEBUG, "quant index huff:");
533 for (j = 0; j < 11; j++)
534 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->quant_index_huffman[i][j]);
535 av_log(s->avctx, AV_LOG_DEBUG, "\n");
536 av_log(s->avctx, AV_LOG_DEBUG, "scalefac adj:");
537 for (j = 0; j < 11; j++)
538 av_log(s->avctx, AV_LOG_DEBUG, " %1.3f", s->scalefactor_adj[i][j]);
539 av_log(s->avctx, AV_LOG_DEBUG, "\n");
546 static int dca_parse_frame_header(DCAContext *s)
548 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
551 skip_bits_long(&s->gb, 32);
554 s->frame_type = get_bits(&s->gb, 1);
555 s->samples_deficit = get_bits(&s->gb, 5) + 1;
556 s->crc_present = get_bits(&s->gb, 1);
557 s->sample_blocks = get_bits(&s->gb, 7) + 1;
558 s->frame_size = get_bits(&s->gb, 14) + 1;
559 if (s->frame_size < 95)
560 return AVERROR_INVALIDDATA;
561 s->amode = get_bits(&s->gb, 6);
562 s->sample_rate = dca_sample_rates[get_bits(&s->gb, 4)];
564 return AVERROR_INVALIDDATA;
565 s->bit_rate_index = get_bits(&s->gb, 5);
566 s->bit_rate = dca_bit_rates[s->bit_rate_index];
568 return AVERROR_INVALIDDATA;
570 s->downmix = get_bits(&s->gb, 1);
571 s->dynrange = get_bits(&s->gb, 1);
572 s->timestamp = get_bits(&s->gb, 1);
573 s->aux_data = get_bits(&s->gb, 1);
574 s->hdcd = get_bits(&s->gb, 1);
575 s->ext_descr = get_bits(&s->gb, 3);
576 s->ext_coding = get_bits(&s->gb, 1);
577 s->aspf = get_bits(&s->gb, 1);
578 s->lfe = get_bits(&s->gb, 2);
579 s->predictor_history = get_bits(&s->gb, 1);
581 /* TODO: check CRC */
583 s->header_crc = get_bits(&s->gb, 16);
585 s->multirate_inter = get_bits(&s->gb, 1);
586 s->version = get_bits(&s->gb, 4);
587 s->copy_history = get_bits(&s->gb, 2);
588 s->source_pcm_res = get_bits(&s->gb, 3);
589 s->front_sum = get_bits(&s->gb, 1);
590 s->surround_sum = get_bits(&s->gb, 1);
591 s->dialog_norm = get_bits(&s->gb, 4);
593 /* FIXME: channels mixing levels */
594 s->output = s->amode;
596 s->output |= DCA_LFE;
599 av_log(s->avctx, AV_LOG_DEBUG, "frame type: %i\n", s->frame_type);
600 av_log(s->avctx, AV_LOG_DEBUG, "samples deficit: %i\n", s->samples_deficit);
601 av_log(s->avctx, AV_LOG_DEBUG, "crc present: %i\n", s->crc_present);
602 av_log(s->avctx, AV_LOG_DEBUG, "sample blocks: %i (%i samples)\n",
603 s->sample_blocks, s->sample_blocks * 32);
604 av_log(s->avctx, AV_LOG_DEBUG, "frame size: %i bytes\n", s->frame_size);
605 av_log(s->avctx, AV_LOG_DEBUG, "amode: %i (%i channels)\n",
606 s->amode, dca_channels[s->amode]);
607 av_log(s->avctx, AV_LOG_DEBUG, "sample rate: %i Hz\n",
609 av_log(s->avctx, AV_LOG_DEBUG, "bit rate: %i bits/s\n",
611 av_log(s->avctx, AV_LOG_DEBUG, "downmix: %i\n", s->downmix);
612 av_log(s->avctx, AV_LOG_DEBUG, "dynrange: %i\n", s->dynrange);
613 av_log(s->avctx, AV_LOG_DEBUG, "timestamp: %i\n", s->timestamp);
614 av_log(s->avctx, AV_LOG_DEBUG, "aux_data: %i\n", s->aux_data);
615 av_log(s->avctx, AV_LOG_DEBUG, "hdcd: %i\n", s->hdcd);
616 av_log(s->avctx, AV_LOG_DEBUG, "ext descr: %i\n", s->ext_descr);
617 av_log(s->avctx, AV_LOG_DEBUG, "ext coding: %i\n", s->ext_coding);
618 av_log(s->avctx, AV_LOG_DEBUG, "aspf: %i\n", s->aspf);
619 av_log(s->avctx, AV_LOG_DEBUG, "lfe: %i\n", s->lfe);
620 av_log(s->avctx, AV_LOG_DEBUG, "predictor history: %i\n",
621 s->predictor_history);
622 av_log(s->avctx, AV_LOG_DEBUG, "header crc: %i\n", s->header_crc);
623 av_log(s->avctx, AV_LOG_DEBUG, "multirate inter: %i\n",
625 av_log(s->avctx, AV_LOG_DEBUG, "version number: %i\n", s->version);
626 av_log(s->avctx, AV_LOG_DEBUG, "copy history: %i\n", s->copy_history);
627 av_log(s->avctx, AV_LOG_DEBUG,
628 "source pcm resolution: %i (%i bits/sample)\n",
629 s->source_pcm_res, dca_bits_per_sample[s->source_pcm_res]);
630 av_log(s->avctx, AV_LOG_DEBUG, "front sum: %i\n", s->front_sum);
631 av_log(s->avctx, AV_LOG_DEBUG, "surround sum: %i\n", s->surround_sum);
632 av_log(s->avctx, AV_LOG_DEBUG, "dialog norm: %i\n", s->dialog_norm);
633 av_log(s->avctx, AV_LOG_DEBUG, "\n");
636 /* Primary audio coding header */
637 s->subframes = get_bits(&s->gb, 4) + 1;
639 return dca_parse_audio_coding_header(s, 0);
643 static inline int get_scale(GetBitContext *gb, int level, int value)
646 /* huffman encoded */
647 value += get_bitalloc(gb, &dca_scalefactor, level);
648 } else if (level < 8)
649 value = get_bits(gb, level + 1);
653 static int dca_subframe_header(DCAContext *s, int base_channel, int block_index)
655 /* Primary audio coding side information */
658 if (get_bits_left(&s->gb) < 0)
659 return AVERROR_INVALIDDATA;
662 s->subsubframes[s->current_subframe] = get_bits(&s->gb, 2) + 1;
663 s->partial_samples[s->current_subframe] = get_bits(&s->gb, 3);
666 for (j = base_channel; j < s->prim_channels; j++) {
667 for (k = 0; k < s->subband_activity[j]; k++)
668 s->prediction_mode[j][k] = get_bits(&s->gb, 1);
671 /* Get prediction codebook */
672 for (j = base_channel; j < s->prim_channels; j++) {
673 for (k = 0; k < s->subband_activity[j]; k++) {
674 if (s->prediction_mode[j][k] > 0) {
675 /* (Prediction coefficient VQ address) */
676 s->prediction_vq[j][k] = get_bits(&s->gb, 12);
681 /* Bit allocation index */
682 for (j = base_channel; j < s->prim_channels; j++) {
683 for (k = 0; k < s->vq_start_subband[j]; k++) {
684 if (s->bitalloc_huffman[j] == 6)
685 s->bitalloc[j][k] = get_bits(&s->gb, 5);
686 else if (s->bitalloc_huffman[j] == 5)
687 s->bitalloc[j][k] = get_bits(&s->gb, 4);
688 else if (s->bitalloc_huffman[j] == 7) {
689 av_log(s->avctx, AV_LOG_ERROR,
690 "Invalid bit allocation index\n");
691 return AVERROR_INVALIDDATA;
694 get_bitalloc(&s->gb, &dca_bitalloc_index, s->bitalloc_huffman[j]);
697 if (s->bitalloc[j][k] > 26) {
698 // av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index [%i][%i] too big (%i)\n",
699 // j, k, s->bitalloc[j][k]);
700 return AVERROR_INVALIDDATA;
705 /* Transition mode */
706 for (j = base_channel; j < s->prim_channels; j++) {
707 for (k = 0; k < s->subband_activity[j]; k++) {
708 s->transition_mode[j][k] = 0;
709 if (s->subsubframes[s->current_subframe] > 1 &&
710 k < s->vq_start_subband[j] && s->bitalloc[j][k] > 0) {
711 s->transition_mode[j][k] =
712 get_bitalloc(&s->gb, &dca_tmode, s->transient_huffman[j]);
717 if (get_bits_left(&s->gb) < 0)
718 return AVERROR_INVALIDDATA;
720 for (j = base_channel; j < s->prim_channels; j++) {
721 const uint32_t *scale_table;
724 memset(s->scale_factor[j], 0,
725 s->subband_activity[j] * sizeof(s->scale_factor[0][0][0]) * 2);
727 if (s->scalefactor_huffman[j] == 6)
728 scale_table = scale_factor_quant7;
730 scale_table = scale_factor_quant6;
732 /* When huffman coded, only the difference is encoded */
735 for (k = 0; k < s->subband_activity[j]; k++) {
736 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0) {
737 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
738 s->scale_factor[j][k][0] = scale_table[scale_sum];
741 if (k < s->vq_start_subband[j] && s->transition_mode[j][k]) {
742 /* Get second scale factor */
743 scale_sum = get_scale(&s->gb, s->scalefactor_huffman[j], scale_sum);
744 s->scale_factor[j][k][1] = scale_table[scale_sum];
749 /* Joint subband scale factor codebook select */
750 for (j = base_channel; j < s->prim_channels; j++) {
751 /* Transmitted only if joint subband coding enabled */
752 if (s->joint_intensity[j] > 0)
753 s->joint_huff[j] = get_bits(&s->gb, 3);
756 if (get_bits_left(&s->gb) < 0)
757 return AVERROR_INVALIDDATA;
759 /* Scale factors for joint subband coding */
760 for (j = base_channel; j < s->prim_channels; j++) {
763 /* Transmitted only if joint subband coding enabled */
764 if (s->joint_intensity[j] > 0) {
766 source_channel = s->joint_intensity[j] - 1;
768 /* When huffman coded, only the difference is encoded
769 * (is this valid as well for joint scales ???) */
771 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++) {
772 scale = get_scale(&s->gb, s->joint_huff[j], 0);
773 scale += 64; /* bias */
774 s->joint_scale_factor[j][k] = scale; /*joint_scale_table[scale]; */
777 if (!(s->debug_flag & 0x02)) {
778 av_log(s->avctx, AV_LOG_DEBUG,
779 "Joint stereo coding not supported\n");
780 s->debug_flag |= 0x02;
785 /* Stereo downmix coefficients */
786 if (!base_channel && s->prim_channels > 2) {
788 for (j = base_channel; j < s->prim_channels; j++) {
789 s->downmix_coef[j][0] = get_bits(&s->gb, 7);
790 s->downmix_coef[j][1] = get_bits(&s->gb, 7);
793 int am = s->amode & DCA_CHANNEL_MASK;
794 for (j = base_channel; j < s->prim_channels; j++) {
795 s->downmix_coef[j][0] = dca_default_coeffs[am][j][0];
796 s->downmix_coef[j][1] = dca_default_coeffs[am][j][1];
801 /* Dynamic range coefficient */
802 if (!base_channel && s->dynrange)
803 s->dynrange_coef = get_bits(&s->gb, 8);
805 /* Side information CRC check word */
806 if (s->crc_present) {
807 get_bits(&s->gb, 16);
811 * Primary audio data arrays
814 /* VQ encoded high frequency subbands */
815 for (j = base_channel; j < s->prim_channels; j++)
816 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
817 /* 1 vector -> 32 samples */
818 s->high_freq_vq[j][k] = get_bits(&s->gb, 10);
820 /* Low frequency effect data */
821 if (!base_channel && s->lfe) {
823 int lfe_samples = 2 * s->lfe * (4 + block_index);
824 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
827 for (j = lfe_samples; j < lfe_end_sample; j++) {
828 /* Signed 8 bits int */
829 s->lfe_data[j] = get_sbits(&s->gb, 8);
832 /* Scale factor index */
833 s->lfe_scale_factor = scale_factor_quant7[get_bits(&s->gb, 8)];
835 /* Quantization step size * scale factor */
836 lfe_scale = 0.035 * s->lfe_scale_factor;
838 for (j = lfe_samples; j < lfe_end_sample; j++)
839 s->lfe_data[j] *= lfe_scale;
843 av_log(s->avctx, AV_LOG_DEBUG, "subsubframes: %i\n",
844 s->subsubframes[s->current_subframe]);
845 av_log(s->avctx, AV_LOG_DEBUG, "partial samples: %i\n",
846 s->partial_samples[s->current_subframe]);
848 for (j = base_channel; j < s->prim_channels; j++) {
849 av_log(s->avctx, AV_LOG_DEBUG, "prediction mode:");
850 for (k = 0; k < s->subband_activity[j]; k++)
851 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->prediction_mode[j][k]);
852 av_log(s->avctx, AV_LOG_DEBUG, "\n");
854 for (j = base_channel; j < s->prim_channels; j++) {
855 for (k = 0; k < s->subband_activity[j]; k++)
856 av_log(s->avctx, AV_LOG_DEBUG,
857 "prediction coefs: %f, %f, %f, %f\n",
858 (float) adpcm_vb[s->prediction_vq[j][k]][0] / 8192,
859 (float) adpcm_vb[s->prediction_vq[j][k]][1] / 8192,
860 (float) adpcm_vb[s->prediction_vq[j][k]][2] / 8192,
861 (float) adpcm_vb[s->prediction_vq[j][k]][3] / 8192);
863 for (j = base_channel; j < s->prim_channels; j++) {
864 av_log(s->avctx, AV_LOG_DEBUG, "bitalloc index: ");
865 for (k = 0; k < s->vq_start_subband[j]; k++)
866 av_log(s->avctx, AV_LOG_DEBUG, "%2.2i ", s->bitalloc[j][k]);
867 av_log(s->avctx, AV_LOG_DEBUG, "\n");
869 for (j = base_channel; j < s->prim_channels; j++) {
870 av_log(s->avctx, AV_LOG_DEBUG, "Transition mode:");
871 for (k = 0; k < s->subband_activity[j]; k++)
872 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->transition_mode[j][k]);
873 av_log(s->avctx, AV_LOG_DEBUG, "\n");
875 for (j = base_channel; j < s->prim_channels; j++) {
876 av_log(s->avctx, AV_LOG_DEBUG, "Scale factor:");
877 for (k = 0; k < s->subband_activity[j]; k++) {
878 if (k >= s->vq_start_subband[j] || s->bitalloc[j][k] > 0)
879 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->scale_factor[j][k][0]);
880 if (k < s->vq_start_subband[j] && s->transition_mode[j][k])
881 av_log(s->avctx, AV_LOG_DEBUG, " %i(t)", s->scale_factor[j][k][1]);
883 av_log(s->avctx, AV_LOG_DEBUG, "\n");
885 for (j = base_channel; j < s->prim_channels; j++) {
886 if (s->joint_intensity[j] > 0) {
887 int source_channel = s->joint_intensity[j] - 1;
888 av_log(s->avctx, AV_LOG_DEBUG, "Joint scale factor index:\n");
889 for (k = s->subband_activity[j]; k < s->subband_activity[source_channel]; k++)
890 av_log(s->avctx, AV_LOG_DEBUG, " %i", s->joint_scale_factor[j][k]);
891 av_log(s->avctx, AV_LOG_DEBUG, "\n");
894 if (!base_channel && s->prim_channels > 2 && s->downmix) {
895 av_log(s->avctx, AV_LOG_DEBUG, "Downmix coeffs:\n");
896 for (j = 0; j < s->prim_channels; j++) {
897 av_log(s->avctx, AV_LOG_DEBUG, "Channel 0, %d = %f\n", j,
898 dca_downmix_coeffs[s->downmix_coef[j][0]]);
899 av_log(s->avctx, AV_LOG_DEBUG, "Channel 1, %d = %f\n", j,
900 dca_downmix_coeffs[s->downmix_coef[j][1]]);
902 av_log(s->avctx, AV_LOG_DEBUG, "\n");
904 for (j = base_channel; j < s->prim_channels; j++)
905 for (k = s->vq_start_subband[j]; k < s->subband_activity[j]; k++)
906 av_log(s->avctx, AV_LOG_DEBUG, "VQ index: %i\n", s->high_freq_vq[j][k]);
907 if (!base_channel && s->lfe) {
908 int lfe_samples = 2 * s->lfe * (4 + block_index);
909 int lfe_end_sample = 2 * s->lfe * (4 + block_index + s->subsubframes[s->current_subframe]);
911 av_log(s->avctx, AV_LOG_DEBUG, "LFE samples:\n");
912 for (j = lfe_samples; j < lfe_end_sample; j++)
913 av_log(s->avctx, AV_LOG_DEBUG, " %f", s->lfe_data[j]);
914 av_log(s->avctx, AV_LOG_DEBUG, "\n");
921 static void qmf_32_subbands(DCAContext *s, int chans,
922 float samples_in[32][8], float *samples_out,
925 const float *prCoeff;
928 int sb_act = s->subband_activity[chans];
931 scale *= sqrt(1 / 8.0);
934 if (!s->multirate_inter) /* Non-perfect reconstruction */
935 prCoeff = fir_32bands_nonperfect;
936 else /* Perfect reconstruction */
937 prCoeff = fir_32bands_perfect;
939 for (i = sb_act; i < 32; i++)
942 /* Reconstructed channel sample index */
943 for (subindex = 0; subindex < 8; subindex++) {
944 /* Load in one sample from each subband and clear inactive subbands */
945 for (i = 0; i < sb_act; i++) {
946 unsigned sign = (i - 1) & 2;
947 uint32_t v = AV_RN32A(&samples_in[i][subindex]) ^ sign << 30;
948 AV_WN32A(&s->raXin[i], v);
951 s->synth.synth_filter_float(&s->imdct,
952 s->subband_fir_hist[chans],
953 &s->hist_index[chans],
954 s->subband_fir_noidea[chans], prCoeff,
955 samples_out, s->raXin, scale);
960 static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
961 int num_deci_sample, float *samples_in,
962 float *samples_out, float scale)
964 /* samples_in: An array holding decimated samples.
965 * Samples in current subframe starts from samples_in[0],
966 * while samples_in[-1], samples_in[-2], ..., stores samples
967 * from last subframe as history.
969 * samples_out: An array holding interpolated samples
973 const float *prCoeff;
976 /* Select decimation filter */
977 if (decimation_select == 1) {
979 prCoeff = lfe_fir_128;
982 prCoeff = lfe_fir_64;
985 for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
986 s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor, scale);
988 samples_out += 2 * decifactor;
992 /* downmixing routines */
993 #define MIX_REAR1(samples, si1, rs, coef) \
994 samples[i] += samples[si1] * coef[rs][0]; \
995 samples[i+256] += samples[si1] * coef[rs][1];
997 #define MIX_REAR2(samples, si1, si2, rs, coef) \
998 samples[i] += samples[si1] * coef[rs][0] + samples[si2] * coef[rs + 1][0]; \
999 samples[i+256] += samples[si1] * coef[rs][1] + samples[si2] * coef[rs + 1][1];
1001 #define MIX_FRONT3(samples, coef) \
1002 t = samples[i + c]; \
1003 u = samples[i + l]; \
1004 v = samples[i + r]; \
1005 samples[i] = t * coef[0][0] + u * coef[1][0] + v * coef[2][0]; \
1006 samples[i+256] = t * coef[0][1] + u * coef[1][1] + v * coef[2][1];
1008 #define DOWNMIX_TO_STEREO(op1, op2) \
1009 for (i = 0; i < 256; i++) { \
1014 static void dca_downmix(float *samples, int srcfmt,
1015 int downmix_coef[DCA_PRIM_CHANNELS_MAX][2],
1016 const int8_t *channel_mapping)
1018 int c, l, r, sl, sr, s;
1021 float coef[DCA_PRIM_CHANNELS_MAX][2];
1023 for (i = 0; i < DCA_PRIM_CHANNELS_MAX; i++) {
1024 coef[i][0] = dca_downmix_coeffs[downmix_coef[i][0]];
1025 coef[i][1] = dca_downmix_coeffs[downmix_coef[i][1]];
1031 case DCA_STEREO_TOTAL:
1032 case DCA_STEREO_SUMDIFF:
1034 av_log(NULL, 0, "Not implemented!\n");
1039 c = channel_mapping[0] * 256;
1040 l = channel_mapping[1] * 256;
1041 r = channel_mapping[2] * 256;
1042 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef), );
1045 s = channel_mapping[2] * 256;
1046 DOWNMIX_TO_STEREO(MIX_REAR1(samples, i + s, 2, coef), );
1049 c = channel_mapping[0] * 256;
1050 l = channel_mapping[1] * 256;
1051 r = channel_mapping[2] * 256;
1052 s = channel_mapping[3] * 256;
1053 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1054 MIX_REAR1(samples, i + s, 3, coef));
1057 sl = channel_mapping[2] * 256;
1058 sr = channel_mapping[3] * 256;
1059 DOWNMIX_TO_STEREO(MIX_REAR2(samples, i + sl, i + sr, 2, coef), );
1062 c = channel_mapping[0] * 256;
1063 l = channel_mapping[1] * 256;
1064 r = channel_mapping[2] * 256;
1065 sl = channel_mapping[3] * 256;
1066 sr = channel_mapping[4] * 256;
1067 DOWNMIX_TO_STEREO(MIX_FRONT3(samples, coef),
1068 MIX_REAR2(samples, i + sl, i + sr, 3, coef));
1074 #ifndef decode_blockcodes
1075 /* Very compact version of the block code decoder that does not use table
1076 * look-up but is slightly slower */
1077 static int decode_blockcode(int code, int levels, int *values)
1080 int offset = (levels - 1) >> 1;
1082 for (i = 0; i < 4; i++) {
1083 int div = FASTDIV(code, levels);
1084 values[i] = code - offset - div * levels;
1091 static int decode_blockcodes(int code1, int code2, int levels, int *values)
1093 return decode_blockcode(code1, levels, values) |
1094 decode_blockcode(code2, levels, values + 4);
1098 static const uint8_t abits_sizes[7] = { 7, 10, 12, 13, 15, 17, 19 };
1099 static const uint8_t abits_levels[7] = { 3, 5, 7, 9, 13, 17, 25 };
1101 #ifndef int8x8_fmul_int32
1102 static inline void int8x8_fmul_int32(float *dst, const int8_t *src, int scale)
1104 float fscale = scale / 16.0;
1106 for (i = 0; i < 8; i++)
1107 dst[i] = src[i] * fscale;
1111 static int dca_subsubframe(DCAContext *s, int base_channel, int block_index)
1114 int subsubframe = s->current_subsubframe;
1116 const float *quant_step_table;
1119 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1120 LOCAL_ALIGNED_16(int, block, [8]);
1126 /* Select quantization step size table */
1127 if (s->bit_rate_index == 0x1f)
1128 quant_step_table = lossless_quant_d;
1130 quant_step_table = lossy_quant_d;
1132 for (k = base_channel; k < s->prim_channels; k++) {
1133 if (get_bits_left(&s->gb) < 0)
1134 return AVERROR_INVALIDDATA;
1136 for (l = 0; l < s->vq_start_subband[k]; l++) {
1139 /* Select the mid-tread linear quantizer */
1140 int abits = s->bitalloc[k][l];
1142 float quant_step_size = quant_step_table[abits];
1145 * Determine quantization index code book and its type
1148 /* Select quantization index code book */
1149 int sel = s->quant_index_huffman[k][abits];
1152 * Extract bits from the bit stream
1155 memset(subband_samples[k][l], 0, 8 * sizeof(subband_samples[0][0][0]));
1157 /* Deal with transients */
1158 int sfi = s->transition_mode[k][l] && subsubframe >= s->transition_mode[k][l];
1159 float rscale = quant_step_size * s->scale_factor[k][l][sfi] *
1160 s->scalefactor_adj[k][sel];
1162 if (abits >= 11 || !dca_smpl_bitalloc[abits].vlc[sel].table) {
1165 int block_code1, block_code2, size, levels, err;
1167 size = abits_sizes[abits - 1];
1168 levels = abits_levels[abits - 1];
1170 block_code1 = get_bits(&s->gb, size);
1171 block_code2 = get_bits(&s->gb, size);
1172 err = decode_blockcodes(block_code1, block_code2,
1175 av_log(s->avctx, AV_LOG_ERROR,
1176 "ERROR: block code look-up failed\n");
1177 return AVERROR_INVALIDDATA;
1181 for (m = 0; m < 8; m++)
1182 block[m] = get_sbits(&s->gb, abits - 3);
1186 for (m = 0; m < 8; m++)
1187 block[m] = get_bitalloc(&s->gb,
1188 &dca_smpl_bitalloc[abits], sel);
1191 s->fmt_conv.int32_to_float_fmul_scalar(subband_samples[k][l],
1196 * Inverse ADPCM if in prediction mode
1198 if (s->prediction_mode[k][l]) {
1200 for (m = 0; m < 8; m++) {
1201 for (n = 1; n <= 4; n++)
1203 subband_samples[k][l][m] +=
1204 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1205 subband_samples[k][l][m - n] / 8192);
1206 else if (s->predictor_history)
1207 subband_samples[k][l][m] +=
1208 (adpcm_vb[s->prediction_vq[k][l]][n - 1] *
1209 s->subband_samples_hist[k][l][m - n + 4] / 8192);
1215 * Decode VQ encoded high frequencies
1217 for (l = s->vq_start_subband[k]; l < s->subband_activity[k]; l++) {
1218 /* 1 vector -> 32 samples but we only need the 8 samples
1219 * for this subsubframe. */
1220 int hfvq = s->high_freq_vq[k][l];
1222 if (!s->debug_flag & 0x01) {
1223 av_log(s->avctx, AV_LOG_DEBUG,
1224 "Stream with high frequencies VQ coding\n");
1225 s->debug_flag |= 0x01;
1228 int8x8_fmul_int32(subband_samples[k][l],
1229 &high_freq_vq[hfvq][subsubframe * 8],
1230 s->scale_factor[k][l][0]);
1234 /* Check for DSYNC after subsubframe */
1235 if (s->aspf || subsubframe == s->subsubframes[s->current_subframe] - 1) {
1236 if (0xFFFF == get_bits(&s->gb, 16)) { /* 0xFFFF */
1238 av_log(s->avctx, AV_LOG_DEBUG, "Got subframe DSYNC\n");
1241 av_log(s->avctx, AV_LOG_ERROR, "Didn't get subframe DSYNC\n");
1245 /* Backup predictor history for adpcm */
1246 for (k = base_channel; k < s->prim_channels; k++)
1247 for (l = 0; l < s->vq_start_subband[k]; l++)
1248 memcpy(s->subband_samples_hist[k][l],
1249 &subband_samples[k][l][4],
1250 4 * sizeof(subband_samples[0][0][0]));
1255 static int dca_filter_channels(DCAContext *s, int block_index)
1257 float (*subband_samples)[DCA_SUBBANDS][8] = s->subband_samples[block_index];
1260 /* 32 subbands QMF */
1261 for (k = 0; k < s->prim_channels; k++) {
1262 /* static float pcm_to_double[8] = { 32768.0, 32768.0, 524288.0, 524288.0,
1263 0, 8388608.0, 8388608.0 };*/
1264 qmf_32_subbands(s, k, subband_samples[k],
1265 &s->samples[256 * s->channel_order_tab[k]],
1266 M_SQRT1_2 * s->scale_bias /* pcm_to_double[s->source_pcm_res] */);
1270 if (s->avctx->request_channels == 2 && s->prim_channels > 2) {
1271 dca_downmix(s->samples, s->amode, s->downmix_coef, s->channel_order_tab);
1274 /* Generate LFE samples for this subsubframe FIXME!!! */
1275 if (s->output & DCA_LFE) {
1276 lfe_interpolation_fir(s, s->lfe, 2 * s->lfe,
1277 s->lfe_data + 2 * s->lfe * (block_index + 4),
1278 &s->samples[256 * dca_lfe_index[s->amode]],
1279 (1.0 / 256.0) * s->scale_bias);
1280 /* Outputs 20bits pcm samples */
1287 static int dca_subframe_footer(DCAContext *s, int base_channel)
1289 int aux_data_count = 0, i;
1292 * Unpack optional information
1295 /* presumably optional information only appears in the core? */
1296 if (!base_channel) {
1298 skip_bits_long(&s->gb, 32);
1301 aux_data_count = get_bits(&s->gb, 6);
1303 for (i = 0; i < aux_data_count; i++)
1304 get_bits(&s->gb, 8);
1306 if (s->crc_present && (s->downmix || s->dynrange))
1307 get_bits(&s->gb, 16);
1314 * Decode a dca frame block
1316 * @param s pointer to the DCAContext
1319 static int dca_decode_block(DCAContext *s, int base_channel, int block_index)
1324 if (s->current_subframe >= s->subframes) {
1325 av_log(s->avctx, AV_LOG_DEBUG, "check failed: %i>%i",
1326 s->current_subframe, s->subframes);
1327 return AVERROR_INVALIDDATA;
1330 if (!s->current_subsubframe) {
1332 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_header\n");
1334 /* Read subframe header */
1335 if ((ret = dca_subframe_header(s, base_channel, block_index)))
1339 /* Read subsubframe */
1341 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subsubframe\n");
1343 if ((ret = dca_subsubframe(s, base_channel, block_index)))
1347 s->current_subsubframe++;
1348 if (s->current_subsubframe >= s->subsubframes[s->current_subframe]) {
1349 s->current_subsubframe = 0;
1350 s->current_subframe++;
1352 if (s->current_subframe >= s->subframes) {
1354 av_log(s->avctx, AV_LOG_DEBUG, "DSYNC dca_subframe_footer\n");
1356 /* Read subframe footer */
1357 if ((ret = dca_subframe_footer(s, base_channel)))
1365 * Return the number of channels in an ExSS speaker mask (HD)
1367 static int dca_exss_mask2count(int mask)
1369 /* count bits that mean speaker pairs twice */
1370 return av_popcount(mask) +
1371 av_popcount(mask & (DCA_EXSS_CENTER_LEFT_RIGHT |
1372 DCA_EXSS_FRONT_LEFT_RIGHT |
1373 DCA_EXSS_FRONT_HIGH_LEFT_RIGHT |
1374 DCA_EXSS_WIDE_LEFT_RIGHT |
1375 DCA_EXSS_SIDE_LEFT_RIGHT |
1376 DCA_EXSS_SIDE_HIGH_LEFT_RIGHT |
1377 DCA_EXSS_SIDE_REAR_LEFT_RIGHT |
1378 DCA_EXSS_REAR_LEFT_RIGHT |
1379 DCA_EXSS_REAR_HIGH_LEFT_RIGHT));
1383 * Skip mixing coefficients of a single mix out configuration (HD)
1385 static void dca_exss_skip_mix_coeffs(GetBitContext *gb, int channels, int out_ch)
1389 for (i = 0; i < channels; i++) {
1390 int mix_map_mask = get_bits(gb, out_ch);
1391 int num_coeffs = av_popcount(mix_map_mask);
1392 skip_bits_long(gb, num_coeffs * 6);
1397 * Parse extension substream asset header (HD)
1399 static int dca_exss_parse_asset_header(DCAContext *s)
1401 int header_pos = get_bits_count(&s->gb);
1404 int embedded_stereo = 0;
1405 int embedded_6ch = 0;
1406 int drc_code_present;
1407 int extensions_mask;
1410 if (get_bits_left(&s->gb) < 16)
1413 /* We will parse just enough to get to the extensions bitmask with which
1414 * we can set the profile value. */
1416 header_size = get_bits(&s->gb, 9) + 1;
1417 skip_bits(&s->gb, 3); // asset index
1419 if (s->static_fields) {
1420 if (get_bits1(&s->gb))
1421 skip_bits(&s->gb, 4); // asset type descriptor
1422 if (get_bits1(&s->gb))
1423 skip_bits_long(&s->gb, 24); // language descriptor
1425 if (get_bits1(&s->gb)) {
1426 /* How can one fit 1024 bytes of text here if the maximum value
1427 * for the asset header size field above was 512 bytes? */
1428 int text_length = get_bits(&s->gb, 10) + 1;
1429 if (get_bits_left(&s->gb) < text_length * 8)
1431 skip_bits_long(&s->gb, text_length * 8); // info text
1434 skip_bits(&s->gb, 5); // bit resolution - 1
1435 skip_bits(&s->gb, 4); // max sample rate code
1436 channels = get_bits(&s->gb, 8) + 1;
1438 if (get_bits1(&s->gb)) { // 1-to-1 channels to speakers
1439 int spkr_remap_sets;
1440 int spkr_mask_size = 16;
1444 embedded_stereo = get_bits1(&s->gb);
1446 embedded_6ch = get_bits1(&s->gb);
1448 if (get_bits1(&s->gb)) {
1449 spkr_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1450 skip_bits(&s->gb, spkr_mask_size); // spkr activity mask
1453 spkr_remap_sets = get_bits(&s->gb, 3);
1455 for (i = 0; i < spkr_remap_sets; i++) {
1456 /* std layout mask for each remap set */
1457 num_spkrs[i] = dca_exss_mask2count(get_bits(&s->gb, spkr_mask_size));
1460 for (i = 0; i < spkr_remap_sets; i++) {
1461 int num_dec_ch_remaps = get_bits(&s->gb, 5) + 1;
1462 if (get_bits_left(&s->gb) < 0)
1465 for (j = 0; j < num_spkrs[i]; j++) {
1466 int remap_dec_ch_mask = get_bits_long(&s->gb, num_dec_ch_remaps);
1467 int num_dec_ch = av_popcount(remap_dec_ch_mask);
1468 skip_bits_long(&s->gb, num_dec_ch * 5); // remap codes
1473 skip_bits(&s->gb, 3); // representation type
1477 drc_code_present = get_bits1(&s->gb);
1478 if (drc_code_present)
1479 get_bits(&s->gb, 8); // drc code
1481 if (get_bits1(&s->gb))
1482 skip_bits(&s->gb, 5); // dialog normalization code
1484 if (drc_code_present && embedded_stereo)
1485 get_bits(&s->gb, 8); // drc stereo code
1487 if (s->mix_metadata && get_bits1(&s->gb)) {
1488 skip_bits(&s->gb, 1); // external mix
1489 skip_bits(&s->gb, 6); // post mix gain code
1491 if (get_bits(&s->gb, 2) != 3) // mixer drc code
1492 skip_bits(&s->gb, 3); // drc limit
1494 skip_bits(&s->gb, 8); // custom drc code
1496 if (get_bits1(&s->gb)) // channel specific scaling
1497 for (i = 0; i < s->num_mix_configs; i++)
1498 skip_bits_long(&s->gb, s->mix_config_num_ch[i] * 6); // scale codes
1500 skip_bits_long(&s->gb, s->num_mix_configs * 6); // scale codes
1502 for (i = 0; i < s->num_mix_configs; i++) {
1503 if (get_bits_left(&s->gb) < 0)
1505 dca_exss_skip_mix_coeffs(&s->gb, channels, s->mix_config_num_ch[i]);
1507 dca_exss_skip_mix_coeffs(&s->gb, 6, s->mix_config_num_ch[i]);
1508 if (embedded_stereo)
1509 dca_exss_skip_mix_coeffs(&s->gb, 2, s->mix_config_num_ch[i]);
1513 switch (get_bits(&s->gb, 2)) {
1514 case 0: extensions_mask = get_bits(&s->gb, 12); break;
1515 case 1: extensions_mask = DCA_EXT_EXSS_XLL; break;
1516 case 2: extensions_mask = DCA_EXT_EXSS_LBR; break;
1517 case 3: extensions_mask = 0; /* aux coding */ break;
1520 /* not parsed further, we were only interested in the extensions mask */
1522 if (get_bits_left(&s->gb) < 0)
1525 if (get_bits_count(&s->gb) - header_pos > header_size * 8) {
1526 av_log(s->avctx, AV_LOG_WARNING, "Asset header size mismatch.\n");
1529 skip_bits_long(&s->gb, header_pos + header_size * 8 - get_bits_count(&s->gb));
1531 if (extensions_mask & DCA_EXT_EXSS_XLL)
1532 s->profile = FF_PROFILE_DTS_HD_MA;
1533 else if (extensions_mask & (DCA_EXT_EXSS_XBR | DCA_EXT_EXSS_X96 |
1535 s->profile = FF_PROFILE_DTS_HD_HRA;
1537 if (!(extensions_mask & DCA_EXT_CORE))
1538 av_log(s->avctx, AV_LOG_WARNING, "DTS core detection mismatch.\n");
1539 if ((extensions_mask & DCA_CORE_EXTS) != s->core_ext_mask)
1540 av_log(s->avctx, AV_LOG_WARNING,
1541 "DTS extensions detection mismatch (%d, %d)\n",
1542 extensions_mask & DCA_CORE_EXTS, s->core_ext_mask);
1548 * Parse extension substream header (HD)
1550 static void dca_exss_parse_header(DCAContext *s)
1556 int active_ss_mask[8];
1559 if (get_bits_left(&s->gb) < 52)
1562 skip_bits(&s->gb, 8); // user data
1563 ss_index = get_bits(&s->gb, 2);
1565 blownup = get_bits1(&s->gb);
1566 skip_bits(&s->gb, 8 + 4 * blownup); // header_size
1567 skip_bits(&s->gb, 16 + 4 * blownup); // hd_size
1569 s->static_fields = get_bits1(&s->gb);
1570 if (s->static_fields) {
1571 skip_bits(&s->gb, 2); // reference clock code
1572 skip_bits(&s->gb, 3); // frame duration code
1574 if (get_bits1(&s->gb))
1575 skip_bits_long(&s->gb, 36); // timestamp
1577 /* a single stream can contain multiple audio assets that can be
1578 * combined to form multiple audio presentations */
1580 num_audiop = get_bits(&s->gb, 3) + 1;
1581 if (num_audiop > 1) {
1582 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio presentations.");
1583 /* ignore such streams for now */
1587 num_assets = get_bits(&s->gb, 3) + 1;
1588 if (num_assets > 1) {
1589 av_log_ask_for_sample(s->avctx, "Multiple DTS-HD audio assets.");
1590 /* ignore such streams for now */
1594 for (i = 0; i < num_audiop; i++)
1595 active_ss_mask[i] = get_bits(&s->gb, ss_index + 1);
1597 for (i = 0; i < num_audiop; i++)
1598 for (j = 0; j <= ss_index; j++)
1599 if (active_ss_mask[i] & (1 << j))
1600 skip_bits(&s->gb, 8); // active asset mask
1602 s->mix_metadata = get_bits1(&s->gb);
1603 if (s->mix_metadata) {
1604 int mix_out_mask_size;
1606 skip_bits(&s->gb, 2); // adjustment level
1607 mix_out_mask_size = (get_bits(&s->gb, 2) + 1) << 2;
1608 s->num_mix_configs = get_bits(&s->gb, 2) + 1;
1610 for (i = 0; i < s->num_mix_configs; i++) {
1611 int mix_out_mask = get_bits(&s->gb, mix_out_mask_size);
1612 s->mix_config_num_ch[i] = dca_exss_mask2count(mix_out_mask);
1617 for (i = 0; i < num_assets; i++)
1618 skip_bits_long(&s->gb, 16 + 4 * blownup); // asset size
1620 for (i = 0; i < num_assets; i++) {
1621 if (dca_exss_parse_asset_header(s))
1625 /* not parsed further, we were only interested in the extensions mask
1626 * from the asset header */
1630 * Main frame decoding function
1631 * FIXME add arguments
1633 static int dca_decode_frame(AVCodecContext *avctx, void *data,
1634 int *got_frame_ptr, AVPacket *avpkt)
1636 const uint8_t *buf = avpkt->data;
1637 int buf_size = avpkt->size;
1640 int num_core_channels = 0;
1643 int16_t *samples_s16;
1644 DCAContext *s = avctx->priv_data;
1651 s->dca_buffer_size = ff_dca_convert_bitstream(buf, buf_size, s->dca_buffer,
1652 DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE);
1653 if (s->dca_buffer_size == AVERROR_INVALIDDATA) {
1654 av_log(avctx, AV_LOG_ERROR, "Not a valid DCA frame\n");
1655 return AVERROR_INVALIDDATA;
1658 init_get_bits(&s->gb, s->dca_buffer, s->dca_buffer_size * 8);
1659 if ((ret = dca_parse_frame_header(s)) < 0) {
1660 //seems like the frame is corrupt, try with the next one
1663 //set AVCodec values with parsed data
1664 avctx->sample_rate = s->sample_rate;
1665 avctx->bit_rate = s->bit_rate;
1667 s->profile = FF_PROFILE_DTS;
1669 for (i = 0; i < (s->sample_blocks / 8); i++) {
1670 if ((ret = dca_decode_block(s, 0, i))) {
1671 av_log(avctx, AV_LOG_ERROR, "error decoding block\n");
1676 /* record number of core channels incase less than max channels are requested */
1677 num_core_channels = s->prim_channels;
1680 s->core_ext_mask = dca_ext_audio_descr_mask[s->ext_descr];
1682 s->core_ext_mask = 0;
1684 core_ss_end = FFMIN(s->frame_size, s->dca_buffer_size) * 8;
1686 /* only scan for extensions if ext_descr was unknown or indicated a
1687 * supported XCh extension */
1688 if (s->core_ext_mask < 0 || s->core_ext_mask & DCA_EXT_XCH) {
1690 /* if ext_descr was unknown, clear s->core_ext_mask so that the
1691 * extensions scan can fill it up */
1692 s->core_ext_mask = FFMAX(s->core_ext_mask, 0);
1694 /* extensions start at 32-bit boundaries into bitstream */
1695 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1697 while (core_ss_end - get_bits_count(&s->gb) >= 32) {
1698 uint32_t bits = get_bits_long(&s->gb, 32);
1702 int ext_amode, xch_fsize;
1704 s->xch_base_channel = s->prim_channels;
1706 /* validate sync word using XCHFSIZE field */
1707 xch_fsize = show_bits(&s->gb, 10);
1708 if ((s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize) &&
1709 (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + xch_fsize + 1))
1712 /* skip length-to-end-of-frame field for the moment */
1713 skip_bits(&s->gb, 10);
1715 s->core_ext_mask |= DCA_EXT_XCH;
1717 /* extension amode(number of channels in extension) should be 1 */
1718 /* AFAIK XCh is not used for more channels */
1719 if ((ext_amode = get_bits(&s->gb, 4)) != 1) {
1720 av_log(avctx, AV_LOG_ERROR, "XCh extension amode %d not"
1721 " supported!\n", ext_amode);
1725 /* much like core primary audio coding header */
1726 dca_parse_audio_coding_header(s, s->xch_base_channel);
1728 for (i = 0; i < (s->sample_blocks / 8); i++)
1729 if ((ret = dca_decode_block(s, s->xch_base_channel, i))) {
1730 av_log(avctx, AV_LOG_ERROR, "error decoding XCh extension\n");
1738 /* XXCh: extended channels */
1739 /* usually found either in core or HD part in DTS-HD HRA streams,
1740 * but not in DTS-ES which contains XCh extensions instead */
1741 s->core_ext_mask |= DCA_EXT_XXCH;
1745 int fsize96 = show_bits(&s->gb, 12) + 1;
1746 if (s->frame_size != (get_bits_count(&s->gb) >> 3) - 4 + fsize96)
1749 av_log(avctx, AV_LOG_DEBUG, "X96 extension found at %d bits\n",
1750 get_bits_count(&s->gb));
1751 skip_bits(&s->gb, 12);
1752 av_log(avctx, AV_LOG_DEBUG, "FSIZE96 = %d bytes\n", fsize96);
1753 av_log(avctx, AV_LOG_DEBUG, "REVNO = %d\n", get_bits(&s->gb, 4));
1755 s->core_ext_mask |= DCA_EXT_X96;
1760 skip_bits_long(&s->gb, (-get_bits_count(&s->gb)) & 31);
1763 /* no supported extensions, skip the rest of the core substream */
1764 skip_bits_long(&s->gb, core_ss_end - get_bits_count(&s->gb));
1767 if (s->core_ext_mask & DCA_EXT_X96)
1768 s->profile = FF_PROFILE_DTS_96_24;
1769 else if (s->core_ext_mask & (DCA_EXT_XCH | DCA_EXT_XXCH))
1770 s->profile = FF_PROFILE_DTS_ES;
1772 /* check for ExSS (HD part) */
1773 if (s->dca_buffer_size - s->frame_size > 32 &&
1774 get_bits_long(&s->gb, 32) == DCA_HD_MARKER)
1775 dca_exss_parse_header(s);
1777 avctx->profile = s->profile;
1779 channels = s->prim_channels + !!s->lfe;
1781 if (s->amode < 16) {
1782 avctx->channel_layout = dca_core_channel_layout[s->amode];
1784 if (s->xch_present && (!avctx->request_channels ||
1785 avctx->request_channels > num_core_channels + !!s->lfe)) {
1786 avctx->channel_layout |= AV_CH_BACK_CENTER;
1788 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1789 s->channel_order_tab = dca_channel_reorder_lfe_xch[s->amode];
1791 s->channel_order_tab = dca_channel_reorder_nolfe_xch[s->amode];
1794 channels = num_core_channels + !!s->lfe;
1795 s->xch_present = 0; /* disable further xch processing */
1797 avctx->channel_layout |= AV_CH_LOW_FREQUENCY;
1798 s->channel_order_tab = dca_channel_reorder_lfe[s->amode];
1800 s->channel_order_tab = dca_channel_reorder_nolfe[s->amode];
1803 if (channels > !!s->lfe &&
1804 s->channel_order_tab[channels - 1 - !!s->lfe] < 0)
1805 return AVERROR_INVALIDDATA;
1807 if (avctx->request_channels == 2 && s->prim_channels > 2) {
1809 s->output = DCA_STEREO;
1810 avctx->channel_layout = AV_CH_LAYOUT_STEREO;
1813 av_log(avctx, AV_LOG_ERROR, "Non standard configuration %d !\n", s->amode);
1814 return AVERROR_INVALIDDATA;
1818 /* There is nothing that prevents a dts frame to change channel configuration
1819 but Libav doesn't support that so only set the channels if it is previously
1820 unset. Ideally during the first probe for channels the crc should be checked
1821 and only set avctx->channels when the crc is ok. Right now the decoder could
1822 set the channels based on a broken first frame.*/
1823 if (s->is_channels_set == 0) {
1824 s->is_channels_set = 1;
1825 avctx->channels = channels;
1827 if (avctx->channels != channels) {
1828 av_log(avctx, AV_LOG_ERROR, "DCA decoder does not support number of "
1829 "channels changing in stream. Skipping frame.\n");
1830 return AVERROR_PATCHWELCOME;
1833 /* get output buffer */
1834 s->frame.nb_samples = 256 * (s->sample_blocks / 8);
1835 if ((ret = avctx->get_buffer(avctx, &s->frame)) < 0) {
1836 av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
1839 samples_flt = (float *) s->frame.data[0];
1840 samples_s16 = (int16_t *) s->frame.data[0];
1842 /* filter to get final output */
1843 for (i = 0; i < (s->sample_blocks / 8); i++) {
1844 dca_filter_channels(s, i);
1846 /* If this was marked as a DTS-ES stream we need to subtract back- */
1847 /* channel from SL & SR to remove matrixed back-channel signal */
1848 if ((s->source_pcm_res & 1) && s->xch_present) {
1849 float *back_chan = s->samples + s->channel_order_tab[s->xch_base_channel] * 256;
1850 float *lt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 2] * 256;
1851 float *rt_chan = s->samples + s->channel_order_tab[s->xch_base_channel - 1] * 256;
1852 s->dsp.vector_fmac_scalar(lt_chan, back_chan, -M_SQRT1_2, 256);
1853 s->dsp.vector_fmac_scalar(rt_chan, back_chan, -M_SQRT1_2, 256);
1856 if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
1857 s->fmt_conv.float_interleave(samples_flt, s->samples_chanptr, 256,
1859 samples_flt += 256 * channels;
1861 s->fmt_conv.float_to_int16_interleave(samples_s16,
1862 s->samples_chanptr, 256,
1864 samples_s16 += 256 * channels;
1868 /* update lfe history */
1869 lfe_samples = 2 * s->lfe * (s->sample_blocks / 8);
1870 for (i = 0; i < 2 * s->lfe * 4; i++)
1871 s->lfe_data[i] = s->lfe_data[i + lfe_samples];
1874 *(AVFrame *) data = s->frame;
1882 * DCA initialization
1884 * @param avctx pointer to the AVCodecContext
1887 static av_cold int dca_decode_init(AVCodecContext *avctx)
1889 DCAContext *s = avctx->priv_data;
1895 ff_dsputil_init(&s->dsp, avctx);
1896 ff_mdct_init(&s->imdct, 6, 1, 1.0);
1897 ff_synth_filter_init(&s->synth);
1898 ff_dcadsp_init(&s->dcadsp);
1899 ff_fmt_convert_init(&s->fmt_conv, avctx);
1901 for (i = 0; i < DCA_PRIM_CHANNELS_MAX + 1; i++)
1902 s->samples_chanptr[i] = s->samples + i * 256;
1904 if (avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT) {
1905 avctx->sample_fmt = AV_SAMPLE_FMT_FLT;
1906 s->scale_bias = 1.0 / 32768.0;
1908 avctx->sample_fmt = AV_SAMPLE_FMT_S16;
1909 s->scale_bias = 1.0;
1912 /* allow downmixing to stereo */
1913 if (avctx->channels > 0 && avctx->request_channels < avctx->channels &&
1914 avctx->request_channels == 2) {
1915 avctx->channels = avctx->request_channels;
1918 avcodec_get_frame_defaults(&s->frame);
1919 avctx->coded_frame = &s->frame;
1924 static av_cold int dca_decode_end(AVCodecContext *avctx)
1926 DCAContext *s = avctx->priv_data;
1927 ff_mdct_end(&s->imdct);
1931 static const AVProfile profiles[] = {
1932 { FF_PROFILE_DTS, "DTS" },
1933 { FF_PROFILE_DTS_ES, "DTS-ES" },
1934 { FF_PROFILE_DTS_96_24, "DTS 96/24" },
1935 { FF_PROFILE_DTS_HD_HRA, "DTS-HD HRA" },
1936 { FF_PROFILE_DTS_HD_MA, "DTS-HD MA" },
1937 { FF_PROFILE_UNKNOWN },
1940 AVCodec ff_dca_decoder = {
1942 .type = AVMEDIA_TYPE_AUDIO,
1944 .priv_data_size = sizeof(DCAContext),
1945 .init = dca_decode_init,
1946 .decode = dca_decode_frame,
1947 .close = dca_decode_end,
1948 .long_name = NULL_IF_CONFIG_SMALL("DCA (DTS Coherent Acoustics)"),
1949 .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
1950 .sample_fmts = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_FLT,
1952 AV_SAMPLE_FMT_NONE },
1953 .profiles = NULL_IF_CONFIG_SMALL(profiles),