Line | Count | Source |
1 | | /*********************************************************************** |
2 | | Copyright (c) 2006-2011, Skype Limited. All rights reserved. |
3 | | Redistribution and use in source and binary forms, with or without |
4 | | modification, are permitted provided that the following conditions |
5 | | are met: |
6 | | - Redistributions of source code must retain the above copyright notice, |
7 | | this list of conditions and the following disclaimer. |
8 | | - Redistributions in binary form must reproduce the above copyright |
9 | | notice, this list of conditions and the following disclaimer in the |
10 | | documentation and/or other materials provided with the distribution. |
11 | | - Neither the name of Internet Society, IETF or IETF Trust, nor the |
12 | | names of specific contributors, may be used to endorse or promote |
13 | | products derived from this software without specific prior written |
14 | | permission. |
15 | | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" |
16 | | AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
17 | | IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
18 | | ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE |
19 | | LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
20 | | CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
21 | | SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
22 | | INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
23 | | CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
24 | | ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
25 | | POSSIBILITY OF SUCH DAMAGE. |
26 | | ***********************************************************************/ |
27 | | |
28 | | #ifdef HAVE_CONFIG_H |
29 | | #include "config.h" |
30 | | #endif |
31 | | #include "API.h" |
32 | | #include "main.h" |
33 | | #include "stack_alloc.h" |
34 | | #include "os_support.h" |
35 | | |
36 | | #ifdef ENABLE_OSCE |
37 | | #include "osce.h" |
38 | | #include "osce_structs.h" |
39 | | #ifdef ENABLE_OSCE_BWE |
40 | | #include "osce_features.h" |
41 | | #endif |
42 | | #endif |
43 | | |
44 | | /************************/ |
45 | | /* Decoder Super Struct */ |
46 | | /************************/ |
47 | | typedef struct { |
48 | | silk_decoder_state channel_state[ DECODER_NUM_CHANNELS ]; |
49 | | stereo_dec_state sStereo; |
50 | | opus_int nChannelsAPI; |
51 | | opus_int nChannelsInternal; |
52 | | opus_int prev_decode_only_middle; |
53 | | #ifdef ENABLE_OSCE |
54 | | OSCEModel osce_model; |
55 | | #endif |
56 | | } silk_decoder; |
57 | | |
58 | | /*********************/ |
59 | | /* Decoder functions */ |
60 | | /*********************/ |
61 | | |
62 | | |
63 | | |
64 | | opus_int silk_LoadOSCEModels(void *decState, const unsigned char *data, int len) |
65 | 403k | { |
66 | | #ifdef ENABLE_OSCE |
67 | | opus_int ret = SILK_NO_ERROR; |
68 | | |
69 | | ret = osce_load_models(&((silk_decoder *)decState)->osce_model, data, len); |
70 | | ((silk_decoder *)decState)->osce_model.loaded = (ret == 0); |
71 | | return ret; |
72 | | #else |
73 | 403k | (void) decState; |
74 | 403k | (void) data; |
75 | 403k | (void) len; |
76 | 403k | return SILK_NO_ERROR; |
77 | 403k | #endif |
78 | 403k | } |
79 | | |
80 | | opus_int silk_Get_Decoder_Size( /* O Returns error code */ |
81 | | opus_int *decSizeBytes /* O Number of bytes in SILK decoder state */ |
82 | | ) |
83 | 1.22M | { |
84 | 1.22M | opus_int ret = SILK_NO_ERROR; |
85 | | |
86 | 1.22M | *decSizeBytes = sizeof( silk_decoder ); |
87 | | |
88 | 1.22M | return ret; |
89 | 1.22M | } |
90 | | |
91 | | /* Reset decoder state */ |
92 | | opus_int silk_ResetDecoder( /* O Returns error code */ |
93 | | void *decState /* I/O State */ |
94 | | ) |
95 | 408 | { |
96 | 408 | opus_int n, ret = SILK_NO_ERROR; |
97 | 408 | silk_decoder_state *channel_state = ((silk_decoder *)decState)->channel_state; |
98 | | |
99 | 1.22k | for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { |
100 | 816 | ret = silk_reset_decoder( &channel_state[ n ] ); |
101 | 816 | } |
102 | 408 | silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); |
103 | | /* Not strictly needed, but it's cleaner that way */ |
104 | 408 | ((silk_decoder *)decState)->prev_decode_only_middle = 0; |
105 | | |
106 | 408 | return ret; |
107 | 408 | } |
108 | | |
109 | | |
110 | | opus_int silk_InitDecoder( /* O Returns error code */ |
111 | | void *decState /* I/O State */ |
112 | | ) |
113 | 403k | { |
114 | 403k | opus_int n, ret = SILK_NO_ERROR; |
115 | 403k | silk_decoder_state *channel_state = ((silk_decoder *)decState)->channel_state; |
116 | | #ifdef ENABLE_OSCE |
117 | | ((silk_decoder *)decState)->osce_model.loaded = 0; |
118 | | #endif |
119 | 403k | #ifndef USE_WEIGHTS_FILE |
120 | | /* load osce models */ |
121 | 403k | silk_LoadOSCEModels(decState, NULL, 0); |
122 | 403k | #endif |
123 | | |
124 | 1.20M | for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { |
125 | 806k | ret = silk_init_decoder( &channel_state[ n ] ); |
126 | 806k | } |
127 | 403k | silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); |
128 | | /* Not strictly needed, but it's cleaner that way */ |
129 | 403k | ((silk_decoder *)decState)->prev_decode_only_middle = 0; |
130 | | |
131 | 403k | return ret; |
132 | 403k | } |
133 | | |
134 | | /* Decode a frame */ |
135 | | opus_int silk_Decode( /* O Returns error code */ |
136 | | void* decState, /* I/O State */ |
137 | | silk_DecControlStruct* decControl, /* I/O Control Structure */ |
138 | | opus_int lostFlag, /* I 0: no loss, 1 loss, 2 decode fec */ |
139 | | opus_int newPacketFlag, /* I Indicates first decoder call for this packet */ |
140 | | ec_dec *psRangeDec, /* I/O Compressor data structure */ |
141 | | opus_res *samplesOut, /* O Decoded output speech vector */ |
142 | | opus_int32 *nSamplesOut, /* O Number of samples decoded */ |
143 | | #ifdef ENABLE_DEEP_PLC |
144 | | LPCNetPLCState *lpcnet, |
145 | | #endif |
146 | | int arch /* I Run-time architecture */ |
147 | | ) |
148 | 796k | { |
149 | 796k | opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR; |
150 | 796k | opus_int32 nSamplesOutDec, LBRR_symbol; |
151 | 796k | opus_int16 *samplesOut1_tmp[ 2 ]; |
152 | 796k | VARDECL( opus_int16, samplesOut1_tmp_storage1 ); |
153 | 796k | VARDECL( opus_int16, samplesOut2_tmp ); |
154 | 796k | opus_int32 MS_pred_Q13[ 2 ] = { 0 }; |
155 | 796k | opus_int16 *resample_out_ptr; |
156 | 796k | silk_decoder *psDec = ( silk_decoder * )decState; |
157 | 796k | silk_decoder_state *channel_state = psDec->channel_state; |
158 | 796k | opus_int has_side; |
159 | 796k | opus_int stereo_to_mono; |
160 | | #ifdef ENABLE_OSCE_BWE |
161 | | ALLOC(resamp_buffer, 3 * MAX_FRAME_LENGTH, opus_int16); |
162 | | #endif |
163 | 796k | SAVE_STACK; |
164 | | |
165 | 796k | celt_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 ); |
166 | | |
167 | | /**********************************/ |
168 | | /* Test if first frame in payload */ |
169 | | /**********************************/ |
170 | 796k | if( newPacketFlag ) { |
171 | 1.37M | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
172 | 856k | channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */ |
173 | 856k | } |
174 | 519k | } |
175 | | |
176 | | /* If Mono -> Stereo transition in bitstream: init state of second channel */ |
177 | 796k | if( decControl->nChannelsInternal > psDec->nChannelsInternal ) { |
178 | 270k | ret += silk_init_decoder( &channel_state[ 1 ] ); |
179 | 270k | } |
180 | | |
181 | 796k | stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 && |
182 | 0 | ( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz ); |
183 | | |
184 | 796k | if( channel_state[ 0 ].nFramesDecoded == 0 ) { |
185 | 1.37M | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
186 | 856k | opus_int fs_kHz_dec; |
187 | 856k | if( decControl->payloadSize_ms == 0 ) { |
188 | | /* Assuming packet loss, use 10 ms */ |
189 | 0 | channel_state[ n ].nFramesPerPacket = 1; |
190 | 0 | channel_state[ n ].nb_subfr = 2; |
191 | 856k | } else if( decControl->payloadSize_ms == 10 ) { |
192 | 419k | channel_state[ n ].nFramesPerPacket = 1; |
193 | 419k | channel_state[ n ].nb_subfr = 2; |
194 | 437k | } else if( decControl->payloadSize_ms == 20 ) { |
195 | 176k | channel_state[ n ].nFramesPerPacket = 1; |
196 | 176k | channel_state[ n ].nb_subfr = 4; |
197 | 260k | } else if( decControl->payloadSize_ms == 40 ) { |
198 | 32.4k | channel_state[ n ].nFramesPerPacket = 2; |
199 | 32.4k | channel_state[ n ].nb_subfr = 4; |
200 | 228k | } else if( decControl->payloadSize_ms == 60 ) { |
201 | 228k | channel_state[ n ].nFramesPerPacket = 3; |
202 | 228k | channel_state[ n ].nb_subfr = 4; |
203 | 228k | } else { |
204 | 0 | celt_assert( 0 ); |
205 | 0 | RESTORE_STACK; |
206 | 0 | return SILK_DEC_INVALID_FRAME_SIZE; |
207 | 0 | } |
208 | 856k | fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1; |
209 | 856k | if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) { |
210 | 0 | celt_assert( 0 ); |
211 | 0 | RESTORE_STACK; |
212 | 0 | return SILK_DEC_INVALID_SAMPLING_FREQUENCY; |
213 | 0 | } |
214 | 856k | ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate ); |
215 | 856k | } |
216 | 519k | } |
217 | | |
218 | 796k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) { |
219 | 0 | silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) ); |
220 | 0 | silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) ); |
221 | 0 | silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) ); |
222 | 0 | } |
223 | 796k | psDec->nChannelsAPI = decControl->nChannelsAPI; |
224 | 796k | psDec->nChannelsInternal = decControl->nChannelsInternal; |
225 | | |
226 | 796k | if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 || decControl->API_sampleRate < 8000 ) { |
227 | 0 | ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY; |
228 | 0 | RESTORE_STACK; |
229 | 0 | return( ret ); |
230 | 0 | } |
231 | | |
232 | 796k | if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) { |
233 | | /* First decoder call for this payload */ |
234 | | /* Decode VAD flags and LBRR flag */ |
235 | 989k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
236 | 1.70M | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { |
237 | 1.09M | channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1); |
238 | 1.09M | } |
239 | 607k | channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1); |
240 | 607k | } |
241 | | /* Decode LBRR flags */ |
242 | 989k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
243 | 607k | silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) ); |
244 | 607k | if( channel_state[ n ].LBRR_flag ) { |
245 | 262k | if( channel_state[ n ].nFramesPerPacket == 1 ) { |
246 | 99.5k | channel_state[ n ].LBRR_flags[ 0 ] = 1; |
247 | 163k | } else { |
248 | 163k | LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1; |
249 | 636k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { |
250 | 472k | channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1; |
251 | 472k | } |
252 | 163k | } |
253 | 262k | } |
254 | 607k | } |
255 | | |
256 | 381k | if( lostFlag == FLAG_DECODE_NORMAL ) { |
257 | | /* Regular decoding: skip all LBRR data */ |
258 | 807k | for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) { |
259 | 1.33M | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
260 | 835k | if( channel_state[ n ].LBRR_flags[ i ] ) { |
261 | 288k | opus_int16 pulses[ MAX_FRAME_LENGTH ]; |
262 | 288k | opus_int condCoding; |
263 | | |
264 | 288k | if( decControl->nChannelsInternal == 2 && n == 0 ) { |
265 | 124k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); |
266 | 124k | if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) { |
267 | 59.1k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); |
268 | 59.1k | } |
269 | 124k | } |
270 | | /* Use conditional coding if previous frame available */ |
271 | 288k | if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) { |
272 | 93.5k | condCoding = CODE_CONDITIONALLY; |
273 | 194k | } else { |
274 | 194k | condCoding = CODE_INDEPENDENTLY; |
275 | 194k | } |
276 | 288k | silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding ); |
277 | 288k | silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType, |
278 | 288k | channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length ); |
279 | 288k | } |
280 | 835k | } |
281 | 499k | } |
282 | 308k | } |
283 | 381k | } |
284 | | |
285 | | /* Get MS predictor index */ |
286 | 796k | if( decControl->nChannelsInternal == 2 ) { |
287 | 549k | if( lostFlag == FLAG_DECODE_NORMAL || |
288 | 213k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) ) |
289 | 378k | { |
290 | 378k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); |
291 | | /* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */ |
292 | 378k | if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) || |
293 | 214k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) ) |
294 | 189k | { |
295 | 189k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); |
296 | 189k | } else { |
297 | 189k | decode_only_middle = 0; |
298 | 189k | } |
299 | 378k | } else { |
300 | 512k | for( n = 0; n < 2; n++ ) { |
301 | 341k | MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ]; |
302 | 341k | } |
303 | 170k | } |
304 | 549k | } |
305 | | |
306 | | /* Reset side channel decoder prediction memory for first frame with side coding */ |
307 | 796k | if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) { |
308 | 40.2k | silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) ); |
309 | 40.2k | silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) ); |
310 | 40.2k | psDec->channel_state[ 1 ].lagPrev = 100; |
311 | 40.2k | psDec->channel_state[ 1 ].LastGainIndex = 10; |
312 | 40.2k | psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY; |
313 | 40.2k | psDec->channel_state[ 1 ].first_frame_after_reset = 1; |
314 | 40.2k | } |
315 | | |
316 | | /* Check if the temp buffer fits into the output PCM buffer. If it fits, |
317 | | we can delay allocating the temp buffer until after the SILK peak stack |
318 | | usage. We need to use a < and not a <= because of the two extra samples. */ |
319 | 796k | ALLOC( samplesOut1_tmp_storage1, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ), |
320 | 796k | opus_int16 ); |
321 | 796k | samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1; |
322 | 796k | samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2; |
323 | | |
324 | 796k | if( lostFlag == FLAG_DECODE_NORMAL ) { |
325 | 499k | has_side = !decode_only_middle; |
326 | 499k | } else { |
327 | 297k | has_side = !psDec->prev_decode_only_middle |
328 | 23.4k | || (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 ); |
329 | 297k | } |
330 | 796k | channel_state[ 0 ].sPLC.enable_deep_plc = decControl->enable_deep_plc; |
331 | | /* Call decoder for one frame */ |
332 | 2.14M | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
333 | 1.34M | if( n == 0 || has_side ) { |
334 | 1.29M | opus_int FrameIndex; |
335 | 1.29M | opus_int condCoding; |
336 | | |
337 | 1.29M | FrameIndex = channel_state[ 0 ].nFramesDecoded - n; |
338 | | /* Use independent coding if no previous frame available */ |
339 | 1.29M | if( FrameIndex <= 0 ) { |
340 | 829k | condCoding = CODE_INDEPENDENTLY; |
341 | 829k | } else if( lostFlag == FLAG_DECODE_LBRR ) { |
342 | 129k | condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY; |
343 | 334k | } else if( n > 0 && psDec->prev_decode_only_middle ) { |
344 | | /* If we skipped a side frame in this packet, we don't |
345 | | need LTP scaling; the LTP state is well-defined. */ |
346 | 11.0k | condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; |
347 | 323k | } else { |
348 | 323k | condCoding = CODE_CONDITIONALLY; |
349 | 323k | } |
350 | | #ifdef ENABLE_OSCE |
351 | | if ( channel_state[n].osce.method != decControl->osce_method ) { |
352 | | osce_reset( &channel_state[n].osce, decControl->osce_method ); |
353 | | } |
354 | | #endif |
355 | 1.29M | ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding, |
356 | | #ifdef ENABLE_DEEP_PLC |
357 | | n == 0 ? lpcnet : NULL, |
358 | | #endif |
359 | | #ifdef ENABLE_OSCE |
360 | | &psDec->osce_model, |
361 | | #endif |
362 | 1.29M | arch); |
363 | 1.29M | } else { |
364 | 52.6k | silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) ); |
365 | 52.6k | } |
366 | 1.34M | channel_state[ n ].nFramesDecoded++; |
367 | 1.34M | } |
368 | | |
369 | 796k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) { |
370 | | /* Convert Mid/Side to Left/Right */ |
371 | 205k | silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec ); |
372 | 591k | } else { |
373 | | /* Buffering */ |
374 | 591k | silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) ); |
375 | 591k | silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) ); |
376 | 591k | } |
377 | | |
378 | | /* Number of output samples */ |
379 | 796k | *nSamplesOut = silk_DIV32( nSamplesOutDec * decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) ); |
380 | | |
381 | | /* Set up pointers to temp buffers */ |
382 | 796k | ALLOC( samplesOut2_tmp, *nSamplesOut, opus_int16 ); |
383 | 796k | resample_out_ptr = samplesOut2_tmp; |
384 | | |
385 | 1.79M | for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) { |
386 | | |
387 | | #ifdef ENABLE_OSCE_BWE |
388 | | /* Resample or extend decoded signal to API_sampleRate */ |
389 | | if (decControl->osce_extended_mode == OSCE_MODE_SILK_BBWE) { |
390 | | silk_assert(decControl->API_sampleRate == 48000); |
391 | | |
392 | | if (decControl->prev_osce_extended_mode != OSCE_MODE_SILK_BBWE) { |
393 | | /* Reset the BWE state */ |
394 | | osce_bwe_reset( &channel_state[ n ].osce_bwe ); |
395 | | } |
396 | | |
397 | | osce_bwe(&psDec->osce_model, &channel_state[ n ].osce_bwe, |
398 | | resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec, arch); |
399 | | |
400 | | if (decControl->prev_osce_extended_mode == OSCE_MODE_SILK_ONLY || |
401 | | decControl->prev_osce_extended_mode == OSCE_MODE_HYBRID) { |
402 | | /* cross-fade with upsampled signal */ |
403 | | silk_resampler( &channel_state[ n ].resampler_state, resamp_buffer, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); |
404 | | osce_bwe_cross_fade_10ms(resample_out_ptr, resamp_buffer, 480); |
405 | | } |
406 | | } else { |
407 | | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); |
408 | | if (decControl->prev_osce_extended_mode == OSCE_MODE_SILK_BBWE) { |
409 | | osce_bwe(&psDec->osce_model, &channel_state[ n ].osce_bwe, |
410 | | resamp_buffer, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec, arch); |
411 | | /* cross-fade with upsampled signal */ |
412 | | osce_bwe_cross_fade_10ms(resample_out_ptr, resamp_buffer, 480); |
413 | | } |
414 | | } |
415 | | #else |
416 | | /* Resample decoded signal to API_sampleRate */ |
417 | 1.00M | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); |
418 | 1.00M | #endif |
419 | | /* Interleave if stereo output and stereo stream */ |
420 | 1.00M | if( decControl->nChannelsAPI == 2 ) { |
421 | 217M | for( i = 0; i < *nSamplesOut; i++ ) { |
422 | 217M | samplesOut[ n + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); |
423 | 217M | } |
424 | 527k | } else { |
425 | 184M | for( i = 0; i < *nSamplesOut; i++ ) { |
426 | 184M | samplesOut[ i ] = INT16TORES(resample_out_ptr[ i ]); |
427 | 184M | } |
428 | 475k | } |
429 | 1.00M | } |
430 | | |
431 | | #ifdef ENABLE_OSCE_BWE |
432 | | decControl->prev_osce_extended_mode = decControl->osce_extended_mode; |
433 | | #endif |
434 | | |
435 | | /* Create two channel output from mono stream */ |
436 | 796k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) { |
437 | 115k | if ( stereo_to_mono ){ |
438 | | /* Resample right channel for newly collapsed stereo just in case |
439 | | we weren't doing collapsing when switching to mono */ |
440 | 0 | ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec ); |
441 | |
|
442 | 0 | for( i = 0; i < *nSamplesOut; i++ ) { |
443 | 0 | samplesOut[ 1 + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); |
444 | 0 | } |
445 | 115k | } else { |
446 | 48.4M | for( i = 0; i < *nSamplesOut; i++ ) { |
447 | 48.3M | samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ]; |
448 | 48.3M | } |
449 | 115k | } |
450 | 115k | } |
451 | | |
452 | | /* Export pitch lag, measured at 48 kHz sampling rate */ |
453 | 796k | if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) { |
454 | 178k | int mult_tab[ 3 ] = { 6, 4, 3 }; |
455 | 178k | decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ]; |
456 | 618k | } else { |
457 | 618k | decControl->prevPitchLag = 0; |
458 | 618k | } |
459 | | |
460 | 796k | if( lostFlag == FLAG_PACKET_LOST ) { |
461 | | /* On packet loss, remove the gain clamping to prevent having the energy "bounce back" |
462 | | if we lose packets when the energy is going down */ |
463 | 387k | for ( i = 0; i < psDec->nChannelsInternal; i++ ) |
464 | 249k | psDec->channel_state[ i ].LastGainIndex = 10; |
465 | 658k | } else { |
466 | 658k | psDec->prev_decode_only_middle = decode_only_middle; |
467 | 658k | } |
468 | 796k | RESTORE_STACK; |
469 | 796k | return ret; |
470 | 796k | } Line | Count | Source | 148 | 398k | { | 149 | 398k | opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR; | 150 | 398k | opus_int32 nSamplesOutDec, LBRR_symbol; | 151 | 398k | opus_int16 *samplesOut1_tmp[ 2 ]; | 152 | 398k | VARDECL( opus_int16, samplesOut1_tmp_storage1 ); | 153 | 398k | VARDECL( opus_int16, samplesOut2_tmp ); | 154 | 398k | opus_int32 MS_pred_Q13[ 2 ] = { 0 }; | 155 | 398k | opus_int16 *resample_out_ptr; | 156 | 398k | silk_decoder *psDec = ( silk_decoder * )decState; | 157 | 398k | silk_decoder_state *channel_state = psDec->channel_state; | 158 | 398k | opus_int has_side; | 159 | 398k | opus_int stereo_to_mono; | 160 | | #ifdef ENABLE_OSCE_BWE | 161 | | ALLOC(resamp_buffer, 3 * MAX_FRAME_LENGTH, opus_int16); | 162 | | #endif | 163 | 398k | SAVE_STACK; | 164 | | | 165 | 398k | celt_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 ); | 166 | | | 167 | | /**********************************/ | 168 | | /* Test if first frame in payload */ | 169 | | /**********************************/ | 170 | 398k | if( newPacketFlag ) { | 171 | 688k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 172 | 428k | channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */ | 173 | 428k | } | 174 | 259k | } | 175 | | | 176 | | /* If Mono -> Stereo transition in bitstream: init state of second channel */ | 177 | 398k | if( decControl->nChannelsInternal > psDec->nChannelsInternal ) { | 178 | 135k | ret += silk_init_decoder( &channel_state[ 1 ] ); | 179 | 135k | } | 180 | | | 181 | 398k | stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 && | 182 | 0 | ( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz ); | 183 | | | 184 | 398k | if( channel_state[ 0 ].nFramesDecoded == 0 ) { | 185 | 688k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 186 | 428k | opus_int fs_kHz_dec; | 187 | 428k | if( decControl->payloadSize_ms == 0 ) { | 188 | | /* Assuming packet loss, use 10 ms */ | 189 | 0 | channel_state[ n ].nFramesPerPacket = 1; | 190 | 0 | channel_state[ n ].nb_subfr = 2; | 191 | 428k | } else if( decControl->payloadSize_ms == 10 ) { | 192 | 209k | channel_state[ n ].nFramesPerPacket = 1; | 193 | 209k | channel_state[ n ].nb_subfr = 2; | 194 | 218k | } else if( decControl->payloadSize_ms == 20 ) { | 195 | 88.1k | channel_state[ n ].nFramesPerPacket = 1; | 196 | 88.1k | channel_state[ n ].nb_subfr = 4; | 197 | 130k | } else if( decControl->payloadSize_ms == 40 ) { | 198 | 16.2k | channel_state[ n ].nFramesPerPacket = 2; | 199 | 16.2k | channel_state[ n ].nb_subfr = 4; | 200 | 114k | } else if( decControl->payloadSize_ms == 60 ) { | 201 | 114k | channel_state[ n ].nFramesPerPacket = 3; | 202 | 114k | channel_state[ n ].nb_subfr = 4; | 203 | 114k | } else { | 204 | 0 | celt_assert( 0 ); | 205 | 0 | RESTORE_STACK; | 206 | 0 | return SILK_DEC_INVALID_FRAME_SIZE; | 207 | 0 | } | 208 | 428k | fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1; | 209 | 428k | if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) { | 210 | 0 | celt_assert( 0 ); | 211 | 0 | RESTORE_STACK; | 212 | 0 | return SILK_DEC_INVALID_SAMPLING_FREQUENCY; | 213 | 0 | } | 214 | 428k | ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate ); | 215 | 428k | } | 216 | 259k | } | 217 | | | 218 | 398k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) { | 219 | 0 | silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) ); | 220 | 0 | silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) ); | 221 | 0 | silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) ); | 222 | 0 | } | 223 | 398k | psDec->nChannelsAPI = decControl->nChannelsAPI; | 224 | 398k | psDec->nChannelsInternal = decControl->nChannelsInternal; | 225 | | | 226 | 398k | if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 || decControl->API_sampleRate < 8000 ) { | 227 | 0 | ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY; | 228 | 0 | RESTORE_STACK; | 229 | 0 | return( ret ); | 230 | 0 | } | 231 | | | 232 | 398k | if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) { | 233 | | /* First decoder call for this payload */ | 234 | | /* Decode VAD flags and LBRR flag */ | 235 | 494k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 236 | 851k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { | 237 | 548k | channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1); | 238 | 548k | } | 239 | 303k | channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1); | 240 | 303k | } | 241 | | /* Decode LBRR flags */ | 242 | 494k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 243 | 303k | silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) ); | 244 | 303k | if( channel_state[ n ].LBRR_flag ) { | 245 | 131k | if( channel_state[ n ].nFramesPerPacket == 1 ) { | 246 | 49.7k | channel_state[ n ].LBRR_flags[ 0 ] = 1; | 247 | 81.7k | } else { | 248 | 81.7k | LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1; | 249 | 318k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { | 250 | 236k | channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1; | 251 | 236k | } | 252 | 81.7k | } | 253 | 131k | } | 254 | 303k | } | 255 | | | 256 | 190k | if( lostFlag == FLAG_DECODE_NORMAL ) { | 257 | | /* Regular decoding: skip all LBRR data */ | 258 | 403k | for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) { | 259 | 667k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 260 | 417k | if( channel_state[ n ].LBRR_flags[ i ] ) { | 261 | 144k | opus_int16 pulses[ MAX_FRAME_LENGTH ]; | 262 | 144k | opus_int condCoding; | 263 | | | 264 | 144k | if( decControl->nChannelsInternal == 2 && n == 0 ) { | 265 | 62.4k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); | 266 | 62.4k | if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) { | 267 | 29.5k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); | 268 | 29.5k | } | 269 | 62.4k | } | 270 | | /* Use conditional coding if previous frame available */ | 271 | 144k | if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) { | 272 | 46.7k | condCoding = CODE_CONDITIONALLY; | 273 | 97.4k | } else { | 274 | 97.4k | condCoding = CODE_INDEPENDENTLY; | 275 | 97.4k | } | 276 | 144k | silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding ); | 277 | 144k | silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType, | 278 | 144k | channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length ); | 279 | 144k | } | 280 | 417k | } | 281 | 249k | } | 282 | 154k | } | 283 | 190k | } | 284 | | | 285 | | /* Get MS predictor index */ | 286 | 398k | if( decControl->nChannelsInternal == 2 ) { | 287 | 274k | if( lostFlag == FLAG_DECODE_NORMAL || | 288 | 106k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) ) | 289 | 189k | { | 290 | 189k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); | 291 | | /* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */ | 292 | 189k | if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) || | 293 | 107k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) ) | 294 | 94.6k | { | 295 | 94.6k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); | 296 | 94.6k | } else { | 297 | 94.5k | decode_only_middle = 0; | 298 | 94.5k | } | 299 | 189k | } else { | 300 | 256k | for( n = 0; n < 2; n++ ) { | 301 | 170k | MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ]; | 302 | 170k | } | 303 | 85.4k | } | 304 | 274k | } | 305 | | | 306 | | /* Reset side channel decoder prediction memory for first frame with side coding */ | 307 | 398k | if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) { | 308 | 20.1k | silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) ); | 309 | 20.1k | silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) ); | 310 | 20.1k | psDec->channel_state[ 1 ].lagPrev = 100; | 311 | 20.1k | psDec->channel_state[ 1 ].LastGainIndex = 10; | 312 | 20.1k | psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY; | 313 | 20.1k | psDec->channel_state[ 1 ].first_frame_after_reset = 1; | 314 | 20.1k | } | 315 | | | 316 | | /* Check if the temp buffer fits into the output PCM buffer. If it fits, | 317 | | we can delay allocating the temp buffer until after the SILK peak stack | 318 | | usage. We need to use a < and not a <= because of the two extra samples. */ | 319 | 398k | ALLOC( samplesOut1_tmp_storage1, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ), | 320 | 398k | opus_int16 ); | 321 | 398k | samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1; | 322 | 398k | samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2; | 323 | | | 324 | 398k | if( lostFlag == FLAG_DECODE_NORMAL ) { | 325 | 249k | has_side = !decode_only_middle; | 326 | 249k | } else { | 327 | 148k | has_side = !psDec->prev_decode_only_middle | 328 | 11.7k | || (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 ); | 329 | 148k | } | 330 | 398k | channel_state[ 0 ].sPLC.enable_deep_plc = decControl->enable_deep_plc; | 331 | | /* Call decoder for one frame */ | 332 | 1.07M | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 333 | 673k | if( n == 0 || has_side ) { | 334 | 646k | opus_int FrameIndex; | 335 | 646k | opus_int condCoding; | 336 | | | 337 | 646k | FrameIndex = channel_state[ 0 ].nFramesDecoded - n; | 338 | | /* Use independent coding if no previous frame available */ | 339 | 646k | if( FrameIndex <= 0 ) { | 340 | 414k | condCoding = CODE_INDEPENDENTLY; | 341 | 414k | } else if( lostFlag == FLAG_DECODE_LBRR ) { | 342 | 64.5k | condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY; | 343 | 167k | } else if( n > 0 && psDec->prev_decode_only_middle ) { | 344 | | /* If we skipped a side frame in this packet, we don't | 345 | | need LTP scaling; the LTP state is well-defined. */ | 346 | 5.50k | condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; | 347 | 161k | } else { | 348 | 161k | condCoding = CODE_CONDITIONALLY; | 349 | 161k | } | 350 | | #ifdef ENABLE_OSCE | 351 | | if ( channel_state[n].osce.method != decControl->osce_method ) { | 352 | | osce_reset( &channel_state[n].osce, decControl->osce_method ); | 353 | | } | 354 | | #endif | 355 | 646k | ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding, | 356 | | #ifdef ENABLE_DEEP_PLC | 357 | | n == 0 ? lpcnet : NULL, | 358 | | #endif | 359 | | #ifdef ENABLE_OSCE | 360 | | &psDec->osce_model, | 361 | | #endif | 362 | 646k | arch); | 363 | 646k | } else { | 364 | 26.3k | silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) ); | 365 | 26.3k | } | 366 | 673k | channel_state[ n ].nFramesDecoded++; | 367 | 673k | } | 368 | | | 369 | 398k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) { | 370 | | /* Convert Mid/Side to Left/Right */ | 371 | 102k | silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec ); | 372 | 295k | } else { | 373 | | /* Buffering */ | 374 | 295k | silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) ); | 375 | 295k | silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) ); | 376 | 295k | } | 377 | | | 378 | | /* Number of output samples */ | 379 | 398k | *nSamplesOut = silk_DIV32( nSamplesOutDec * decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) ); | 380 | | | 381 | | /* Set up pointers to temp buffers */ | 382 | 398k | ALLOC( samplesOut2_tmp, *nSamplesOut, opus_int16 ); | 383 | 398k | resample_out_ptr = samplesOut2_tmp; | 384 | | | 385 | 899k | for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) { | 386 | | | 387 | | #ifdef ENABLE_OSCE_BWE | 388 | | /* Resample or extend decoded signal to API_sampleRate */ | 389 | | if (decControl->osce_extended_mode == OSCE_MODE_SILK_BBWE) { | 390 | | silk_assert(decControl->API_sampleRate == 48000); | 391 | | | 392 | | if (decControl->prev_osce_extended_mode != OSCE_MODE_SILK_BBWE) { | 393 | | /* Reset the BWE state */ | 394 | | osce_bwe_reset( &channel_state[ n ].osce_bwe ); | 395 | | } | 396 | | | 397 | | osce_bwe(&psDec->osce_model, &channel_state[ n ].osce_bwe, | 398 | | resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec, arch); | 399 | | | 400 | | if (decControl->prev_osce_extended_mode == OSCE_MODE_SILK_ONLY || | 401 | | decControl->prev_osce_extended_mode == OSCE_MODE_HYBRID) { | 402 | | /* cross-fade with upsampled signal */ | 403 | | silk_resampler( &channel_state[ n ].resampler_state, resamp_buffer, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); | 404 | | osce_bwe_cross_fade_10ms(resample_out_ptr, resamp_buffer, 480); | 405 | | } | 406 | | } else { | 407 | | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); | 408 | | if (decControl->prev_osce_extended_mode == OSCE_MODE_SILK_BBWE) { | 409 | | osce_bwe(&psDec->osce_model, &channel_state[ n ].osce_bwe, | 410 | | resamp_buffer, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec, arch); | 411 | | /* cross-fade with upsampled signal */ | 412 | | osce_bwe_cross_fade_10ms(resample_out_ptr, resamp_buffer, 480); | 413 | | } | 414 | | } | 415 | | #else | 416 | | /* Resample decoded signal to API_sampleRate */ | 417 | 501k | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); | 418 | 501k | #endif | 419 | | /* Interleave if stereo output and stereo stream */ | 420 | 501k | if( decControl->nChannelsAPI == 2 ) { | 421 | 108M | for( i = 0; i < *nSamplesOut; i++ ) { | 422 | 108M | samplesOut[ n + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); | 423 | 108M | } | 424 | 263k | } else { | 425 | 92.2M | for( i = 0; i < *nSamplesOut; i++ ) { | 426 | 92.0M | samplesOut[ i ] = INT16TORES(resample_out_ptr[ i ]); | 427 | 92.0M | } | 428 | 237k | } | 429 | 501k | } | 430 | | | 431 | | #ifdef ENABLE_OSCE_BWE | 432 | | decControl->prev_osce_extended_mode = decControl->osce_extended_mode; | 433 | | #endif | 434 | | | 435 | | /* Create two channel output from mono stream */ | 436 | 398k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) { | 437 | 57.7k | if ( stereo_to_mono ){ | 438 | | /* Resample right channel for newly collapsed stereo just in case | 439 | | we weren't doing collapsing when switching to mono */ | 440 | 0 | ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec ); | 441 | |
| 442 | 0 | for( i = 0; i < *nSamplesOut; i++ ) { | 443 | 0 | samplesOut[ 1 + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); | 444 | 0 | } | 445 | 57.7k | } else { | 446 | 24.2M | for( i = 0; i < *nSamplesOut; i++ ) { | 447 | 24.1M | samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ]; | 448 | 24.1M | } | 449 | 57.7k | } | 450 | 57.7k | } | 451 | | | 452 | | /* Export pitch lag, measured at 48 kHz sampling rate */ | 453 | 398k | if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) { | 454 | 89.2k | int mult_tab[ 3 ] = { 6, 4, 3 }; | 455 | 89.2k | decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ]; | 456 | 309k | } else { | 457 | 309k | decControl->prevPitchLag = 0; | 458 | 309k | } | 459 | | | 460 | 398k | if( lostFlag == FLAG_PACKET_LOST ) { | 461 | | /* On packet loss, remove the gain clamping to prevent having the energy "bounce back" | 462 | | if we lose packets when the energy is going down */ | 463 | 193k | for ( i = 0; i < psDec->nChannelsInternal; i++ ) | 464 | 124k | psDec->channel_state[ i ].LastGainIndex = 10; | 465 | 329k | } else { | 466 | 329k | psDec->prev_decode_only_middle = decode_only_middle; | 467 | 329k | } | 468 | 398k | RESTORE_STACK; | 469 | 398k | return ret; | 470 | 398k | } |
Line | Count | Source | 148 | 398k | { | 149 | 398k | opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR; | 150 | 398k | opus_int32 nSamplesOutDec, LBRR_symbol; | 151 | 398k | opus_int16 *samplesOut1_tmp[ 2 ]; | 152 | 398k | VARDECL( opus_int16, samplesOut1_tmp_storage1 ); | 153 | 398k | VARDECL( opus_int16, samplesOut2_tmp ); | 154 | 398k | opus_int32 MS_pred_Q13[ 2 ] = { 0 }; | 155 | 398k | opus_int16 *resample_out_ptr; | 156 | 398k | silk_decoder *psDec = ( silk_decoder * )decState; | 157 | 398k | silk_decoder_state *channel_state = psDec->channel_state; | 158 | 398k | opus_int has_side; | 159 | 398k | opus_int stereo_to_mono; | 160 | | #ifdef ENABLE_OSCE_BWE | 161 | | ALLOC(resamp_buffer, 3 * MAX_FRAME_LENGTH, opus_int16); | 162 | | #endif | 163 | 398k | SAVE_STACK; | 164 | | | 165 | 398k | celt_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 ); | 166 | | | 167 | | /**********************************/ | 168 | | /* Test if first frame in payload */ | 169 | | /**********************************/ | 170 | 398k | if( newPacketFlag ) { | 171 | 688k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 172 | 428k | channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */ | 173 | 428k | } | 174 | 259k | } | 175 | | | 176 | | /* If Mono -> Stereo transition in bitstream: init state of second channel */ | 177 | 398k | if( decControl->nChannelsInternal > psDec->nChannelsInternal ) { | 178 | 135k | ret += silk_init_decoder( &channel_state[ 1 ] ); | 179 | 135k | } | 180 | | | 181 | 398k | stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 && | 182 | 0 | ( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz ); | 183 | | | 184 | 398k | if( channel_state[ 0 ].nFramesDecoded == 0 ) { | 185 | 688k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 186 | 428k | opus_int fs_kHz_dec; | 187 | 428k | if( decControl->payloadSize_ms == 0 ) { | 188 | | /* Assuming packet loss, use 10 ms */ | 189 | 0 | channel_state[ n ].nFramesPerPacket = 1; | 190 | 0 | channel_state[ n ].nb_subfr = 2; | 191 | 428k | } else if( decControl->payloadSize_ms == 10 ) { | 192 | 209k | channel_state[ n ].nFramesPerPacket = 1; | 193 | 209k | channel_state[ n ].nb_subfr = 2; | 194 | 218k | } else if( decControl->payloadSize_ms == 20 ) { | 195 | 88.1k | channel_state[ n ].nFramesPerPacket = 1; | 196 | 88.1k | channel_state[ n ].nb_subfr = 4; | 197 | 130k | } else if( decControl->payloadSize_ms == 40 ) { | 198 | 16.2k | channel_state[ n ].nFramesPerPacket = 2; | 199 | 16.2k | channel_state[ n ].nb_subfr = 4; | 200 | 114k | } else if( decControl->payloadSize_ms == 60 ) { | 201 | 114k | channel_state[ n ].nFramesPerPacket = 3; | 202 | 114k | channel_state[ n ].nb_subfr = 4; | 203 | 114k | } else { | 204 | 0 | celt_assert( 0 ); | 205 | 0 | RESTORE_STACK; | 206 | 0 | return SILK_DEC_INVALID_FRAME_SIZE; | 207 | 0 | } | 208 | 428k | fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1; | 209 | 428k | if( fs_kHz_dec != 8 && fs_kHz_dec != 12 && fs_kHz_dec != 16 ) { | 210 | 0 | celt_assert( 0 ); | 211 | 0 | RESTORE_STACK; | 212 | 0 | return SILK_DEC_INVALID_SAMPLING_FREQUENCY; | 213 | 0 | } | 214 | 428k | ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate ); | 215 | 428k | } | 216 | 259k | } | 217 | | | 218 | 398k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) { | 219 | 0 | silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) ); | 220 | 0 | silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) ); | 221 | 0 | silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) ); | 222 | 0 | } | 223 | 398k | psDec->nChannelsAPI = decControl->nChannelsAPI; | 224 | 398k | psDec->nChannelsInternal = decControl->nChannelsInternal; | 225 | | | 226 | 398k | if( decControl->API_sampleRate > (opus_int32)MAX_API_FS_KHZ * 1000 || decControl->API_sampleRate < 8000 ) { | 227 | 0 | ret = SILK_DEC_INVALID_SAMPLING_FREQUENCY; | 228 | 0 | RESTORE_STACK; | 229 | 0 | return( ret ); | 230 | 0 | } | 231 | | | 232 | 398k | if( lostFlag != FLAG_PACKET_LOST && channel_state[ 0 ].nFramesDecoded == 0 ) { | 233 | | /* First decoder call for this payload */ | 234 | | /* Decode VAD flags and LBRR flag */ | 235 | 494k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 236 | 851k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { | 237 | 548k | channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1); | 238 | 548k | } | 239 | 303k | channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1); | 240 | 303k | } | 241 | | /* Decode LBRR flags */ | 242 | 494k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 243 | 303k | silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) ); | 244 | 303k | if( channel_state[ n ].LBRR_flag ) { | 245 | 131k | if( channel_state[ n ].nFramesPerPacket == 1 ) { | 246 | 49.7k | channel_state[ n ].LBRR_flags[ 0 ] = 1; | 247 | 81.7k | } else { | 248 | 81.7k | LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1; | 249 | 318k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { | 250 | 236k | channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1; | 251 | 236k | } | 252 | 81.7k | } | 253 | 131k | } | 254 | 303k | } | 255 | | | 256 | 190k | if( lostFlag == FLAG_DECODE_NORMAL ) { | 257 | | /* Regular decoding: skip all LBRR data */ | 258 | 403k | for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) { | 259 | 667k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 260 | 417k | if( channel_state[ n ].LBRR_flags[ i ] ) { | 261 | 144k | opus_int16 pulses[ MAX_FRAME_LENGTH ]; | 262 | 144k | opus_int condCoding; | 263 | | | 264 | 144k | if( decControl->nChannelsInternal == 2 && n == 0 ) { | 265 | 62.4k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); | 266 | 62.4k | if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) { | 267 | 29.5k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); | 268 | 29.5k | } | 269 | 62.4k | } | 270 | | /* Use conditional coding if previous frame available */ | 271 | 144k | if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) { | 272 | 46.7k | condCoding = CODE_CONDITIONALLY; | 273 | 97.4k | } else { | 274 | 97.4k | condCoding = CODE_INDEPENDENTLY; | 275 | 97.4k | } | 276 | 144k | silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding ); | 277 | 144k | silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType, | 278 | 144k | channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length ); | 279 | 144k | } | 280 | 417k | } | 281 | 249k | } | 282 | 154k | } | 283 | 190k | } | 284 | | | 285 | | /* Get MS predictor index */ | 286 | 398k | if( decControl->nChannelsInternal == 2 ) { | 287 | 274k | if( lostFlag == FLAG_DECODE_NORMAL || | 288 | 106k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) ) | 289 | 189k | { | 290 | 189k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); | 291 | | /* For LBRR data, decode mid-only flag only if side-channel's LBRR flag is false */ | 292 | 189k | if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) || | 293 | 107k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) ) | 294 | 94.6k | { | 295 | 94.6k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); | 296 | 94.6k | } else { | 297 | 94.5k | decode_only_middle = 0; | 298 | 94.5k | } | 299 | 189k | } else { | 300 | 256k | for( n = 0; n < 2; n++ ) { | 301 | 170k | MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ]; | 302 | 170k | } | 303 | 85.4k | } | 304 | 274k | } | 305 | | | 306 | | /* Reset side channel decoder prediction memory for first frame with side coding */ | 307 | 398k | if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) { | 308 | 20.1k | silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) ); | 309 | 20.1k | silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) ); | 310 | 20.1k | psDec->channel_state[ 1 ].lagPrev = 100; | 311 | 20.1k | psDec->channel_state[ 1 ].LastGainIndex = 10; | 312 | 20.1k | psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY; | 313 | 20.1k | psDec->channel_state[ 1 ].first_frame_after_reset = 1; | 314 | 20.1k | } | 315 | | | 316 | | /* Check if the temp buffer fits into the output PCM buffer. If it fits, | 317 | | we can delay allocating the temp buffer until after the SILK peak stack | 318 | | usage. We need to use a < and not a <= because of the two extra samples. */ | 319 | 398k | ALLOC( samplesOut1_tmp_storage1, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ), | 320 | 398k | opus_int16 ); | 321 | 398k | samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1; | 322 | 398k | samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2; | 323 | | | 324 | 398k | if( lostFlag == FLAG_DECODE_NORMAL ) { | 325 | 249k | has_side = !decode_only_middle; | 326 | 249k | } else { | 327 | 148k | has_side = !psDec->prev_decode_only_middle | 328 | 11.7k | || (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 ); | 329 | 148k | } | 330 | 398k | channel_state[ 0 ].sPLC.enable_deep_plc = decControl->enable_deep_plc; | 331 | | /* Call decoder for one frame */ | 332 | 1.07M | for( n = 0; n < decControl->nChannelsInternal; n++ ) { | 333 | 673k | if( n == 0 || has_side ) { | 334 | 646k | opus_int FrameIndex; | 335 | 646k | opus_int condCoding; | 336 | | | 337 | 646k | FrameIndex = channel_state[ 0 ].nFramesDecoded - n; | 338 | | /* Use independent coding if no previous frame available */ | 339 | 646k | if( FrameIndex <= 0 ) { | 340 | 414k | condCoding = CODE_INDEPENDENTLY; | 341 | 414k | } else if( lostFlag == FLAG_DECODE_LBRR ) { | 342 | 64.5k | condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY; | 343 | 167k | } else if( n > 0 && psDec->prev_decode_only_middle ) { | 344 | | /* If we skipped a side frame in this packet, we don't | 345 | | need LTP scaling; the LTP state is well-defined. */ | 346 | 5.50k | condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; | 347 | 161k | } else { | 348 | 161k | condCoding = CODE_CONDITIONALLY; | 349 | 161k | } | 350 | | #ifdef ENABLE_OSCE | 351 | | if ( channel_state[n].osce.method != decControl->osce_method ) { | 352 | | osce_reset( &channel_state[n].osce, decControl->osce_method ); | 353 | | } | 354 | | #endif | 355 | 646k | ret += silk_decode_frame( &channel_state[ n ], psRangeDec, &samplesOut1_tmp[ n ][ 2 ], &nSamplesOutDec, lostFlag, condCoding, | 356 | | #ifdef ENABLE_DEEP_PLC | 357 | | n == 0 ? lpcnet : NULL, | 358 | | #endif | 359 | | #ifdef ENABLE_OSCE | 360 | | &psDec->osce_model, | 361 | | #endif | 362 | 646k | arch); | 363 | 646k | } else { | 364 | 26.3k | silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) ); | 365 | 26.3k | } | 366 | 673k | channel_state[ n ].nFramesDecoded++; | 367 | 673k | } | 368 | | | 369 | 398k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) { | 370 | | /* Convert Mid/Side to Left/Right */ | 371 | 102k | silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec ); | 372 | 295k | } else { | 373 | | /* Buffering */ | 374 | 295k | silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) ); | 375 | 295k | silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) ); | 376 | 295k | } | 377 | | | 378 | | /* Number of output samples */ | 379 | 398k | *nSamplesOut = silk_DIV32( nSamplesOutDec * decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) ); | 380 | | | 381 | | /* Set up pointers to temp buffers */ | 382 | 398k | ALLOC( samplesOut2_tmp, *nSamplesOut, opus_int16 ); | 383 | 398k | resample_out_ptr = samplesOut2_tmp; | 384 | | | 385 | 899k | for( n = 0; n < silk_min( decControl->nChannelsAPI, decControl->nChannelsInternal ); n++ ) { | 386 | | | 387 | | #ifdef ENABLE_OSCE_BWE | 388 | | /* Resample or extend decoded signal to API_sampleRate */ | 389 | | if (decControl->osce_extended_mode == OSCE_MODE_SILK_BBWE) { | 390 | | silk_assert(decControl->API_sampleRate == 48000); | 391 | | | 392 | | if (decControl->prev_osce_extended_mode != OSCE_MODE_SILK_BBWE) { | 393 | | /* Reset the BWE state */ | 394 | | osce_bwe_reset( &channel_state[ n ].osce_bwe ); | 395 | | } | 396 | | | 397 | | osce_bwe(&psDec->osce_model, &channel_state[ n ].osce_bwe, | 398 | | resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec, arch); | 399 | | | 400 | | if (decControl->prev_osce_extended_mode == OSCE_MODE_SILK_ONLY || | 401 | | decControl->prev_osce_extended_mode == OSCE_MODE_HYBRID) { | 402 | | /* cross-fade with upsampled signal */ | 403 | | silk_resampler( &channel_state[ n ].resampler_state, resamp_buffer, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); | 404 | | osce_bwe_cross_fade_10ms(resample_out_ptr, resamp_buffer, 480); | 405 | | } | 406 | | } else { | 407 | | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); | 408 | | if (decControl->prev_osce_extended_mode == OSCE_MODE_SILK_BBWE) { | 409 | | osce_bwe(&psDec->osce_model, &channel_state[ n ].osce_bwe, | 410 | | resamp_buffer, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec, arch); | 411 | | /* cross-fade with upsampled signal */ | 412 | | osce_bwe_cross_fade_10ms(resample_out_ptr, resamp_buffer, 480); | 413 | | } | 414 | | } | 415 | | #else | 416 | | /* Resample decoded signal to API_sampleRate */ | 417 | 501k | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); | 418 | 501k | #endif | 419 | | /* Interleave if stereo output and stereo stream */ | 420 | 501k | if( decControl->nChannelsAPI == 2 ) { | 421 | 108M | for( i = 0; i < *nSamplesOut; i++ ) { | 422 | 108M | samplesOut[ n + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); | 423 | 108M | } | 424 | 263k | } else { | 425 | 92.2M | for( i = 0; i < *nSamplesOut; i++ ) { | 426 | 92.0M | samplesOut[ i ] = INT16TORES(resample_out_ptr[ i ]); | 427 | 92.0M | } | 428 | 237k | } | 429 | 501k | } | 430 | | | 431 | | #ifdef ENABLE_OSCE_BWE | 432 | | decControl->prev_osce_extended_mode = decControl->osce_extended_mode; | 433 | | #endif | 434 | | | 435 | | /* Create two channel output from mono stream */ | 436 | 398k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) { | 437 | 57.7k | if ( stereo_to_mono ){ | 438 | | /* Resample right channel for newly collapsed stereo just in case | 439 | | we weren't doing collapsing when switching to mono */ | 440 | 0 | ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec ); | 441 | |
| 442 | 0 | for( i = 0; i < *nSamplesOut; i++ ) { | 443 | 0 | samplesOut[ 1 + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); | 444 | 0 | } | 445 | 57.7k | } else { | 446 | 24.2M | for( i = 0; i < *nSamplesOut; i++ ) { | 447 | 24.1M | samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ]; | 448 | 24.1M | } | 449 | 57.7k | } | 450 | 57.7k | } | 451 | | | 452 | | /* Export pitch lag, measured at 48 kHz sampling rate */ | 453 | 398k | if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) { | 454 | 89.2k | int mult_tab[ 3 ] = { 6, 4, 3 }; | 455 | 89.2k | decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ]; | 456 | 309k | } else { | 457 | 309k | decControl->prevPitchLag = 0; | 458 | 309k | } | 459 | | | 460 | 398k | if( lostFlag == FLAG_PACKET_LOST ) { | 461 | | /* On packet loss, remove the gain clamping to prevent having the energy "bounce back" | 462 | | if we lose packets when the energy is going down */ | 463 | 193k | for ( i = 0; i < psDec->nChannelsInternal; i++ ) | 464 | 124k | psDec->channel_state[ i ].LastGainIndex = 10; | 465 | 329k | } else { | 466 | 329k | psDec->prev_decode_only_middle = decode_only_middle; | 467 | 329k | } | 468 | 398k | RESTORE_STACK; | 469 | 398k | return ret; | 470 | 398k | } |
|
471 | | |
472 | | #if 0 |
473 | | /* Getting table of contents for a packet */ |
474 | | opus_int silk_get_TOC( |
475 | | const opus_uint8 *payload, /* I Payload data */ |
476 | | const opus_int nBytesIn, /* I Number of input bytes */ |
477 | | const opus_int nFramesPerPayload, /* I Number of SILK frames per payload */ |
478 | | silk_TOC_struct *Silk_TOC /* O Type of content */ |
479 | | ) |
480 | | { |
481 | | opus_int i, flags, ret = SILK_NO_ERROR; |
482 | | |
483 | | if( nBytesIn < 1 ) { |
484 | | return -1; |
485 | | } |
486 | | if( nFramesPerPayload < 0 || nFramesPerPayload > 3 ) { |
487 | | return -1; |
488 | | } |
489 | | |
490 | | silk_memset( Silk_TOC, 0, sizeof( *Silk_TOC ) ); |
491 | | |
492 | | /* For stereo, extract the flags for the mid channel */ |
493 | | flags = silk_RSHIFT( payload[ 0 ], 7 - nFramesPerPayload ) & ( silk_LSHIFT( 1, nFramesPerPayload + 1 ) - 1 ); |
494 | | |
495 | | Silk_TOC->inbandFECFlag = flags & 1; |
496 | | for( i = nFramesPerPayload - 1; i >= 0 ; i-- ) { |
497 | | flags = silk_RSHIFT( flags, 1 ); |
498 | | Silk_TOC->VADFlags[ i ] = flags & 1; |
499 | | Silk_TOC->VADFlag |= flags & 1; |
500 | | } |
501 | | |
502 | | return ret; |
503 | | } |
504 | | #endif |