Line | Count | Source (jump to first uncovered line) |
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 | 59.7k | { |
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 | 59.7k | (void) decState; |
74 | 59.7k | (void) data; |
75 | 59.7k | (void) len; |
76 | 59.7k | return SILK_NO_ERROR; |
77 | 59.7k | #endif |
78 | 59.7k | } |
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 | 1.99M | { |
84 | 1.99M | opus_int ret = SILK_NO_ERROR; |
85 | | |
86 | 1.99M | *decSizeBytes = sizeof( silk_decoder ); |
87 | | |
88 | 1.99M | return ret; |
89 | 1.99M | } |
90 | | |
91 | | /* Reset decoder state */ |
92 | | opus_int silk_ResetDecoder( /* O Returns error code */ |
93 | | void *decState /* I/O State */ |
94 | | ) |
95 | 3.19M | { |
96 | 3.19M | opus_int n, ret = SILK_NO_ERROR; |
97 | 3.19M | silk_decoder_state *channel_state = ((silk_decoder *)decState)->channel_state; |
98 | | |
99 | 9.59M | for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { |
100 | 6.39M | ret = silk_reset_decoder( &channel_state[ n ] ); |
101 | 6.39M | } |
102 | 3.19M | silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); |
103 | | /* Not strictly needed, but it's cleaner that way */ |
104 | 3.19M | ((silk_decoder *)decState)->prev_decode_only_middle = 0; |
105 | | |
106 | 3.19M | return ret; |
107 | 3.19M | } |
108 | | |
109 | | |
110 | | opus_int silk_InitDecoder( /* O Returns error code */ |
111 | | void *decState /* I/O State */ |
112 | | ) |
113 | 59.7k | { |
114 | 59.7k | opus_int n, ret = SILK_NO_ERROR; |
115 | 59.7k | 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 | 59.7k | #ifndef USE_WEIGHTS_FILE |
120 | | /* load osce models */ |
121 | 59.7k | silk_LoadOSCEModels(decState, NULL, 0); |
122 | 59.7k | #endif |
123 | | |
124 | 179k | for( n = 0; n < DECODER_NUM_CHANNELS; n++ ) { |
125 | 119k | ret = silk_init_decoder( &channel_state[ n ] ); |
126 | 119k | } |
127 | 59.7k | silk_memset(&((silk_decoder *)decState)->sStereo, 0, sizeof(((silk_decoder *)decState)->sStereo)); |
128 | | /* Not strictly needed, but it's cleaner that way */ |
129 | 59.7k | ((silk_decoder *)decState)->prev_decode_only_middle = 0; |
130 | | |
131 | 59.7k | return ret; |
132 | 59.7k | } |
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 | 334k | { |
149 | 334k | opus_int i, n, decode_only_middle = 0, ret = SILK_NO_ERROR; |
150 | 334k | opus_int32 nSamplesOutDec, LBRR_symbol; |
151 | 334k | opus_int16 *samplesOut1_tmp[ 2 ]; |
152 | 334k | VARDECL( opus_int16, samplesOut1_tmp_storage1 ); |
153 | 334k | VARDECL( opus_int16, samplesOut2_tmp ); |
154 | 334k | opus_int32 MS_pred_Q13[ 2 ] = { 0 }; |
155 | 334k | opus_int16 *resample_out_ptr; |
156 | 334k | silk_decoder *psDec = ( silk_decoder * )decState; |
157 | 334k | silk_decoder_state *channel_state = psDec->channel_state; |
158 | 334k | opus_int has_side; |
159 | 334k | opus_int stereo_to_mono; |
160 | | #ifdef ENABLE_OSCE_BWE |
161 | | ALLOC(resamp_buffer, 3 * MAX_FRAME_LENGTH, opus_int16); |
162 | | #endif |
163 | 334k | SAVE_STACK; |
164 | | |
165 | 334k | celt_assert( decControl->nChannelsInternal == 1 || decControl->nChannelsInternal == 2 ); |
166 | | |
167 | | /**********************************/ |
168 | | /* Test if first frame in payload */ |
169 | | /**********************************/ |
170 | 334k | if( newPacketFlag ) { |
171 | 662k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
172 | 365k | channel_state[ n ].nFramesDecoded = 0; /* Used to count frames in packet */ |
173 | 365k | } |
174 | 297k | } |
175 | | |
176 | | /* If Mono -> Stereo transition in bitstream: init state of second channel */ |
177 | 334k | if( decControl->nChannelsInternal > psDec->nChannelsInternal ) { |
178 | 8.76k | ret += silk_init_decoder( &channel_state[ 1 ] ); |
179 | 8.76k | } |
180 | | |
181 | 334k | stereo_to_mono = decControl->nChannelsInternal == 1 && psDec->nChannelsInternal == 2 && |
182 | 334k | ( decControl->internalSampleRate == 1000*channel_state[ 0 ].fs_kHz ); |
183 | | |
184 | 334k | if( channel_state[ 0 ].nFramesDecoded == 0 ) { |
185 | 662k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
186 | 365k | opus_int fs_kHz_dec; |
187 | 365k | 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 | 365k | } else if( decControl->payloadSize_ms == 10 ) { |
192 | 235k | channel_state[ n ].nFramesPerPacket = 1; |
193 | 235k | channel_state[ n ].nb_subfr = 2; |
194 | 235k | } else if( decControl->payloadSize_ms == 20 ) { |
195 | 75.0k | channel_state[ n ].nFramesPerPacket = 1; |
196 | 75.0k | channel_state[ n ].nb_subfr = 4; |
197 | 75.0k | } else if( decControl->payloadSize_ms == 40 ) { |
198 | 41.0k | channel_state[ n ].nFramesPerPacket = 2; |
199 | 41.0k | channel_state[ n ].nb_subfr = 4; |
200 | 41.0k | } else if( decControl->payloadSize_ms == 60 ) { |
201 | 13.4k | channel_state[ n ].nFramesPerPacket = 3; |
202 | 13.4k | channel_state[ n ].nb_subfr = 4; |
203 | 13.4k | } else { |
204 | 0 | celt_assert( 0 ); |
205 | 0 | RESTORE_STACK; |
206 | 0 | return SILK_DEC_INVALID_FRAME_SIZE; |
207 | 0 | } |
208 | 365k | fs_kHz_dec = ( decControl->internalSampleRate >> 10 ) + 1; |
209 | 365k | 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 | 365k | ret += silk_decoder_set_fs( &channel_state[ n ], fs_kHz_dec, decControl->API_sampleRate ); |
215 | 365k | } |
216 | 297k | } |
217 | | |
218 | 334k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 && ( psDec->nChannelsAPI == 1 || psDec->nChannelsInternal == 1 ) ) { |
219 | 5.12k | silk_memset( psDec->sStereo.pred_prev_Q13, 0, sizeof( psDec->sStereo.pred_prev_Q13 ) ); |
220 | 5.12k | silk_memset( psDec->sStereo.sSide, 0, sizeof( psDec->sStereo.sSide ) ); |
221 | 5.12k | silk_memcpy( &channel_state[ 1 ].resampler_state, &channel_state[ 0 ].resampler_state, sizeof( silk_resampler_state_struct ) ); |
222 | 5.12k | } |
223 | 334k | psDec->nChannelsAPI = decControl->nChannelsAPI; |
224 | 334k | psDec->nChannelsInternal = decControl->nChannelsInternal; |
225 | | |
226 | 334k | 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 | 334k | 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 | 304k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
236 | 407k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { |
237 | 237k | channel_state[ n ].VAD_flags[ i ] = ec_dec_bit_logp(psRangeDec, 1); |
238 | 237k | } |
239 | 169k | channel_state[ n ].LBRR_flag = ec_dec_bit_logp(psRangeDec, 1); |
240 | 169k | } |
241 | | /* Decode LBRR flags */ |
242 | 304k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
243 | 169k | silk_memset( channel_state[ n ].LBRR_flags, 0, sizeof( channel_state[ n ].LBRR_flags ) ); |
244 | 169k | if( channel_state[ n ].LBRR_flag ) { |
245 | 62.5k | if( channel_state[ n ].nFramesPerPacket == 1 ) { |
246 | 45.9k | channel_state[ n ].LBRR_flags[ 0 ] = 1; |
247 | 45.9k | } else { |
248 | 16.5k | LBRR_symbol = ec_dec_icdf( psRangeDec, silk_LBRR_flags_iCDF_ptr[ channel_state[ n ].nFramesPerPacket - 2 ], 8 ) + 1; |
249 | 56.3k | for( i = 0; i < channel_state[ n ].nFramesPerPacket; i++ ) { |
250 | 39.7k | channel_state[ n ].LBRR_flags[ i ] = silk_RSHIFT( LBRR_symbol, i ) & 1; |
251 | 39.7k | } |
252 | 16.5k | } |
253 | 62.5k | } |
254 | 169k | } |
255 | | |
256 | 134k | if( lostFlag == FLAG_DECODE_NORMAL ) { |
257 | | /* Regular decoding: skip all LBRR data */ |
258 | 306k | for( i = 0; i < channel_state[ 0 ].nFramesPerPacket; i++ ) { |
259 | 409k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
260 | 237k | if( channel_state[ n ].LBRR_flags[ i ] ) { |
261 | 73.2k | opus_int16 pulses[ MAX_FRAME_LENGTH ]; |
262 | 73.2k | opus_int condCoding; |
263 | | |
264 | 73.2k | if( decControl->nChannelsInternal == 2 && n == 0 ) { |
265 | 16.5k | silk_stereo_decode_pred( psRangeDec, MS_pred_Q13 ); |
266 | 16.5k | if( channel_state[ 1 ].LBRR_flags[ i ] == 0 ) { |
267 | 6.90k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); |
268 | 6.90k | } |
269 | 16.5k | } |
270 | | /* Use conditional coding if previous frame available */ |
271 | 73.2k | if( i > 0 && channel_state[ n ].LBRR_flags[ i - 1 ] ) { |
272 | 10.5k | condCoding = CODE_CONDITIONALLY; |
273 | 62.7k | } else { |
274 | 62.7k | condCoding = CODE_INDEPENDENTLY; |
275 | 62.7k | } |
276 | 73.2k | silk_decode_indices( &channel_state[ n ], psRangeDec, i, 1, condCoding ); |
277 | 73.2k | silk_decode_pulses( psRangeDec, pulses, channel_state[ n ].indices.signalType, |
278 | 73.2k | channel_state[ n ].indices.quantOffsetType, channel_state[ n ].frame_length ); |
279 | 73.2k | } |
280 | 237k | } |
281 | 172k | } |
282 | 134k | } |
283 | 134k | } |
284 | | |
285 | | /* Get MS predictor index */ |
286 | 334k | if( decControl->nChannelsInternal == 2 ) { |
287 | 98.7k | if( lostFlag == FLAG_DECODE_NORMAL || |
288 | 98.7k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 0 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 1 ) ) |
289 | 65.5k | { |
290 | 65.5k | 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 | 65.5k | if( ( lostFlag == FLAG_DECODE_NORMAL && channel_state[ 1 ].VAD_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) || |
293 | 65.5k | ( lostFlag == FLAG_DECODE_LBRR && channel_state[ 1 ].LBRR_flags[ channel_state[ 0 ].nFramesDecoded ] == 0 ) ) |
294 | 43.6k | { |
295 | 43.6k | silk_stereo_decode_mid_only( psRangeDec, &decode_only_middle ); |
296 | 43.6k | } else { |
297 | 21.9k | decode_only_middle = 0; |
298 | 21.9k | } |
299 | 65.5k | } else { |
300 | 99.7k | for( n = 0; n < 2; n++ ) { |
301 | 66.4k | MS_pred_Q13[ n ] = psDec->sStereo.pred_prev_Q13[ n ]; |
302 | 66.4k | } |
303 | 33.2k | } |
304 | 98.7k | } |
305 | | |
306 | | /* Reset side channel decoder prediction memory for first frame with side coding */ |
307 | 334k | if( decControl->nChannelsInternal == 2 && decode_only_middle == 0 && psDec->prev_decode_only_middle == 1 ) { |
308 | 9.61k | silk_memset( psDec->channel_state[ 1 ].outBuf, 0, sizeof(psDec->channel_state[ 1 ].outBuf) ); |
309 | 9.61k | silk_memset( psDec->channel_state[ 1 ].sLPC_Q14_buf, 0, sizeof(psDec->channel_state[ 1 ].sLPC_Q14_buf) ); |
310 | 9.61k | psDec->channel_state[ 1 ].lagPrev = 100; |
311 | 9.61k | psDec->channel_state[ 1 ].LastGainIndex = 10; |
312 | 9.61k | psDec->channel_state[ 1 ].prevSignalType = TYPE_NO_VOICE_ACTIVITY; |
313 | 9.61k | psDec->channel_state[ 1 ].first_frame_after_reset = 1; |
314 | 9.61k | } |
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 | 334k | ALLOC( samplesOut1_tmp_storage1, decControl->nChannelsInternal*(channel_state[ 0 ].frame_length + 2 ), |
320 | 334k | opus_int16 ); |
321 | 334k | samplesOut1_tmp[ 0 ] = samplesOut1_tmp_storage1; |
322 | 334k | samplesOut1_tmp[ 1 ] = samplesOut1_tmp_storage1 + channel_state[ 0 ].frame_length + 2; |
323 | | |
324 | 334k | if( lostFlag == FLAG_DECODE_NORMAL ) { |
325 | 172k | has_side = !decode_only_middle; |
326 | 172k | } else { |
327 | 162k | has_side = !psDec->prev_decode_only_middle |
328 | 162k | || (decControl->nChannelsInternal == 2 && lostFlag == FLAG_DECODE_LBRR && channel_state[1].LBRR_flags[ channel_state[1].nFramesDecoded ] == 1 ); |
329 | 162k | } |
330 | 334k | channel_state[ 0 ].sPLC.enable_deep_plc = decControl->enable_deep_plc; |
331 | | /* Call decoder for one frame */ |
332 | 768k | for( n = 0; n < decControl->nChannelsInternal; n++ ) { |
333 | 433k | if( n == 0 || has_side ) { |
334 | 420k | opus_int FrameIndex; |
335 | 420k | opus_int condCoding; |
336 | | |
337 | 420k | FrameIndex = channel_state[ 0 ].nFramesDecoded - n; |
338 | | /* Use independent coding if no previous frame available */ |
339 | 420k | if( FrameIndex <= 0 ) { |
340 | 357k | condCoding = CODE_INDEPENDENTLY; |
341 | 357k | } else if( lostFlag == FLAG_DECODE_LBRR ) { |
342 | 0 | condCoding = channel_state[ n ].LBRR_flags[ FrameIndex - 1 ] ? CODE_CONDITIONALLY : CODE_INDEPENDENTLY; |
343 | 63.5k | } 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 | 2.16k | condCoding = CODE_INDEPENDENTLY_NO_LTP_SCALING; |
347 | 61.4k | } else { |
348 | 61.4k | condCoding = CODE_CONDITIONALLY; |
349 | 61.4k | } |
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 | 420k | 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 | 420k | arch); |
363 | 420k | } else { |
364 | 12.8k | silk_memset( &samplesOut1_tmp[ n ][ 2 ], 0, nSamplesOutDec * sizeof( opus_int16 ) ); |
365 | 12.8k | } |
366 | 433k | channel_state[ n ].nFramesDecoded++; |
367 | 433k | } |
368 | | |
369 | 334k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 2 ) { |
370 | | /* Convert Mid/Side to Left/Right */ |
371 | 96.9k | silk_stereo_MS_to_LR( &psDec->sStereo, samplesOut1_tmp[ 0 ], samplesOut1_tmp[ 1 ], MS_pred_Q13, channel_state[ 0 ].fs_kHz, nSamplesOutDec ); |
372 | 237k | } else { |
373 | | /* Buffering */ |
374 | 237k | silk_memcpy( samplesOut1_tmp[ 0 ], psDec->sStereo.sMid, 2 * sizeof( opus_int16 ) ); |
375 | 237k | silk_memcpy( psDec->sStereo.sMid, &samplesOut1_tmp[ 0 ][ nSamplesOutDec ], 2 * sizeof( opus_int16 ) ); |
376 | 237k | } |
377 | | |
378 | | /* Number of output samples */ |
379 | 334k | *nSamplesOut = silk_DIV32( nSamplesOutDec * decControl->API_sampleRate, silk_SMULBB( channel_state[ 0 ].fs_kHz, 1000 ) ); |
380 | | |
381 | | /* Set up pointers to temp buffers */ |
382 | 334k | ALLOC( samplesOut2_tmp, *nSamplesOut, opus_int16 ); |
383 | 334k | resample_out_ptr = samplesOut2_tmp; |
384 | | |
385 | 766k | 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 | 431k | ret += silk_resampler( &channel_state[ n ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ n ][ 1 ], nSamplesOutDec ); |
418 | 431k | #endif |
419 | | /* Interleave if stereo output and stereo stream */ |
420 | 431k | if( decControl->nChannelsAPI == 2 ) { |
421 | 286M | for( i = 0; i < *nSamplesOut; i++ ) { |
422 | 286M | samplesOut[ n + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); |
423 | 286M | } |
424 | 406k | } else { |
425 | 15.1M | for( i = 0; i < *nSamplesOut; i++ ) { |
426 | 15.0M | samplesOut[ i ] = INT16TORES(resample_out_ptr[ i ]); |
427 | 15.0M | } |
428 | 25.2k | } |
429 | 431k | } |
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 | 334k | if( decControl->nChannelsAPI == 2 && decControl->nChannelsInternal == 1 ) { |
437 | 212k | 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 | 1.03k | ret += silk_resampler( &channel_state[ 1 ].resampler_state, resample_out_ptr, &samplesOut1_tmp[ 0 ][ 1 ], nSamplesOutDec ); |
441 | | |
442 | 768k | for( i = 0; i < *nSamplesOut; i++ ) { |
443 | 767k | samplesOut[ 1 + 2 * i ] = INT16TORES(resample_out_ptr[ i ]); |
444 | 767k | } |
445 | 211k | } else { |
446 | 120M | for( i = 0; i < *nSamplesOut; i++ ) { |
447 | 120M | samplesOut[ 1 + 2 * i ] = samplesOut[ 0 + 2 * i ]; |
448 | 120M | } |
449 | 211k | } |
450 | 212k | } |
451 | | |
452 | | /* Export pitch lag, measured at 48 kHz sampling rate */ |
453 | 334k | if( channel_state[ 0 ].prevSignalType == TYPE_VOICED ) { |
454 | 31.1k | int mult_tab[ 3 ] = { 6, 4, 3 }; |
455 | 31.1k | decControl->prevPitchLag = channel_state[ 0 ].lagPrev * mult_tab[ ( channel_state[ 0 ].fs_kHz - 8 ) >> 2 ]; |
456 | 303k | } else { |
457 | 303k | decControl->prevPitchLag = 0; |
458 | 303k | } |
459 | | |
460 | 334k | 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 | 358k | for ( i = 0; i < psDec->nChannelsInternal; i++ ) |
464 | 195k | psDec->channel_state[ i ].LastGainIndex = 10; |
465 | 172k | } else { |
466 | 172k | psDec->prev_decode_only_middle = decode_only_middle; |
467 | 172k | } |
468 | 334k | RESTORE_STACK; |
469 | 334k | return ret; |
470 | 334k | } |
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 |