diff options
Diffstat (limited to 'lib/rbcodec/codecs/libopus/opus_decoder.c')
-rw-r--r-- | lib/rbcodec/codecs/libopus/opus_decoder.c | 999 |
1 files changed, 999 insertions, 0 deletions
diff --git a/lib/rbcodec/codecs/libopus/opus_decoder.c b/lib/rbcodec/codecs/libopus/opus_decoder.c new file mode 100644 index 0000000000..7103b183b8 --- /dev/null +++ b/lib/rbcodec/codecs/libopus/opus_decoder.c | |||
@@ -0,0 +1,999 @@ | |||
1 | /* Copyright (c) 2010 Xiph.Org Foundation, Skype Limited | ||
2 | Written by Jean-Marc Valin and Koen Vos */ | ||
3 | /* | ||
4 | Redistribution and use in source and binary forms, with or without | ||
5 | modification, are permitted provided that the following conditions | ||
6 | are met: | ||
7 | |||
8 | - Redistributions of source code must retain the above copyright | ||
9 | notice, this list of conditions and the following disclaimer. | ||
10 | |||
11 | - Redistributions in binary form must reproduce the above copyright | ||
12 | notice, this list of conditions and the following disclaimer in the | ||
13 | documentation and/or other materials provided with the distribution. | ||
14 | |||
15 | THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | ||
16 | ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | ||
17 | LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | ||
18 | A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER | ||
19 | OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, | ||
20 | EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, | ||
21 | PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR | ||
22 | PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF | ||
23 | LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING | ||
24 | NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS | ||
25 | SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | ||
26 | */ | ||
27 | |||
28 | #ifdef HAVE_CONFIG_H | ||
29 | #include "opus_config.h" | ||
30 | #endif | ||
31 | |||
32 | #ifndef OPUS_BUILD | ||
33 | #error "OPUS_BUILD _MUST_ be defined to build Opus and you probably want a decent config.h, see README for more details." | ||
34 | #endif | ||
35 | |||
36 | #include <stdarg.h> | ||
37 | #include "celt.h" | ||
38 | #include "opus.h" | ||
39 | #include "entdec.h" | ||
40 | #include "modes.h" | ||
41 | #include "API.h" | ||
42 | #include "stack_alloc.h" | ||
43 | #include "float_cast.h" | ||
44 | #include "opus_private.h" | ||
45 | #include "os_support.h" | ||
46 | #include "structs.h" | ||
47 | #include "define.h" | ||
48 | #include "mathops.h" | ||
49 | |||
50 | struct OpusDecoder { | ||
51 | int celt_dec_offset; | ||
52 | int silk_dec_offset; | ||
53 | int channels; | ||
54 | opus_int32 Fs; /** Sampling rate (at the API level) */ | ||
55 | silk_DecControlStruct DecControl; | ||
56 | int decode_gain; | ||
57 | |||
58 | /* Everything beyond this point gets cleared on a reset */ | ||
59 | #define OPUS_DECODER_RESET_START stream_channels | ||
60 | int stream_channels; | ||
61 | |||
62 | int bandwidth; | ||
63 | int mode; | ||
64 | int prev_mode; | ||
65 | int frame_size; | ||
66 | int prev_redundancy; | ||
67 | |||
68 | opus_uint32 rangeFinal; | ||
69 | }; | ||
70 | |||
71 | #ifdef FIXED_POINT | ||
72 | static inline opus_int16 SAT16(opus_int32 x) { | ||
73 | return x > 32767 ? 32767 : x < -32768 ? -32768 : (opus_int16)x; | ||
74 | } | ||
75 | #endif | ||
76 | |||
77 | |||
78 | int opus_decoder_get_size(int channels) | ||
79 | { | ||
80 | int silkDecSizeBytes, celtDecSizeBytes; | ||
81 | int ret; | ||
82 | if (channels<1 || channels > 2) | ||
83 | return 0; | ||
84 | ret = silk_Get_Decoder_Size( &silkDecSizeBytes ); | ||
85 | if(ret) | ||
86 | return 0; | ||
87 | silkDecSizeBytes = align(silkDecSizeBytes); | ||
88 | celtDecSizeBytes = celt_decoder_get_size(channels); | ||
89 | return align(sizeof(OpusDecoder))+silkDecSizeBytes+celtDecSizeBytes; | ||
90 | } | ||
91 | |||
92 | int opus_decoder_init(OpusDecoder *st, opus_int32 Fs, int channels) | ||
93 | { | ||
94 | void *silk_dec; | ||
95 | CELTDecoder *celt_dec; | ||
96 | int ret, silkDecSizeBytes; | ||
97 | |||
98 | if ((Fs!=48000&&Fs!=24000&&Fs!=16000&&Fs!=12000&&Fs!=8000) | ||
99 | || (channels!=1&&channels!=2)) | ||
100 | return OPUS_BAD_ARG; | ||
101 | |||
102 | OPUS_CLEAR((char*)st, opus_decoder_get_size(channels)); | ||
103 | /* Initialize SILK encoder */ | ||
104 | ret = silk_Get_Decoder_Size(&silkDecSizeBytes); | ||
105 | if (ret) | ||
106 | return OPUS_INTERNAL_ERROR; | ||
107 | |||
108 | silkDecSizeBytes = align(silkDecSizeBytes); | ||
109 | st->silk_dec_offset = align(sizeof(OpusDecoder)); | ||
110 | st->celt_dec_offset = st->silk_dec_offset+silkDecSizeBytes; | ||
111 | silk_dec = (char*)st+st->silk_dec_offset; | ||
112 | celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); | ||
113 | st->stream_channels = st->channels = channels; | ||
114 | |||
115 | st->Fs = Fs; | ||
116 | st->DecControl.API_sampleRate = st->Fs; | ||
117 | st->DecControl.nChannelsAPI = st->channels; | ||
118 | |||
119 | /* Reset decoder */ | ||
120 | ret = silk_InitDecoder( silk_dec ); | ||
121 | if(ret)return OPUS_INTERNAL_ERROR; | ||
122 | |||
123 | /* Initialize CELT decoder */ | ||
124 | ret = celt_decoder_init(celt_dec, Fs, channels); | ||
125 | if(ret!=OPUS_OK)return OPUS_INTERNAL_ERROR; | ||
126 | |||
127 | celt_decoder_ctl(celt_dec, CELT_SET_SIGNALLING(0)); | ||
128 | |||
129 | st->prev_mode = 0; | ||
130 | st->frame_size = Fs/400; | ||
131 | return OPUS_OK; | ||
132 | } | ||
133 | |||
134 | OpusDecoder *opus_decoder_create(opus_int32 Fs, int channels, int *error) | ||
135 | { | ||
136 | int ret; | ||
137 | OpusDecoder *st; | ||
138 | if ((Fs!=48000&&Fs!=24000&&Fs!=16000&&Fs!=12000&&Fs!=8000) | ||
139 | || (channels!=1&&channels!=2)) | ||
140 | { | ||
141 | if (error) | ||
142 | *error = OPUS_BAD_ARG; | ||
143 | return NULL; | ||
144 | } | ||
145 | st = (OpusDecoder *)opus_alloc(opus_decoder_get_size(channels)); | ||
146 | if (st == NULL) | ||
147 | { | ||
148 | if (error) | ||
149 | *error = OPUS_ALLOC_FAIL; | ||
150 | return NULL; | ||
151 | } | ||
152 | ret = opus_decoder_init(st, Fs, channels); | ||
153 | if (error) | ||
154 | *error = ret; | ||
155 | if (ret != OPUS_OK) | ||
156 | { | ||
157 | opus_free(st); | ||
158 | st = NULL; | ||
159 | } | ||
160 | return st; | ||
161 | } | ||
162 | |||
163 | static void smooth_fade(const opus_val16 *in1, const opus_val16 *in2, | ||
164 | opus_val16 *out, int overlap, int channels, | ||
165 | const opus_val16 *window, opus_int32 Fs) | ||
166 | { | ||
167 | int i, c; | ||
168 | int inc = 48000/Fs; | ||
169 | for (c=0;c<channels;c++) | ||
170 | { | ||
171 | for (i=0;i<overlap;i++) | ||
172 | { | ||
173 | opus_val16 w = MULT16_16_Q15(window[i*inc], window[i*inc]); | ||
174 | out[i*channels+c] = SHR32(MAC16_16(MULT16_16(w,in2[i*channels+c]), | ||
175 | Q15ONE-w, in1[i*channels+c]), 15); | ||
176 | } | ||
177 | } | ||
178 | } | ||
179 | |||
180 | static int opus_packet_get_mode(const unsigned char *data) | ||
181 | { | ||
182 | int mode; | ||
183 | if (data[0]&0x80) | ||
184 | { | ||
185 | mode = MODE_CELT_ONLY; | ||
186 | } else if ((data[0]&0x60) == 0x60) | ||
187 | { | ||
188 | mode = MODE_HYBRID; | ||
189 | } else { | ||
190 | mode = MODE_SILK_ONLY; | ||
191 | } | ||
192 | return mode; | ||
193 | } | ||
194 | |||
195 | static int opus_decode_frame(OpusDecoder *st, const unsigned char *data, | ||
196 | opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) | ||
197 | { | ||
198 | void *silk_dec; | ||
199 | CELTDecoder *celt_dec; | ||
200 | int i, silk_ret=0, celt_ret=0; | ||
201 | ec_dec dec; | ||
202 | opus_int32 silk_frame_size; | ||
203 | VARDECL(opus_int16, pcm_silk); | ||
204 | VARDECL(opus_val16, pcm_transition); | ||
205 | VARDECL(opus_val16, redundant_audio); | ||
206 | |||
207 | int audiosize; | ||
208 | int mode; | ||
209 | int transition=0; | ||
210 | int start_band; | ||
211 | int redundancy=0; | ||
212 | int redundancy_bytes = 0; | ||
213 | int celt_to_silk=0; | ||
214 | int c; | ||
215 | int F2_5, F5, F10, F20; | ||
216 | const opus_val16 *window; | ||
217 | opus_uint32 redundant_rng = 0; | ||
218 | ALLOC_STACK; | ||
219 | |||
220 | silk_dec = (char*)st+st->silk_dec_offset; | ||
221 | celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); | ||
222 | F20 = st->Fs/50; | ||
223 | F10 = F20>>1; | ||
224 | F5 = F10>>1; | ||
225 | F2_5 = F5>>1; | ||
226 | if (frame_size < F2_5) | ||
227 | { | ||
228 | RESTORE_STACK; | ||
229 | return OPUS_BUFFER_TOO_SMALL; | ||
230 | } | ||
231 | /* Limit frame_size to avoid excessive stack allocations. */ | ||
232 | frame_size = IMIN(frame_size, st->Fs/25*3); | ||
233 | /* Payloads of 1 (2 including ToC) or 0 trigger the PLC/DTX */ | ||
234 | if (len<=1) | ||
235 | { | ||
236 | data = NULL; | ||
237 | /* In that case, don't conceal more than what the ToC says */ | ||
238 | frame_size = IMIN(frame_size, st->frame_size); | ||
239 | } | ||
240 | if (data != NULL) | ||
241 | { | ||
242 | audiosize = st->frame_size; | ||
243 | mode = st->mode; | ||
244 | ec_dec_init(&dec,(unsigned char*)data,len); | ||
245 | } else { | ||
246 | audiosize = frame_size; | ||
247 | |||
248 | if (st->prev_mode == 0) | ||
249 | { | ||
250 | /* If we haven't got any packet yet, all we can do is return zeros */ | ||
251 | for (i=0;i<audiosize*st->channels;i++) | ||
252 | pcm[i] = 0; | ||
253 | RESTORE_STACK; | ||
254 | return audiosize; | ||
255 | } else { | ||
256 | mode = st->prev_mode; | ||
257 | } | ||
258 | } | ||
259 | |||
260 | /* For CELT/hybrid PLC of more than 20 ms, do multiple calls */ | ||
261 | if (data==NULL && frame_size > F20 && mode != MODE_SILK_ONLY) | ||
262 | { | ||
263 | int nb_samples = 0; | ||
264 | do { | ||
265 | int ret = opus_decode_frame(st, NULL, 0, pcm, F20, 0); | ||
266 | if (ret != F20) | ||
267 | { | ||
268 | RESTORE_STACK; | ||
269 | return OPUS_INTERNAL_ERROR; | ||
270 | } | ||
271 | pcm += F20*st->channels; | ||
272 | nb_samples += F20; | ||
273 | } while (nb_samples < frame_size); | ||
274 | RESTORE_STACK; | ||
275 | return frame_size; | ||
276 | } | ||
277 | ALLOC(pcm_transition, F5*st->channels, opus_val16); | ||
278 | |||
279 | if (data!=NULL && st->prev_mode > 0 && ( | ||
280 | (mode == MODE_CELT_ONLY && st->prev_mode != MODE_CELT_ONLY && !st->prev_redundancy) | ||
281 | || (mode != MODE_CELT_ONLY && st->prev_mode == MODE_CELT_ONLY) ) | ||
282 | ) | ||
283 | { | ||
284 | transition = 1; | ||
285 | if (mode == MODE_CELT_ONLY) | ||
286 | opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); | ||
287 | } | ||
288 | if (audiosize > frame_size) | ||
289 | { | ||
290 | /*fprintf(stderr, "PCM buffer too small: %d vs %d (mode = %d)\n", audiosize, frame_size, mode);*/ | ||
291 | RESTORE_STACK; | ||
292 | return OPUS_BAD_ARG; | ||
293 | } else { | ||
294 | frame_size = audiosize; | ||
295 | } | ||
296 | |||
297 | ALLOC(pcm_silk, IMAX(F10, frame_size)*st->channels, opus_int16); | ||
298 | ALLOC(redundant_audio, F5*st->channels, opus_val16); | ||
299 | |||
300 | /* SILK processing */ | ||
301 | if (mode != MODE_CELT_ONLY) | ||
302 | { | ||
303 | int lost_flag, decoded_samples; | ||
304 | opus_int16 *pcm_ptr = pcm_silk; | ||
305 | |||
306 | if (st->prev_mode==MODE_CELT_ONLY) | ||
307 | silk_InitDecoder( silk_dec ); | ||
308 | |||
309 | /* The SILK PLC cannot produce frames of less than 10 ms */ | ||
310 | st->DecControl.payloadSize_ms = IMAX(10, 1000 * audiosize / st->Fs); | ||
311 | |||
312 | if (data != NULL) | ||
313 | { | ||
314 | st->DecControl.nChannelsInternal = st->stream_channels; | ||
315 | if( mode == MODE_SILK_ONLY ) { | ||
316 | if( st->bandwidth == OPUS_BANDWIDTH_NARROWBAND ) { | ||
317 | st->DecControl.internalSampleRate = 8000; | ||
318 | } else if( st->bandwidth == OPUS_BANDWIDTH_MEDIUMBAND ) { | ||
319 | st->DecControl.internalSampleRate = 12000; | ||
320 | } else if( st->bandwidth == OPUS_BANDWIDTH_WIDEBAND ) { | ||
321 | st->DecControl.internalSampleRate = 16000; | ||
322 | } else { | ||
323 | st->DecControl.internalSampleRate = 16000; | ||
324 | silk_assert( 0 ); | ||
325 | } | ||
326 | } else { | ||
327 | /* Hybrid mode */ | ||
328 | st->DecControl.internalSampleRate = 16000; | ||
329 | } | ||
330 | } | ||
331 | |||
332 | lost_flag = data == NULL ? 1 : 2 * decode_fec; | ||
333 | decoded_samples = 0; | ||
334 | do { | ||
335 | /* Call SILK decoder */ | ||
336 | int first_frame = decoded_samples == 0; | ||
337 | silk_ret = silk_Decode( silk_dec, &st->DecControl, | ||
338 | lost_flag, first_frame, &dec, pcm_ptr, &silk_frame_size ); | ||
339 | if( silk_ret ) { | ||
340 | if (lost_flag) { | ||
341 | /* PLC failure should not be fatal */ | ||
342 | silk_frame_size = frame_size; | ||
343 | for (i=0;i<frame_size*st->channels;i++) | ||
344 | pcm_ptr[i] = 0; | ||
345 | } else { | ||
346 | RESTORE_STACK; | ||
347 | return OPUS_INVALID_PACKET; | ||
348 | } | ||
349 | } | ||
350 | pcm_ptr += silk_frame_size * st->channels; | ||
351 | decoded_samples += silk_frame_size; | ||
352 | } while( decoded_samples < frame_size ); | ||
353 | } | ||
354 | |||
355 | start_band = 0; | ||
356 | if (!decode_fec && mode != MODE_CELT_ONLY && data != NULL | ||
357 | && ec_tell(&dec)+17+20*(st->mode == MODE_HYBRID) <= 8*len) | ||
358 | { | ||
359 | /* Check if we have a redundant 0-8 kHz band */ | ||
360 | if (mode == MODE_HYBRID) | ||
361 | redundancy = ec_dec_bit_logp(&dec, 12); | ||
362 | else | ||
363 | redundancy = 1; | ||
364 | if (redundancy) | ||
365 | { | ||
366 | celt_to_silk = ec_dec_bit_logp(&dec, 1); | ||
367 | /* redundancy_bytes will be at least two, in the non-hybrid | ||
368 | case due to the ec_tell() check above */ | ||
369 | redundancy_bytes = mode==MODE_HYBRID ? | ||
370 | (opus_int32)ec_dec_uint(&dec, 256)+2 : | ||
371 | len-((ec_tell(&dec)+7)>>3); | ||
372 | len -= redundancy_bytes; | ||
373 | /* This is a sanity check. It should never happen for a valid | ||
374 | packet, so the exact behaviour is not normative. */ | ||
375 | if (len*8 < ec_tell(&dec)) | ||
376 | { | ||
377 | len = 0; | ||
378 | redundancy_bytes = 0; | ||
379 | redundancy = 0; | ||
380 | } | ||
381 | /* Shrink decoder because of raw bits */ | ||
382 | dec.storage -= redundancy_bytes; | ||
383 | } | ||
384 | } | ||
385 | if (mode != MODE_CELT_ONLY) | ||
386 | start_band = 17; | ||
387 | |||
388 | { | ||
389 | int endband=21; | ||
390 | |||
391 | switch(st->bandwidth) | ||
392 | { | ||
393 | case OPUS_BANDWIDTH_NARROWBAND: | ||
394 | endband = 13; | ||
395 | break; | ||
396 | case OPUS_BANDWIDTH_MEDIUMBAND: | ||
397 | case OPUS_BANDWIDTH_WIDEBAND: | ||
398 | endband = 17; | ||
399 | break; | ||
400 | case OPUS_BANDWIDTH_SUPERWIDEBAND: | ||
401 | endband = 19; | ||
402 | break; | ||
403 | case OPUS_BANDWIDTH_FULLBAND: | ||
404 | endband = 21; | ||
405 | break; | ||
406 | } | ||
407 | celt_decoder_ctl(celt_dec, CELT_SET_END_BAND(endband)); | ||
408 | celt_decoder_ctl(celt_dec, CELT_SET_CHANNELS(st->stream_channels)); | ||
409 | } | ||
410 | |||
411 | if (redundancy) | ||
412 | transition = 0; | ||
413 | |||
414 | if (transition && mode != MODE_CELT_ONLY) | ||
415 | opus_decode_frame(st, NULL, 0, pcm_transition, IMIN(F5, audiosize), 0); | ||
416 | |||
417 | /* 5 ms redundant frame for CELT->SILK*/ | ||
418 | if (redundancy && celt_to_silk) | ||
419 | { | ||
420 | celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0)); | ||
421 | celt_decode_with_ec(celt_dec, data+len, redundancy_bytes, | ||
422 | redundant_audio, F5, NULL); | ||
423 | celt_decoder_ctl(celt_dec, OPUS_GET_FINAL_RANGE(&redundant_rng)); | ||
424 | } | ||
425 | |||
426 | /* MUST be after PLC */ | ||
427 | celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(start_band)); | ||
428 | |||
429 | if (mode != MODE_SILK_ONLY) | ||
430 | { | ||
431 | int celt_frame_size = IMIN(F20, frame_size); | ||
432 | /* Make sure to discard any previous CELT state */ | ||
433 | if (mode != st->prev_mode && st->prev_mode > 0 && !st->prev_redundancy) | ||
434 | celt_decoder_ctl(celt_dec, OPUS_RESET_STATE); | ||
435 | /* Decode CELT */ | ||
436 | celt_ret = celt_decode_with_ec(celt_dec, decode_fec ? NULL : data, | ||
437 | len, pcm, celt_frame_size, &dec); | ||
438 | } else { | ||
439 | unsigned char silence[2] = {0xFF, 0xFF}; | ||
440 | for (i=0;i<frame_size*st->channels;i++) | ||
441 | pcm[i] = 0; | ||
442 | /* For hybrid -> SILK transitions, we let the CELT MDCT | ||
443 | do a fade-out by decoding a silence frame */ | ||
444 | if (st->prev_mode == MODE_HYBRID && !(redundancy && celt_to_silk && st->prev_redundancy) ) | ||
445 | { | ||
446 | celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0)); | ||
447 | celt_decode_with_ec(celt_dec, silence, 2, pcm, F2_5, NULL); | ||
448 | } | ||
449 | } | ||
450 | |||
451 | if (mode != MODE_CELT_ONLY) | ||
452 | { | ||
453 | #ifdef FIXED_POINT | ||
454 | for (i=0;i<frame_size*st->channels;i++) | ||
455 | pcm[i] = SAT16(pcm[i] + pcm_silk[i]); | ||
456 | #else | ||
457 | for (i=0;i<frame_size*st->channels;i++) | ||
458 | pcm[i] = pcm[i] + (opus_val16)((1.f/32768.f)*pcm_silk[i]); | ||
459 | #endif | ||
460 | } | ||
461 | |||
462 | { | ||
463 | const CELTMode *celt_mode; | ||
464 | celt_decoder_ctl(celt_dec, CELT_GET_MODE(&celt_mode)); | ||
465 | window = celt_mode->window; | ||
466 | } | ||
467 | |||
468 | /* 5 ms redundant frame for SILK->CELT */ | ||
469 | if (redundancy && !celt_to_silk) | ||
470 | { | ||
471 | celt_decoder_ctl(celt_dec, OPUS_RESET_STATE); | ||
472 | celt_decoder_ctl(celt_dec, CELT_SET_START_BAND(0)); | ||
473 | |||
474 | celt_decode_with_ec(celt_dec, data+len, redundancy_bytes, redundant_audio, F5, NULL); | ||
475 | celt_decoder_ctl(celt_dec, OPUS_GET_FINAL_RANGE(&redundant_rng)); | ||
476 | smooth_fade(pcm+st->channels*(frame_size-F2_5), redundant_audio+st->channels*F2_5, | ||
477 | pcm+st->channels*(frame_size-F2_5), F2_5, st->channels, window, st->Fs); | ||
478 | } | ||
479 | if (redundancy && celt_to_silk) | ||
480 | { | ||
481 | for (c=0;c<st->channels;c++) | ||
482 | { | ||
483 | for (i=0;i<F2_5;i++) | ||
484 | pcm[st->channels*i+c] = redundant_audio[st->channels*i+c]; | ||
485 | } | ||
486 | smooth_fade(redundant_audio+st->channels*F2_5, pcm+st->channels*F2_5, | ||
487 | pcm+st->channels*F2_5, F2_5, st->channels, window, st->Fs); | ||
488 | } | ||
489 | if (transition) | ||
490 | { | ||
491 | if (audiosize >= F5) | ||
492 | { | ||
493 | for (i=0;i<st->channels*F2_5;i++) | ||
494 | pcm[i] = pcm_transition[i]; | ||
495 | smooth_fade(pcm_transition+st->channels*F2_5, pcm+st->channels*F2_5, | ||
496 | pcm+st->channels*F2_5, F2_5, | ||
497 | st->channels, window, st->Fs); | ||
498 | } else { | ||
499 | /* Not enough time to do a clean transition, but we do it anyway | ||
500 | This will not preserve amplitude perfectly and may introduce | ||
501 | a bit of temporal aliasing, but it shouldn't be too bad and | ||
502 | that's pretty much the best we can do. In any case, generating this | ||
503 | transition it pretty silly in the first place */ | ||
504 | smooth_fade(pcm_transition, pcm, | ||
505 | pcm, F2_5, | ||
506 | st->channels, window, st->Fs); | ||
507 | } | ||
508 | } | ||
509 | |||
510 | if(st->decode_gain) | ||
511 | { | ||
512 | opus_val32 gain; | ||
513 | gain = celt_exp2(MULT16_16_P15(QCONST16(6.48814081e-4f, 25), st->decode_gain)); | ||
514 | for (i=0;i<frame_size*st->channels;i++) | ||
515 | { | ||
516 | opus_val32 x; | ||
517 | x = MULT16_32_P16(pcm[i],gain); | ||
518 | pcm[i] = SATURATE(x, 32767); | ||
519 | } | ||
520 | } | ||
521 | |||
522 | if (len <= 1) | ||
523 | st->rangeFinal = 0; | ||
524 | else | ||
525 | st->rangeFinal = dec.rng ^ redundant_rng; | ||
526 | |||
527 | st->prev_mode = mode; | ||
528 | st->prev_redundancy = redundancy && !celt_to_silk; | ||
529 | RESTORE_STACK; | ||
530 | return celt_ret < 0 ? celt_ret : audiosize; | ||
531 | |||
532 | } | ||
533 | |||
534 | static int parse_size(const unsigned char *data, opus_int32 len, short *size) | ||
535 | { | ||
536 | if (len<1) | ||
537 | { | ||
538 | *size = -1; | ||
539 | return -1; | ||
540 | } else if (data[0]<252) | ||
541 | { | ||
542 | *size = data[0]; | ||
543 | return 1; | ||
544 | } else if (len<2) | ||
545 | { | ||
546 | *size = -1; | ||
547 | return -1; | ||
548 | } else { | ||
549 | *size = 4*data[1] + data[0]; | ||
550 | return 2; | ||
551 | } | ||
552 | } | ||
553 | |||
554 | static int opus_packet_parse_impl(const unsigned char *data, opus_int32 len, | ||
555 | int self_delimited, unsigned char *out_toc, | ||
556 | const unsigned char *frames[48], short size[48], int *payload_offset) | ||
557 | { | ||
558 | int i, bytes; | ||
559 | int count; | ||
560 | int cbr; | ||
561 | unsigned char ch, toc; | ||
562 | int framesize; | ||
563 | int last_size; | ||
564 | const unsigned char *data0 = data; | ||
565 | |||
566 | if (size==NULL) | ||
567 | return OPUS_BAD_ARG; | ||
568 | |||
569 | framesize = opus_packet_get_samples_per_frame(data, 48000); | ||
570 | |||
571 | cbr = 0; | ||
572 | toc = *data++; | ||
573 | len--; | ||
574 | last_size = len; | ||
575 | switch (toc&0x3) | ||
576 | { | ||
577 | /* One frame */ | ||
578 | case 0: | ||
579 | count=1; | ||
580 | break; | ||
581 | /* Two CBR frames */ | ||
582 | case 1: | ||
583 | count=2; | ||
584 | cbr = 1; | ||
585 | if (!self_delimited) | ||
586 | { | ||
587 | if (len&0x1) | ||
588 | return OPUS_INVALID_PACKET; | ||
589 | size[0] = last_size = len/2; | ||
590 | } | ||
591 | break; | ||
592 | /* Two VBR frames */ | ||
593 | case 2: | ||
594 | count = 2; | ||
595 | bytes = parse_size(data, len, size); | ||
596 | len -= bytes; | ||
597 | if (size[0]<0 || size[0] > len) | ||
598 | return OPUS_INVALID_PACKET; | ||
599 | data += bytes; | ||
600 | last_size = len-size[0]; | ||
601 | break; | ||
602 | /* Multiple CBR/VBR frames (from 0 to 120 ms) */ | ||
603 | default: /*case 3:*/ | ||
604 | if (len<1) | ||
605 | return OPUS_INVALID_PACKET; | ||
606 | /* Number of frames encoded in bits 0 to 5 */ | ||
607 | ch = *data++; | ||
608 | count = ch&0x3F; | ||
609 | if (count <= 0 || framesize*count > 5760) | ||
610 | return OPUS_INVALID_PACKET; | ||
611 | len--; | ||
612 | /* Padding flag is bit 6 */ | ||
613 | if (ch&0x40) | ||
614 | { | ||
615 | int padding=0; | ||
616 | int p; | ||
617 | do { | ||
618 | if (len<=0) | ||
619 | return OPUS_INVALID_PACKET; | ||
620 | p = *data++; | ||
621 | len--; | ||
622 | padding += p==255 ? 254: p; | ||
623 | } while (p==255); | ||
624 | len -= padding; | ||
625 | } | ||
626 | if (len<0) | ||
627 | return OPUS_INVALID_PACKET; | ||
628 | /* VBR flag is bit 7 */ | ||
629 | cbr = !(ch&0x80); | ||
630 | if (!cbr) | ||
631 | { | ||
632 | /* VBR case */ | ||
633 | last_size = len; | ||
634 | for (i=0;i<count-1;i++) | ||
635 | { | ||
636 | bytes = parse_size(data, len, size+i); | ||
637 | len -= bytes; | ||
638 | if (size[i]<0 || size[i] > len) | ||
639 | return OPUS_INVALID_PACKET; | ||
640 | data += bytes; | ||
641 | last_size -= bytes+size[i]; | ||
642 | } | ||
643 | if (last_size<0) | ||
644 | return OPUS_INVALID_PACKET; | ||
645 | } else if (!self_delimited) | ||
646 | { | ||
647 | /* CBR case */ | ||
648 | last_size = len/count; | ||
649 | if (last_size*count!=len) | ||
650 | return OPUS_INVALID_PACKET; | ||
651 | for (i=0;i<count-1;i++) | ||
652 | size[i] = last_size; | ||
653 | } | ||
654 | break; | ||
655 | } | ||
656 | /* Self-delimited framing has an extra size for the last frame. */ | ||
657 | if (self_delimited) | ||
658 | { | ||
659 | bytes = parse_size(data, len, size+count-1); | ||
660 | len -= bytes; | ||
661 | if (size[count-1]<0 || size[count-1] > len) | ||
662 | return OPUS_INVALID_PACKET; | ||
663 | data += bytes; | ||
664 | /* For CBR packets, apply the size to all the frames. */ | ||
665 | if (cbr) | ||
666 | { | ||
667 | if (size[count-1]*count > len) | ||
668 | return OPUS_INVALID_PACKET; | ||
669 | for (i=0;i<count-1;i++) | ||
670 | size[i] = size[count-1]; | ||
671 | } else if(size[count-1] > last_size) | ||
672 | return OPUS_INVALID_PACKET; | ||
673 | } else | ||
674 | { | ||
675 | /* Because it's not encoded explicitly, it's possible the size of the | ||
676 | last packet (or all the packets, for the CBR case) is larger than | ||
677 | 1275. Reject them here.*/ | ||
678 | if (last_size > 1275) | ||
679 | return OPUS_INVALID_PACKET; | ||
680 | size[count-1] = last_size; | ||
681 | } | ||
682 | |||
683 | if (frames) | ||
684 | { | ||
685 | for (i=0;i<count;i++) | ||
686 | { | ||
687 | frames[i] = data; | ||
688 | data += size[i]; | ||
689 | } | ||
690 | } | ||
691 | |||
692 | if (out_toc) | ||
693 | *out_toc = toc; | ||
694 | |||
695 | if (payload_offset) | ||
696 | *payload_offset = data-data0; | ||
697 | |||
698 | return count; | ||
699 | } | ||
700 | |||
701 | int opus_packet_parse(const unsigned char *data, opus_int32 len, | ||
702 | unsigned char *out_toc, const unsigned char *frames[48], | ||
703 | short size[48], int *payload_offset) | ||
704 | { | ||
705 | return opus_packet_parse_impl(data, len, 0, out_toc, | ||
706 | frames, size, payload_offset); | ||
707 | } | ||
708 | |||
709 | int opus_decode_native(OpusDecoder *st, const unsigned char *data, | ||
710 | opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec, | ||
711 | int self_delimited, int *packet_offset) | ||
712 | { | ||
713 | int i, nb_samples; | ||
714 | int count, offset; | ||
715 | unsigned char toc; | ||
716 | int tot_offset; | ||
717 | /* 48 x 2.5 ms = 120 ms */ | ||
718 | short size[48]; | ||
719 | if (decode_fec<0 || decode_fec>1) | ||
720 | return OPUS_BAD_ARG; | ||
721 | if (len==0 || data==NULL) | ||
722 | return opus_decode_frame(st, NULL, 0, pcm, frame_size, 0); | ||
723 | else if (len<0) | ||
724 | return OPUS_BAD_ARG; | ||
725 | |||
726 | tot_offset = 0; | ||
727 | st->mode = opus_packet_get_mode(data); | ||
728 | st->bandwidth = opus_packet_get_bandwidth(data); | ||
729 | st->frame_size = opus_packet_get_samples_per_frame(data, st->Fs); | ||
730 | st->stream_channels = opus_packet_get_nb_channels(data); | ||
731 | |||
732 | count = opus_packet_parse_impl(data, len, self_delimited, &toc, NULL, size, &offset); | ||
733 | if (count < 0) | ||
734 | return count; | ||
735 | |||
736 | data += offset; | ||
737 | tot_offset += offset; | ||
738 | |||
739 | if (count*st->frame_size > frame_size) | ||
740 | return OPUS_BUFFER_TOO_SMALL; | ||
741 | nb_samples=0; | ||
742 | for (i=0;i<count;i++) | ||
743 | { | ||
744 | int ret; | ||
745 | ret = opus_decode_frame(st, data, size[i], pcm, frame_size-nb_samples, decode_fec); | ||
746 | if (ret<0) | ||
747 | return ret; | ||
748 | data += size[i]; | ||
749 | tot_offset += size[i]; | ||
750 | pcm += ret*st->channels; | ||
751 | nb_samples += ret; | ||
752 | } | ||
753 | if (packet_offset != NULL) | ||
754 | *packet_offset = tot_offset; | ||
755 | return nb_samples; | ||
756 | } | ||
757 | |||
758 | #ifdef FIXED_POINT | ||
759 | |||
760 | int opus_decode(OpusDecoder *st, const unsigned char *data, | ||
761 | opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) | ||
762 | { | ||
763 | return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL); | ||
764 | } | ||
765 | |||
766 | #ifndef DISABLE_FLOAT_API | ||
767 | int opus_decode_float(OpusDecoder *st, const unsigned char *data, | ||
768 | opus_int32 len, float *pcm, int frame_size, int decode_fec) | ||
769 | { | ||
770 | VARDECL(opus_int16, out); | ||
771 | int ret, i; | ||
772 | ALLOC_STACK; | ||
773 | |||
774 | ALLOC(out, frame_size*st->channels, opus_int16); | ||
775 | |||
776 | ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL); | ||
777 | if (ret > 0) | ||
778 | { | ||
779 | for (i=0;i<ret*st->channels;i++) | ||
780 | pcm[i] = (1.f/32768.f)*(out[i]); | ||
781 | } | ||
782 | RESTORE_STACK; | ||
783 | return ret; | ||
784 | } | ||
785 | #endif | ||
786 | |||
787 | |||
788 | #else | ||
789 | int opus_decode(OpusDecoder *st, const unsigned char *data, | ||
790 | opus_int32 len, opus_int16 *pcm, int frame_size, int decode_fec) | ||
791 | { | ||
792 | VARDECL(float, out); | ||
793 | int ret, i; | ||
794 | ALLOC_STACK; | ||
795 | |||
796 | if(frame_size<0) | ||
797 | { | ||
798 | RESTORE_STACK; | ||
799 | return OPUS_BAD_ARG; | ||
800 | } | ||
801 | |||
802 | ALLOC(out, frame_size*st->channels, float); | ||
803 | |||
804 | ret = opus_decode_native(st, data, len, out, frame_size, decode_fec, 0, NULL); | ||
805 | if (ret > 0) | ||
806 | { | ||
807 | for (i=0;i<ret*st->channels;i++) | ||
808 | pcm[i] = FLOAT2INT16(out[i]); | ||
809 | } | ||
810 | RESTORE_STACK; | ||
811 | return ret; | ||
812 | } | ||
813 | |||
814 | int opus_decode_float(OpusDecoder *st, const unsigned char *data, | ||
815 | opus_int32 len, opus_val16 *pcm, int frame_size, int decode_fec) | ||
816 | { | ||
817 | return opus_decode_native(st, data, len, pcm, frame_size, decode_fec, 0, NULL); | ||
818 | } | ||
819 | |||
820 | #endif | ||
821 | |||
822 | int opus_decoder_ctl(OpusDecoder *st, int request, ...) | ||
823 | { | ||
824 | int ret = OPUS_OK; | ||
825 | va_list ap; | ||
826 | void *silk_dec; | ||
827 | CELTDecoder *celt_dec; | ||
828 | |||
829 | silk_dec = (char*)st+st->silk_dec_offset; | ||
830 | celt_dec = (CELTDecoder*)((char*)st+st->celt_dec_offset); | ||
831 | |||
832 | |||
833 | va_start(ap, request); | ||
834 | |||
835 | switch (request) | ||
836 | { | ||
837 | case OPUS_GET_BANDWIDTH_REQUEST: | ||
838 | { | ||
839 | opus_int32 *value = va_arg(ap, opus_int32*); | ||
840 | *value = st->bandwidth; | ||
841 | } | ||
842 | break; | ||
843 | case OPUS_GET_FINAL_RANGE_REQUEST: | ||
844 | { | ||
845 | opus_uint32 *value = va_arg(ap, opus_uint32*); | ||
846 | *value = st->rangeFinal; | ||
847 | } | ||
848 | break; | ||
849 | case OPUS_RESET_STATE: | ||
850 | { | ||
851 | OPUS_CLEAR((char*)&st->OPUS_DECODER_RESET_START, | ||
852 | sizeof(OpusDecoder)- | ||
853 | ((char*)&st->OPUS_DECODER_RESET_START - (char*)st)); | ||
854 | |||
855 | celt_decoder_ctl(celt_dec, OPUS_RESET_STATE); | ||
856 | silk_InitDecoder( silk_dec ); | ||
857 | st->stream_channels = st->channels; | ||
858 | st->frame_size = st->Fs/400; | ||
859 | } | ||
860 | break; | ||
861 | case OPUS_GET_SAMPLE_RATE_REQUEST: | ||
862 | { | ||
863 | opus_int32 *value = va_arg(ap, opus_int32*); | ||
864 | if (value==NULL) | ||
865 | { | ||
866 | ret = OPUS_BAD_ARG; | ||
867 | break; | ||
868 | } | ||
869 | *value = st->Fs; | ||
870 | } | ||
871 | break; | ||
872 | case OPUS_GET_PITCH_REQUEST: | ||
873 | { | ||
874 | opus_int32 *value = va_arg(ap, opus_int32*); | ||
875 | if (value==NULL) | ||
876 | { | ||
877 | ret = OPUS_BAD_ARG; | ||
878 | break; | ||
879 | } | ||
880 | if (st->prev_mode == MODE_CELT_ONLY) | ||
881 | celt_decoder_ctl(celt_dec, OPUS_GET_PITCH(value)); | ||
882 | else | ||
883 | *value = st->DecControl.prevPitchLag; | ||
884 | } | ||
885 | break; | ||
886 | case OPUS_GET_GAIN_REQUEST: | ||
887 | { | ||
888 | opus_int32 *value = va_arg(ap, opus_int32*); | ||
889 | if (value==NULL) | ||
890 | { | ||
891 | ret = OPUS_BAD_ARG; | ||
892 | break; | ||
893 | } | ||
894 | *value = st->decode_gain; | ||
895 | } | ||
896 | break; | ||
897 | case OPUS_SET_GAIN_REQUEST: | ||
898 | { | ||
899 | opus_int32 value = va_arg(ap, opus_int32); | ||
900 | if (value<-32768 || value>32767) | ||
901 | { | ||
902 | ret = OPUS_BAD_ARG; | ||
903 | break; | ||
904 | } | ||
905 | st->decode_gain = value; | ||
906 | } | ||
907 | break; | ||
908 | default: | ||
909 | /*fprintf(stderr, "unknown opus_decoder_ctl() request: %d", request);*/ | ||
910 | ret = OPUS_UNIMPLEMENTED; | ||
911 | break; | ||
912 | } | ||
913 | |||
914 | va_end(ap); | ||
915 | return ret; | ||
916 | } | ||
917 | |||
918 | void opus_decoder_destroy(OpusDecoder *st) | ||
919 | { | ||
920 | opus_free(st); | ||
921 | } | ||
922 | |||
923 | |||
924 | int opus_packet_get_bandwidth(const unsigned char *data) | ||
925 | { | ||
926 | int bandwidth; | ||
927 | if (data[0]&0x80) | ||
928 | { | ||
929 | bandwidth = OPUS_BANDWIDTH_MEDIUMBAND + ((data[0]>>5)&0x3); | ||
930 | if (bandwidth == OPUS_BANDWIDTH_MEDIUMBAND) | ||
931 | bandwidth = OPUS_BANDWIDTH_NARROWBAND; | ||
932 | } else if ((data[0]&0x60) == 0x60) | ||
933 | { | ||
934 | bandwidth = (data[0]&0x10) ? OPUS_BANDWIDTH_FULLBAND : | ||
935 | OPUS_BANDWIDTH_SUPERWIDEBAND; | ||
936 | } else { | ||
937 | bandwidth = OPUS_BANDWIDTH_NARROWBAND + ((data[0]>>5)&0x3); | ||
938 | } | ||
939 | return bandwidth; | ||
940 | } | ||
941 | |||
942 | int opus_packet_get_samples_per_frame(const unsigned char *data, | ||
943 | opus_int32 Fs) | ||
944 | { | ||
945 | int audiosize; | ||
946 | if (data[0]&0x80) | ||
947 | { | ||
948 | audiosize = ((data[0]>>3)&0x3); | ||
949 | audiosize = (Fs<<audiosize)/400; | ||
950 | } else if ((data[0]&0x60) == 0x60) | ||
951 | { | ||
952 | audiosize = (data[0]&0x08) ? Fs/50 : Fs/100; | ||
953 | } else { | ||
954 | audiosize = ((data[0]>>3)&0x3); | ||
955 | if (audiosize == 3) | ||
956 | audiosize = Fs*60/1000; | ||
957 | else | ||
958 | audiosize = (Fs<<audiosize)/100; | ||
959 | } | ||
960 | return audiosize; | ||
961 | } | ||
962 | |||
963 | int opus_packet_get_nb_channels(const unsigned char *data) | ||
964 | { | ||
965 | return (data[0]&0x4) ? 2 : 1; | ||
966 | } | ||
967 | |||
968 | int opus_packet_get_nb_frames(const unsigned char packet[], opus_int32 len) | ||
969 | { | ||
970 | int count; | ||
971 | if (len<1) | ||
972 | return OPUS_BAD_ARG; | ||
973 | count = packet[0]&0x3; | ||
974 | if (count==0) | ||
975 | return 1; | ||
976 | else if (count!=3) | ||
977 | return 2; | ||
978 | else if (len<2) | ||
979 | return OPUS_INVALID_PACKET; | ||
980 | else | ||
981 | return packet[1]&0x3F; | ||
982 | } | ||
983 | |||
984 | int opus_decoder_get_nb_samples(const OpusDecoder *dec, | ||
985 | const unsigned char packet[], opus_int32 len) | ||
986 | { | ||
987 | int samples; | ||
988 | int count = opus_packet_get_nb_frames(packet, len); | ||
989 | |||
990 | if (count<0) | ||
991 | return count; | ||
992 | |||
993 | samples = count*opus_packet_get_samples_per_frame(packet, dec->Fs); | ||
994 | /* Can't have more than 120 ms */ | ||
995 | if (samples*25 > dec->Fs*3) | ||
996 | return OPUS_INVALID_PACKET; | ||
997 | else | ||
998 | return samples; | ||
999 | } | ||