00001 /* 00002 * Copyright (C) 2011 Michael Niedermayer (michaelni@gmx.at) 00003 * 00004 * This file is part of libswresample 00005 * 00006 * libswresample is free software; you can redistribute it and/or 00007 * modify it under the terms of the GNU Lesser General Public 00008 * License as published by the Free Software Foundation; either 00009 * version 2.1 of the License, or (at your option) any later version. 00010 * 00011 * libswresample is distributed in the hope that it will be useful, 00012 * but WITHOUT ANY WARRANTY; without even the implied warranty of 00013 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 00014 * Lesser General Public License for more details. 00015 * 00016 * You should have received a copy of the GNU Lesser General Public 00017 * License along with libswresample; if not, write to the Free Software 00018 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 00019 */ 00020 00021 #include "libavutil/opt.h" 00022 #include "swresample_internal.h" 00023 #include "audioconvert.h" 00024 #include "libavutil/avassert.h" 00025 #include "libavutil/audioconvert.h" 00026 00027 #define C30DB M_SQRT2 00028 #define C15DB 1.189207115 00029 #define C__0DB 1.0 00030 #define C_15DB 0.840896415 00031 #define C_30DB M_SQRT1_2 00032 #define C_45DB 0.594603558 00033 #define C_60DB 0.5 00034 00035 00036 //TODO split options array out? 00037 #define OFFSET(x) offsetof(SwrContext,x) 00038 static const AVOption options[]={ 00039 {"ich", "input channel count", OFFSET( in.ch_count ), AV_OPT_TYPE_INT, {.dbl=2}, 0, SWR_CH_MAX, 0}, 00040 {"och", "output channel count", OFFSET(out.ch_count ), AV_OPT_TYPE_INT, {.dbl=2}, 0, SWR_CH_MAX, 0}, 00041 {"uch", "used channel count", OFFSET(used_ch_count ), AV_OPT_TYPE_INT, {.dbl=0}, 0, SWR_CH_MAX, 0}, 00042 {"isr", "input sample rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT, {.dbl=48000}, 1, INT_MAX, 0}, 00043 {"osr", "output sample rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT, {.dbl=48000}, 1, INT_MAX, 0}, 00044 //{"ip" , "input planar" , OFFSET( in.planar ), AV_OPT_TYPE_INT, {.dbl=0}, 0, 1, 0}, 00045 //{"op" , "output planar" , OFFSET(out.planar ), AV_OPT_TYPE_INT, {.dbl=0}, 0, 1, 0}, 00046 {"isf", "input sample format", OFFSET( in_sample_fmt ), AV_OPT_TYPE_INT, {.dbl=AV_SAMPLE_FMT_S16}, 0, AV_SAMPLE_FMT_NB-1+256, 0}, 00047 {"osf", "output sample format", OFFSET(out_sample_fmt ), AV_OPT_TYPE_INT, {.dbl=AV_SAMPLE_FMT_S16}, 0, AV_SAMPLE_FMT_NB-1+256, 0}, 00048 {"tsf", "internal sample format", OFFSET(int_sample_fmt ), AV_OPT_TYPE_INT, {.dbl=AV_SAMPLE_FMT_NONE}, -1, AV_SAMPLE_FMT_FLT, 0}, 00049 {"icl", "input channel layout" , OFFSET( in_ch_layout), AV_OPT_TYPE_INT64, {.dbl=0}, 0, INT64_MAX, 0, "channel_layout"}, 00050 {"ocl", "output channel layout", OFFSET(out_ch_layout), AV_OPT_TYPE_INT64, {.dbl=0}, 0, INT64_MAX, 0, "channel_layout"}, 00051 {"clev", "center mix level" , OFFSET(clev) , AV_OPT_TYPE_FLOAT, {.dbl=C_30DB}, 0, 4, 0}, 00052 {"slev", "sourround mix level" , OFFSET(slev) , AV_OPT_TYPE_FLOAT, {.dbl=C_30DB}, 0, 4, 0}, 00053 {"rmvol", "rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0}, -1000, 1000, 0}, 00054 {"flags", NULL , OFFSET(flags) , AV_OPT_TYPE_FLAGS, {.dbl=0}, 0, UINT_MAX, 0, "flags"}, 00055 {"res", "force resampling", 0, AV_OPT_TYPE_CONST, {.dbl=SWR_FLAG_RESAMPLE}, INT_MIN, INT_MAX, 0, "flags"}, 00056 00057 {0} 00058 }; 00059 00060 static const char* context_to_name(void* ptr) { 00061 return "SWR"; 00062 } 00063 00064 static const AVClass av_class = { 00065 .class_name = "SwrContext", 00066 .item_name = context_to_name, 00067 .option = options, 00068 .version = LIBAVUTIL_VERSION_INT, 00069 .log_level_offset_offset = OFFSET(log_level_offset), 00070 .parent_log_context_offset = OFFSET(log_ctx), 00071 }; 00072 00073 unsigned swresample_version(void) 00074 { 00075 av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100); 00076 return LIBSWRESAMPLE_VERSION_INT; 00077 } 00078 00079 const char *swresample_configuration(void) 00080 { 00081 return FFMPEG_CONFIGURATION; 00082 } 00083 00084 const char *swresample_license(void) 00085 { 00086 #define LICENSE_PREFIX "libswresample license: " 00087 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1; 00088 } 00089 00090 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){ 00091 if(!s || s->in_convert) // s needs to be allocated but not initialized 00092 return AVERROR(EINVAL); 00093 s->channel_map = channel_map; 00094 return 0; 00095 } 00096 00097 struct SwrContext *swr_alloc(void){ 00098 SwrContext *s= av_mallocz(sizeof(SwrContext)); 00099 if(s){ 00100 s->av_class= &av_class; 00101 av_opt_set_defaults(s); 00102 } 00103 return s; 00104 } 00105 00106 struct SwrContext *swr_alloc_set_opts(struct SwrContext *s, 00107 int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate, 00108 int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate, 00109 int log_offset, void *log_ctx){ 00110 if(!s) s= swr_alloc(); 00111 if(!s) return NULL; 00112 00113 s->log_level_offset= log_offset; 00114 s->log_ctx= log_ctx; 00115 00116 av_opt_set_int(s, "ocl", out_ch_layout, 0); 00117 av_opt_set_int(s, "osf", out_sample_fmt, 0); 00118 av_opt_set_int(s, "osr", out_sample_rate, 0); 00119 av_opt_set_int(s, "icl", in_ch_layout, 0); 00120 av_opt_set_int(s, "isf", in_sample_fmt, 0); 00121 av_opt_set_int(s, "isr", in_sample_rate, 0); 00122 av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_S16, 0); 00123 av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0); 00124 av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0); 00125 av_opt_set_int(s, "uch", 0, 0); 00126 return s; 00127 } 00128 00129 00130 static void free_temp(AudioData *a){ 00131 av_free(a->data); 00132 memset(a, 0, sizeof(*a)); 00133 } 00134 00135 void swr_free(SwrContext **ss){ 00136 SwrContext *s= *ss; 00137 if(s){ 00138 free_temp(&s->postin); 00139 free_temp(&s->midbuf); 00140 free_temp(&s->preout); 00141 free_temp(&s->in_buffer); 00142 swri_audio_convert_free(&s-> in_convert); 00143 swri_audio_convert_free(&s->out_convert); 00144 swri_audio_convert_free(&s->full_convert); 00145 swri_resample_free(&s->resample); 00146 } 00147 00148 av_freep(ss); 00149 } 00150 00151 int swr_init(struct SwrContext *s){ 00152 s->in_buffer_index= 0; 00153 s->in_buffer_count= 0; 00154 s->resample_in_constraint= 0; 00155 free_temp(&s->postin); 00156 free_temp(&s->midbuf); 00157 free_temp(&s->preout); 00158 free_temp(&s->in_buffer); 00159 swri_audio_convert_free(&s-> in_convert); 00160 swri_audio_convert_free(&s->out_convert); 00161 swri_audio_convert_free(&s->full_convert); 00162 00163 s-> in.planar= av_sample_fmt_is_planar(s-> in_sample_fmt); 00164 s->out.planar= av_sample_fmt_is_planar(s->out_sample_fmt); 00165 s-> in_sample_fmt= av_get_alt_sample_fmt(s-> in_sample_fmt, 0); 00166 s->out_sample_fmt= av_get_alt_sample_fmt(s->out_sample_fmt, 0); 00167 00168 if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){ 00169 av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt); 00170 return AVERROR(EINVAL); 00171 } 00172 if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){ 00173 av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt); 00174 return AVERROR(EINVAL); 00175 } 00176 00177 if( s->int_sample_fmt != AV_SAMPLE_FMT_S16 00178 &&s->int_sample_fmt != AV_SAMPLE_FMT_FLT){ 00179 av_log(s, AV_LOG_ERROR, "Requested sample format %s is not supported internally, only float & S16 is supported\n", av_get_sample_fmt_name(s->int_sample_fmt)); 00180 return AVERROR(EINVAL); 00181 } 00182 00183 //FIXME should we allow/support using FLT on material that doesnt need it ? 00184 if(s->in_sample_fmt <= AV_SAMPLE_FMT_S16 || s->int_sample_fmt==AV_SAMPLE_FMT_S16){ 00185 s->int_sample_fmt= AV_SAMPLE_FMT_S16; 00186 }else 00187 s->int_sample_fmt= AV_SAMPLE_FMT_FLT; 00188 00189 00190 if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){ 00191 s->resample = swri_resample_init(s->resample, s->out_sample_rate, s->in_sample_rate, 16, 10, 0, 0.8); 00192 }else 00193 swri_resample_free(&s->resample); 00194 if(s->int_sample_fmt != AV_SAMPLE_FMT_S16 && s->resample){ 00195 av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16 currently\n"); //FIXME 00196 return -1; 00197 } 00198 00199 if(!s->used_ch_count) 00200 s->used_ch_count= s->in.ch_count; 00201 00202 if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){ 00203 av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n"); 00204 s-> in_ch_layout= 0; 00205 } 00206 00207 if(!s-> in_ch_layout) 00208 s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count); 00209 if(!s->out_ch_layout) 00210 s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count); 00211 00212 s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0; 00213 00214 #define RSC 1 //FIXME finetune 00215 if(!s-> in.ch_count) 00216 s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout); 00217 if(!s->used_ch_count) 00218 s->used_ch_count= s->in.ch_count; 00219 if(!s->out.ch_count) 00220 s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout); 00221 00222 if(!s-> in.ch_count){ 00223 av_assert0(!s->in_ch_layout); 00224 av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n"); 00225 return -1; 00226 } 00227 00228 av_assert0(s->used_ch_count); 00229 av_assert0(s->out.ch_count); 00230 s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0; 00231 00232 s-> in.bps= av_get_bytes_per_sample(s-> in_sample_fmt); 00233 s->int_bps= av_get_bytes_per_sample(s->int_sample_fmt); 00234 s->out.bps= av_get_bytes_per_sample(s->out_sample_fmt); 00235 00236 if(!s->resample && !s->rematrix && !s->channel_map){ 00237 s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt, 00238 s-> in_sample_fmt, s-> in.ch_count, NULL, 0); 00239 return 0; 00240 } 00241 00242 s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt, 00243 s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0); 00244 s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt, 00245 s->int_sample_fmt, s->out.ch_count, NULL, 0); 00246 00247 00248 s->postin= s->in; 00249 s->preout= s->out; 00250 s->midbuf= s->in; 00251 s->in_buffer= s->in; 00252 if(s->channel_map){ 00253 s->postin.ch_count= 00254 s->midbuf.ch_count= 00255 s->in_buffer.ch_count= s->used_ch_count; 00256 } 00257 if(!s->resample_first){ 00258 s->midbuf.ch_count= s->out.ch_count; 00259 s->in_buffer.ch_count = s->out.ch_count; 00260 } 00261 00262 s->in_buffer.bps = s->postin.bps = s->midbuf.bps = s->preout.bps = s->int_bps; 00263 s->in_buffer.planar = s->postin.planar = s->midbuf.planar = s->preout.planar = 1; 00264 00265 00266 if(s->rematrix) 00267 return swri_rematrix_init(s); 00268 00269 return 0; 00270 } 00271 00272 static int realloc_audio(AudioData *a, int count){ 00273 int i, countb; 00274 AudioData old; 00275 00276 if(a->count >= count) 00277 return 0; 00278 00279 count*=2; 00280 00281 countb= FFALIGN(count*a->bps, 32); 00282 old= *a; 00283 00284 av_assert0(a->planar); 00285 av_assert0(a->bps); 00286 av_assert0(a->ch_count); 00287 00288 a->data= av_malloc(countb*a->ch_count); 00289 if(!a->data) 00290 return AVERROR(ENOMEM); 00291 for(i=0; i<a->ch_count; i++){ 00292 a->ch[i]= a->data + i*(a->planar ? countb : a->bps); 00293 if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps); 00294 } 00295 av_free(old.data); 00296 a->count= count; 00297 00298 return 1; 00299 } 00300 00301 static void copy(AudioData *out, AudioData *in, 00302 int count){ 00303 av_assert0(out->planar == in->planar); 00304 av_assert0(out->bps == in->bps); 00305 av_assert0(out->ch_count == in->ch_count); 00306 if(out->planar){ 00307 int ch; 00308 for(ch=0; ch<out->ch_count; ch++) 00309 memcpy(out->ch[ch], in->ch[ch], count*out->bps); 00310 }else 00311 memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps); 00312 } 00313 00314 static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){ 00315 int i; 00316 if(out->planar){ 00317 for(i=0; i<out->ch_count; i++) 00318 out->ch[i]= in_arg[i]; 00319 }else{ 00320 for(i=0; i<out->ch_count; i++) 00321 out->ch[i]= in_arg[0] + i*out->bps; 00322 } 00323 } 00324 00329 static void buf_set(AudioData *out, AudioData *in, int count){ 00330 if(in->planar){ 00331 int ch; 00332 for(ch=0; ch<out->ch_count; ch++) 00333 out->ch[ch]= in->ch[ch] + count*out->bps; 00334 }else 00335 out->ch[0]= in->ch[0] + count*out->ch_count*out->bps; 00336 } 00337 00342 static int resample(SwrContext *s, AudioData *out_param, int out_count, 00343 const AudioData * in_param, int in_count){ 00344 AudioData in, out, tmp; 00345 int ret_sum=0; 00346 int border=0; 00347 00348 tmp=out=*out_param; 00349 in = *in_param; 00350 00351 do{ 00352 int ret, size, consumed; 00353 if(!s->resample_in_constraint && s->in_buffer_count){ 00354 buf_set(&tmp, &s->in_buffer, s->in_buffer_index); 00355 ret= swri_multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed); 00356 out_count -= ret; 00357 ret_sum += ret; 00358 buf_set(&out, &out, ret); 00359 s->in_buffer_count -= consumed; 00360 s->in_buffer_index += consumed; 00361 00362 if(!in_count) 00363 break; 00364 if(s->in_buffer_count <= border){ 00365 buf_set(&in, &in, -s->in_buffer_count); 00366 in_count += s->in_buffer_count; 00367 s->in_buffer_count=0; 00368 s->in_buffer_index=0; 00369 border = 0; 00370 } 00371 } 00372 00373 if(in_count && !s->in_buffer_count){ 00374 s->in_buffer_index=0; 00375 ret= swri_multiple_resample(s->resample, &out, out_count, &in, in_count, &consumed); 00376 out_count -= ret; 00377 ret_sum += ret; 00378 buf_set(&out, &out, ret); 00379 in_count -= consumed; 00380 buf_set(&in, &in, consumed); 00381 } 00382 00383 //TODO is this check sane considering the advanced copy avoidance below 00384 size= s->in_buffer_index + s->in_buffer_count + in_count; 00385 if( size > s->in_buffer.count 00386 && s->in_buffer_count + in_count <= s->in_buffer_index){ 00387 buf_set(&tmp, &s->in_buffer, s->in_buffer_index); 00388 copy(&s->in_buffer, &tmp, s->in_buffer_count); 00389 s->in_buffer_index=0; 00390 }else 00391 if((ret=realloc_audio(&s->in_buffer, size)) < 0) 00392 return ret; 00393 00394 if(in_count){ 00395 int count= in_count; 00396 if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2; 00397 00398 buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count); 00399 copy(&tmp, &in, /*in_*/count); 00400 s->in_buffer_count += count; 00401 in_count -= count; 00402 border += count; 00403 buf_set(&in, &in, count); 00404 s->resample_in_constraint= 0; 00405 if(s->in_buffer_count != count || in_count) 00406 continue; 00407 } 00408 break; 00409 }while(1); 00410 00411 s->resample_in_constraint= !!out_count; 00412 00413 return ret_sum; 00414 } 00415 00416 int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count, 00417 const uint8_t *in_arg [SWR_CH_MAX], int in_count){ 00418 AudioData *postin, *midbuf, *preout; 00419 int ret/*, in_max*/; 00420 AudioData * in= &s->in; 00421 AudioData *out= &s->out; 00422 AudioData preout_tmp, midbuf_tmp; 00423 00424 if(!s->resample){ 00425 if(in_count > out_count) 00426 return -1; 00427 out_count = in_count; 00428 } 00429 00430 if(!in_arg){ 00431 if(s->in_buffer_count){ 00432 AudioData *a= &s->in_buffer; 00433 int i, j, ret; 00434 if((ret=realloc_audio(a, s->in_buffer_index + 2*s->in_buffer_count)) < 0) 00435 return ret; 00436 av_assert0(a->planar); 00437 for(i=0; i<a->ch_count; i++){ 00438 for(j=0; j<s->in_buffer_count; j++){ 00439 memcpy(a->ch[i] + (s->in_buffer_index+s->in_buffer_count+j )*a->bps, 00440 a->ch[i] + (s->in_buffer_index+s->in_buffer_count-j-1)*a->bps, a->bps); 00441 } 00442 } 00443 s->in_buffer_count += (s->in_buffer_count+1)/2; 00444 s->resample_in_constraint = 0; 00445 }else{ 00446 return 0; 00447 } 00448 }else 00449 fill_audiodata(in , (void*)in_arg); 00450 fill_audiodata(out, out_arg); 00451 00452 if(s->full_convert){ 00453 av_assert0(!s->resample); 00454 swri_audio_convert(s->full_convert, out, in, in_count); 00455 return out_count; 00456 } 00457 00458 // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps; 00459 // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count); 00460 00461 if((ret=realloc_audio(&s->postin, in_count))<0) 00462 return ret; 00463 if(s->resample_first){ 00464 av_assert0(s->midbuf.ch_count == s->used_ch_count); 00465 if((ret=realloc_audio(&s->midbuf, out_count))<0) 00466 return ret; 00467 }else{ 00468 av_assert0(s->midbuf.ch_count == s->out.ch_count); 00469 if((ret=realloc_audio(&s->midbuf, in_count))<0) 00470 return ret; 00471 } 00472 if((ret=realloc_audio(&s->preout, out_count))<0) 00473 return ret; 00474 00475 postin= &s->postin; 00476 00477 midbuf_tmp= s->midbuf; 00478 midbuf= &midbuf_tmp; 00479 preout_tmp= s->preout; 00480 preout= &preout_tmp; 00481 00482 if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar) 00483 postin= in; 00484 00485 if(s->resample_first ? !s->resample : !s->rematrix) 00486 midbuf= postin; 00487 00488 if(s->resample_first ? !s->rematrix : !s->resample) 00489 preout= midbuf; 00490 00491 if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar){ 00492 if(preout==in){ 00493 out_count= FFMIN(out_count, in_count); //TODO check at teh end if this is needed or redundant 00494 av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though 00495 copy(out, in, out_count); 00496 return out_count; 00497 } 00498 else if(preout==postin) preout= midbuf= postin= out; 00499 else if(preout==midbuf) preout= midbuf= out; 00500 else preout= out; 00501 } 00502 00503 if(in != postin){ 00504 swri_audio_convert(s->in_convert, postin, in, in_count); 00505 } 00506 00507 if(s->resample_first){ 00508 if(postin != midbuf) 00509 out_count= resample(s, midbuf, out_count, postin, in_count); 00510 if(midbuf != preout) 00511 swri_rematrix(s, preout, midbuf, out_count, preout==out); 00512 }else{ 00513 if(postin != midbuf) 00514 swri_rematrix(s, midbuf, postin, in_count, midbuf==out); 00515 if(midbuf != preout) 00516 out_count= resample(s, preout, out_count, midbuf, in_count); 00517 } 00518 00519 if(preout != out){ 00520 //FIXME packed doesnt need more than 1 chan here! 00521 swri_audio_convert(s->out_convert, out, preout, out_count); 00522 } 00523 if(!in_arg) 00524 s->in_buffer_count = 0; 00525 return out_count; 00526 } 00527