2 * Copyright (C) 2011-2012 Michael Niedermayer (michaelni@gmx.at)
4 * This file is part of libswresample
6 * libswresample is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * libswresample is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with libswresample; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 #include "libavutil/opt.h"
22 #include "swresample_internal.h"
23 #include "audioconvert.h"
24 #include "libavutil/avassert.h"
25 #include "libavutil/channel_layout.h"
30 #define C15DB 1.189207115
32 #define C_15DB 0.840896415
33 #define C_30DB M_SQRT1_2
34 #define C_45DB 0.594603558
39 //TODO split options array out?
40 #define OFFSET(x) offsetof(SwrContext,x)
41 #define PARAM AV_OPT_FLAG_AUDIO_PARAM
43 static const AVOption options[]={
44 {"ich" , "set input channel count" , OFFSET( in.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
45 {"in_channel_count" , "set input channel count" , OFFSET( in.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
46 {"och" , "set output channel count" , OFFSET(out.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
47 {"out_channel_count" , "set output channel count" , OFFSET(out.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
48 {"uch" , "set used channel count" , OFFSET(used_ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
49 {"used_channel_count" , "set used channel count" , OFFSET(used_ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
50 {"isr" , "set input sample rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
51 {"in_sample_rate" , "set input sample rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
52 {"osr" , "set output sample rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
53 {"out_sample_rate" , "set output sample rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
54 {"isf" , "set input sample format" , OFFSET( in_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , AV_SAMPLE_FMT_NB-1, PARAM},
55 {"in_sample_fmt" , "set input sample format" , OFFSET( in_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , AV_SAMPLE_FMT_NB-1, PARAM},
56 {"osf" , "set output sample format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , AV_SAMPLE_FMT_NB-1, PARAM},
57 {"out_sample_fmt" , "set output sample format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , AV_SAMPLE_FMT_NB-1, PARAM},
58 {"tsf" , "set internal sample format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , AV_SAMPLE_FMT_NB-1, PARAM},
59 {"internal_sample_fmt" , "set internal sample format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , AV_SAMPLE_FMT_NB-1, PARAM},
60 {"icl" , "set input channel layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
61 {"in_channel_layout" , "set input channel layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
62 {"ocl" , "set output channel layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
63 {"out_channel_layout" , "set output channel layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
64 {"clev" , "set center mix level" , OFFSET(clev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
65 {"center_mix_level" , "set center mix level" , OFFSET(clev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
66 {"slev" , "set surround mix level" , OFFSET(slev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
67 {"surround_mix_level" , "set surround mix Level" , OFFSET(slev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
68 {"lfe_mix_level" , "set LFE mix level" , OFFSET(lfe_mix_level ), AV_OPT_TYPE_FLOAT, {.dbl=0 }, -32 , 32 , PARAM},
69 {"rmvol" , "set rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
70 {"rematrix_volume" , "set rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
72 {"flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
73 {"swr_flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
74 {"res" , "force resampling" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_FLAG_RESAMPLE }, INT_MIN, INT_MAX , PARAM, "flags"},
76 {"dither_scale" , "set dither scale" , OFFSET(dither_scale ), AV_OPT_TYPE_FLOAT, {.dbl=1 }, 0 , INT_MAX , PARAM},
78 {"dither_method" , "set dither method" , OFFSET(dither_method ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_DITHER_NB-1, PARAM, "dither_method"},
79 {"rectangular" , "select rectangular dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX , PARAM, "dither_method"},
80 {"triangular" , "select triangular dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX , PARAM, "dither_method"},
81 {"triangular_hp" , "select triangular dither with high pass" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR_HIGHPASS }, INT_MIN, INT_MAX, PARAM, "dither_method"},
83 {"filter_size" , "set swr resampling filter size", OFFSET(filter_size) , AV_OPT_TYPE_INT , {.i64=16 }, 0 , INT_MAX , PARAM },
84 {"phase_shift" , "set swr resampling phase shift", OFFSET(phase_shift) , AV_OPT_TYPE_INT , {.i64=10 }, 0 , 30 , PARAM },
85 {"linear_interp" , "enable linear interpolation" , OFFSET(linear_interp) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
86 {"cutoff" , "set cutoff frequency ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0. }, 0 , 1 , PARAM },
87 {"resampler" , "set resampling Engine" , OFFSET(engine) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_ENGINE_NB-1, PARAM, "resampler"},
88 {"swr" , "select SW Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SWR }, INT_MIN, INT_MAX , PARAM, "resampler"},
89 {"soxr" , "select SoX Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SOXR }, INT_MIN, INT_MAX , PARAM, "resampler"},
90 {"precision" , "set soxr resampling precision (in bits)"
91 , OFFSET(precision) , AV_OPT_TYPE_DOUBLE,{.dbl=20.0 }, 15.0 , 33.0 , PARAM },
92 {"cheby" , "enable soxr Chebyshev passband & higher-precision irrational ratio approximation"
93 , OFFSET(cheby) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
94 {"min_comp" , "set minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
95 , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX }, 0 , FLT_MAX , PARAM },
96 {"min_hard_comp" , "set minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
97 , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1 }, 0 , INT_MAX , PARAM },
98 {"comp_duration" , "set duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
99 , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1 }, 0 , INT_MAX , PARAM },
100 {"max_soft_comp" , "set maximum factor by which data is stretched/squeezed to make it match the timestamps."
101 , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
102 {"async" , "simplified 1 parameter audio timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum stretch/squeeze in samples per second)"
103 , OFFSET(async) , AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
105 { "matrix_encoding" , "set matrixed stereo encoding" , OFFSET(matrix_encoding), AV_OPT_TYPE_INT ,{.i64 = AV_MATRIX_ENCODING_NONE}, AV_MATRIX_ENCODING_NONE, AV_MATRIX_ENCODING_NB-1, PARAM, "matrix_encoding" },
106 { "none", "select none", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
107 { "dolby", "select Dolby", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
108 { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
110 { "filter_type" , "select swr filter type" , OFFSET(filter_type) , AV_OPT_TYPE_INT , { .i64 = SWR_FILTER_TYPE_KAISER }, SWR_FILTER_TYPE_CUBIC, SWR_FILTER_TYPE_KAISER, PARAM, "filter_type" },
111 { "cubic" , "select cubic" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC }, INT_MIN, INT_MAX, PARAM, "filter_type" },
112 { "blackman_nuttall", "select Blackman Nuttall Windowed Sinc", 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_BLACKMAN_NUTTALL }, INT_MIN, INT_MAX, PARAM, "filter_type" },
113 { "kaiser" , "select Kaiser Windowed Sinc" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER }, INT_MIN, INT_MAX, PARAM, "filter_type" },
115 { "kaiser_beta" , "set swr Kaiser Window Beta" , OFFSET(kaiser_beta) , AV_OPT_TYPE_INT , {.i64=9 }, 2 , 16 , PARAM },
120 static const char* context_to_name(void* ptr) {
124 static const AVClass av_class = {
125 .class_name = "SWResampler",
126 .item_name = context_to_name,
128 .version = LIBAVUTIL_VERSION_INT,
129 .log_level_offset_offset = OFFSET(log_level_offset),
130 .parent_log_context_offset = OFFSET(log_ctx),
131 .category = AV_CLASS_CATEGORY_SWRESAMPLER,
134 unsigned swresample_version(void)
136 av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
137 return LIBSWRESAMPLE_VERSION_INT;
140 const char *swresample_configuration(void)
142 return FFMPEG_CONFIGURATION;
145 const char *swresample_license(void)
147 #define LICENSE_PREFIX "libswresample license: "
148 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
151 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
152 if(!s || s->in_convert) // s needs to be allocated but not initialized
153 return AVERROR(EINVAL);
154 s->channel_map = channel_map;
158 const AVClass *swr_get_class(void)
163 av_cold struct SwrContext *swr_alloc(void){
164 SwrContext *s= av_mallocz(sizeof(SwrContext));
166 s->av_class= &av_class;
167 av_opt_set_defaults(s);
172 struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
173 int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
174 int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate,
175 int log_offset, void *log_ctx){
176 if(!s) s= swr_alloc();
179 s->log_level_offset= log_offset;
182 av_opt_set_int(s, "ocl", out_ch_layout, 0);
183 av_opt_set_int(s, "osf", out_sample_fmt, 0);
184 av_opt_set_int(s, "osr", out_sample_rate, 0);
185 av_opt_set_int(s, "icl", in_ch_layout, 0);
186 av_opt_set_int(s, "isf", in_sample_fmt, 0);
187 av_opt_set_int(s, "isr", in_sample_rate, 0);
188 av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE, 0);
189 av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
190 av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
191 av_opt_set_int(s, "uch", 0, 0);
195 static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
197 a->bps = av_get_bytes_per_sample(fmt);
198 a->planar= av_sample_fmt_is_planar(fmt);
201 static void free_temp(AudioData *a){
203 memset(a, 0, sizeof(*a));
206 av_cold void swr_free(SwrContext **ss){
209 free_temp(&s->postin);
210 free_temp(&s->midbuf);
211 free_temp(&s->preout);
212 free_temp(&s->in_buffer);
213 free_temp(&s->dither);
214 swri_audio_convert_free(&s-> in_convert);
215 swri_audio_convert_free(&s->out_convert);
216 swri_audio_convert_free(&s->full_convert);
218 s->resampler->free(&s->resample);
219 swri_rematrix_free(s);
225 av_cold int swr_init(struct SwrContext *s){
226 s->in_buffer_index= 0;
227 s->in_buffer_count= 0;
228 s->resample_in_constraint= 0;
229 free_temp(&s->postin);
230 free_temp(&s->midbuf);
231 free_temp(&s->preout);
232 free_temp(&s->in_buffer);
233 free_temp(&s->dither);
234 memset(s->in.ch, 0, sizeof(s->in.ch));
235 memset(s->out.ch, 0, sizeof(s->out.ch));
236 swri_audio_convert_free(&s-> in_convert);
237 swri_audio_convert_free(&s->out_convert);
238 swri_audio_convert_free(&s->full_convert);
239 swri_rematrix_free(s);
243 if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
244 av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
245 return AVERROR(EINVAL);
247 if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
248 av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
249 return AVERROR(EINVAL);
252 if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
253 if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
254 s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
255 }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
256 s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
258 av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
259 s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
263 if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
264 &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
265 &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
266 &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
267 av_log(s, AV_LOG_ERROR, "Requested sample format %s is not supported internally, S16/S32/FLT/DBL is supported\n", av_get_sample_fmt_name(s->int_sample_fmt));
268 return AVERROR(EINVAL);
273 extern struct Resampler const soxr_resampler;
274 case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
276 case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
278 av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
279 return AVERROR(EINVAL);
282 set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
283 set_audiodata_fmt(&s->out, s->out_sample_fmt);
286 if (s->min_compensation >= FLT_MAX/2)
287 s->min_compensation = 0.001;
288 if (s->async > 1.0001) {
289 s->max_soft_compensation = s->async / (double) s->in_sample_rate;
293 if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
294 s->resample = s->resampler->init(s->resample, s->out_sample_rate, s->in_sample_rate, s->filter_size, s->phase_shift, s->linear_interp, s->cutoff, s->int_sample_fmt, s->filter_type, s->kaiser_beta, s->precision, s->cheby);
296 s->resampler->free(&s->resample);
297 if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
298 && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
299 && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
300 && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
302 av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
306 if(!s->used_ch_count)
307 s->used_ch_count= s->in.ch_count;
309 if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
310 av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
314 if(!s-> in_ch_layout)
315 s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
316 if(!s->out_ch_layout)
317 s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);
319 s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
322 #define RSC 1 //FIXME finetune
324 s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
325 if(!s->used_ch_count)
326 s->used_ch_count= s->in.ch_count;
328 s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);
330 if(!s-> in.ch_count){
331 av_assert0(!s->in_ch_layout);
332 av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
336 if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
337 av_log(s, AV_LOG_ERROR, "Rematrix is needed but there is not enough information to do it\n");
341 av_assert0(s->used_ch_count);
342 av_assert0(s->out.ch_count);
343 s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
347 if(!s->resample && !s->rematrix && !s->channel_map && !s->dither_method){
348 s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
349 s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
353 s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
354 s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
355 s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
356 s->int_sample_fmt, s->out.ch_count, NULL, 0);
365 s->midbuf.ch_count= s->used_ch_count;
367 s->in_buffer.ch_count= s->used_ch_count;
369 if(!s->resample_first){
370 s->midbuf.ch_count= s->out.ch_count;
372 s->in_buffer.ch_count = s->out.ch_count;
375 set_audiodata_fmt(&s->postin, s->int_sample_fmt);
376 set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
377 set_audiodata_fmt(&s->preout, s->int_sample_fmt);
380 set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
383 s->dither = s->preout;
385 if(s->rematrix || s->dither_method)
386 return swri_rematrix_init(s);
391 int swri_realloc_audio(AudioData *a, int count){
395 if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
396 return AVERROR(EINVAL);
398 if(a->count >= count)
403 countb= FFALIGN(count*a->bps, ALIGN);
407 av_assert0(a->ch_count);
409 a->data= av_mallocz(countb*a->ch_count);
411 return AVERROR(ENOMEM);
412 for(i=0; i<a->ch_count; i++){
413 a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
414 if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
416 if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
423 static void copy(AudioData *out, AudioData *in,
425 av_assert0(out->planar == in->planar);
426 av_assert0(out->bps == in->bps);
427 av_assert0(out->ch_count == in->ch_count);
430 for(ch=0; ch<out->ch_count; ch++)
431 memcpy(out->ch[ch], in->ch[ch], count*out->bps);
433 memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
436 static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
439 memset(out->ch, 0, sizeof(out->ch));
440 }else if(out->planar){
441 for(i=0; i<out->ch_count; i++)
442 out->ch[i]= in_arg[i];
444 for(i=0; i<out->ch_count; i++)
445 out->ch[i]= in_arg[0] + i*out->bps;
449 static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
452 for(i=0; i<out->ch_count; i++)
453 in_arg[i]= out->ch[i];
455 in_arg[0]= out->ch[0];
461 * out may be equal in.
463 static void buf_set(AudioData *out, AudioData *in, int count){
466 for(ch=0; ch<out->ch_count; ch++)
467 out->ch[ch]= in->ch[ch] + count*out->bps;
469 for(ch=out->ch_count-1; ch>=0; ch--)
470 out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
476 * @return number of samples output per channel
478 static int resample(SwrContext *s, AudioData *out_param, int out_count,
479 const AudioData * in_param, int in_count){
480 AudioData in, out, tmp;
484 av_assert1(s->in_buffer.ch_count == in_param->ch_count);
485 av_assert1(s->in_buffer.planar == in_param->planar);
486 av_assert1(s->in_buffer.fmt == in_param->fmt);
492 int ret, size, consumed;
493 if(!s->resample_in_constraint && s->in_buffer_count){
494 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
495 ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
498 buf_set(&out, &out, ret);
499 s->in_buffer_count -= consumed;
500 s->in_buffer_index += consumed;
504 if(s->in_buffer_count <= border){
505 buf_set(&in, &in, -s->in_buffer_count);
506 in_count += s->in_buffer_count;
507 s->in_buffer_count=0;
508 s->in_buffer_index=0;
513 if((s->flushed || in_count) && !s->in_buffer_count){
514 s->in_buffer_index=0;
515 ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, in_count, &consumed);
518 buf_set(&out, &out, ret);
519 in_count -= consumed;
520 buf_set(&in, &in, consumed);
523 //TODO is this check sane considering the advanced copy avoidance below
524 size= s->in_buffer_index + s->in_buffer_count + in_count;
525 if( size > s->in_buffer.count
526 && s->in_buffer_count + in_count <= s->in_buffer_index){
527 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
528 copy(&s->in_buffer, &tmp, s->in_buffer_count);
529 s->in_buffer_index=0;
531 if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
536 if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
538 buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
539 copy(&tmp, &in, /*in_*/count);
540 s->in_buffer_count += count;
543 buf_set(&in, &in, count);
544 s->resample_in_constraint= 0;
545 if(s->in_buffer_count != count || in_count)
551 s->resample_in_constraint= !!out_count;
556 static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
557 AudioData *in , int in_count){
558 AudioData *postin, *midbuf, *preout;
560 AudioData preout_tmp, midbuf_tmp;
563 av_assert0(!s->resample);
564 swri_audio_convert(s->full_convert, out, in, in_count);
568 // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
569 // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
571 if((ret=swri_realloc_audio(&s->postin, in_count))<0)
573 if(s->resample_first){
574 av_assert0(s->midbuf.ch_count == s->used_ch_count);
575 if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
578 av_assert0(s->midbuf.ch_count == s->out.ch_count);
579 if((ret=swri_realloc_audio(&s->midbuf, in_count))<0)
582 if((ret=swri_realloc_audio(&s->preout, out_count))<0)
587 midbuf_tmp= s->midbuf;
589 preout_tmp= s->preout;
592 if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
595 if(s->resample_first ? !s->resample : !s->rematrix)
598 if(s->resample_first ? !s->rematrix : !s->resample)
601 if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar){
603 out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
604 av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
605 copy(out, in, out_count);
608 else if(preout==postin) preout= midbuf= postin= out;
609 else if(preout==midbuf) preout= midbuf= out;
614 swri_audio_convert(s->in_convert, postin, in, in_count);
617 if(s->resample_first){
619 out_count= resample(s, midbuf, out_count, postin, in_count);
621 swri_rematrix(s, preout, midbuf, out_count, preout==out);
624 swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
626 out_count= resample(s, preout, out_count, midbuf, in_count);
629 if(preout != out && out_count){
630 if(s->dither_method){
632 int dither_count= FFMAX(out_count, 1<<16);
633 av_assert0(preout != in);
635 if((ret=swri_realloc_audio(&s->dither, dither_count))<0)
638 for(ch=0; ch<s->dither.ch_count; ch++)
639 swri_get_dither(s, s->dither.ch[ch], s->dither.count, 12345678913579<<ch, s->out_sample_fmt, s->int_sample_fmt);
640 av_assert0(s->dither.ch_count == preout->ch_count);
642 if(s->dither_pos + out_count > s->dither.count)
645 for(ch=0; ch<preout->ch_count; ch++)
646 s->mix_2_1_f(preout->ch[ch], preout->ch[ch], s->dither.ch[ch] + s->dither.bps * s->dither_pos, s->native_one, 0, 0, out_count);
648 s->dither_pos += out_count;
650 //FIXME packed doesnt need more than 1 chan here!
651 swri_audio_convert(s->out_convert, out, preout, out_count);
656 int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
657 const uint8_t *in_arg [SWR_CH_MAX], int in_count){
658 AudioData * in= &s->in;
659 AudioData *out= &s->out;
661 if(s->drop_output > 0){
663 AudioData tmp = s->out;
664 uint8_t *tmp_arg[SWR_CH_MAX];
667 if((ret=swri_realloc_audio(&tmp, s->drop_output))<0)
670 reversefill_audiodata(&tmp, tmp_arg);
671 s->drop_output *= -1; //FIXME find a less hackish solution
672 ret = swr_convert(s, tmp_arg, -s->drop_output, in_arg, in_count); //FIXME optimize but this is as good as never called so maybe it doesnt matter
673 s->drop_output *= -1;
675 s->drop_output -= ret;
678 if(s->drop_output || !out_arg)
686 s->resampler->flush(s);
687 s->resample_in_constraint = 0;
689 }else if(!s->in_buffer_count){
693 fill_audiodata(in , (void*)in_arg);
695 fill_audiodata(out, out_arg);
698 int ret = swr_convert_internal(s, out, out_count, in, in_count);
699 if(ret>0 && !s->drop_output)
700 s->outpts += ret * (int64_t)s->in_sample_rate;
706 size = FFMIN(out_count, s->in_buffer_count);
708 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
709 ret= swr_convert_internal(s, out, size, &tmp, size);
713 s->in_buffer_count -= ret;
714 s->in_buffer_index += ret;
715 buf_set(out, out, ret);
717 if(!s->in_buffer_count)
718 s->in_buffer_index = 0;
722 size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
724 if(in_count > out_count) { //FIXME move after swr_convert_internal
725 if( size > s->in_buffer.count
726 && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
727 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
728 copy(&s->in_buffer, &tmp, s->in_buffer_count);
729 s->in_buffer_index=0;
731 if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
736 size = FFMIN(in_count, out_count);
737 ret= swr_convert_internal(s, out, size, in, size);
740 buf_set(in, in, ret);
745 buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
746 copy(&tmp, in, in_count);
747 s->in_buffer_count += in_count;
750 if(ret2>0 && !s->drop_output)
751 s->outpts += ret2 * (int64_t)s->in_sample_rate;
756 int swr_drop_output(struct SwrContext *s, int count){
757 s->drop_output += count;
759 if(s->drop_output <= 0)
762 av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
763 return swr_convert(s, NULL, s->drop_output, NULL, 0);
766 int swr_inject_silence(struct SwrContext *s, int count){
768 AudioData silence = s->in;
769 uint8_t *tmp_arg[SWR_CH_MAX];
776 if((ret=swri_realloc_audio(&silence, count))<0)
779 if(silence.planar) for(i=0; i<silence.ch_count; i++) {
780 memset(silence.ch[i], silence.bps==1 ? 0x80 : 0, count*silence.bps);
782 memset(silence.ch[0], silence.bps==1 ? 0x80 : 0, count*silence.bps*silence.ch_count);
784 reversefill_audiodata(&silence, tmp_arg);
785 av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
786 ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
787 av_freep(&silence.data);
791 int64_t swr_get_delay(struct SwrContext *s, int64_t base){
792 if (s->resampler && s->resample){
793 return s->resampler->get_delay(s, base);
795 return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
799 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
802 if (!s || compensation_distance < 0)
803 return AVERROR(EINVAL);
804 if (!compensation_distance && sample_delta)
805 return AVERROR(EINVAL);
807 s->flags |= SWR_FLAG_RESAMPLE;
812 if (!s->resampler->set_compensation){
813 return AVERROR(EINVAL);
815 return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
819 int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
822 if(s->min_compensation >= FLT_MAX) {
823 return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
825 int64_t delta = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate) - s->outpts;
826 double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
828 if(fabs(fdelta) > s->min_compensation) {
829 if(!s->outpts || fabs(fdelta) > s->min_hard_compensation){
831 if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
832 else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
834 av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
836 } else if(s->soft_compensation_duration && s->max_soft_compensation) {
837 int duration = s->out_sample_rate * s->soft_compensation_duration;
838 double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
839 int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
840 av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
841 swr_set_compensation(s, comp, duration);