swresample: use the internal buffer for resampling the last few samples
[ffmpeg.git] / libswresample / swresample.c
1 /*
2  * Copyright (C) 2011-2013 Michael Niedermayer (michaelni@gmx.at)
3  *
4  * This file is part of libswresample
5  *
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.
10  *
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.
15  *
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
19  */
20
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"
26
27 #include <float.h>
28
29 #define  C30DB  M_SQRT2
30 #define  C15DB  1.189207115
31 #define C__0DB  1.0
32 #define C_15DB  0.840896415
33 #define C_30DB  M_SQRT1_2
34 #define C_45DB  0.594603558
35 #define C_60DB  0.5
36
37 #define ALIGN 32
38
39 //TODO split options array out?
40 #define OFFSET(x) offsetof(SwrContext,x)
41 #define PARAM AV_OPT_FLAG_AUDIO_PARAM
42
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_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
61 {"in_channel_layout"    , "set input channel layout"    , OFFSET( in_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
62 {"ocl"                  , "set output channel layout"   , OFFSET(out_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
63 {"out_channel_layout"   , "set output channel layout"   , OFFSET(out_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.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},
71 {"rematrix_maxval"      , "set rematrix maxval"         , OFFSET(rematrix_maxval), AV_OPT_TYPE_FLOAT, {.dbl=0.0                   }, 0      , 1000      , PARAM},
72
73 {"flags"                , "set flags"                   , OFFSET(flags          ), AV_OPT_TYPE_FLAGS, {.i64=0                     }, 0      , UINT_MAX  , PARAM, "flags"},
74 {"swr_flags"            , "set flags"                   , OFFSET(flags          ), AV_OPT_TYPE_FLAGS, {.i64=0                     }, 0      , UINT_MAX  , PARAM, "flags"},
75 {"res"                  , "force resampling"            , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_FLAG_RESAMPLE     }, INT_MIN, INT_MAX   , PARAM, "flags"},
76
77 {"dither_scale"         , "set dither scale"            , OFFSET(dither.scale   ), AV_OPT_TYPE_FLOAT, {.dbl=1                     }, 0      , INT_MAX   , PARAM},
78
79 {"dither_method"        , "set dither method"           , OFFSET(dither.method  ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_DITHER_NB-1, PARAM, "dither_method"},
80 {"rectangular"          , "select rectangular dither"   , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX   , PARAM, "dither_method"},
81 {"triangular"           , "select triangular dither"    , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX   , PARAM, "dither_method"},
82 {"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 {"lipshitz"             , "select lipshitz noise shaping dither" , 0             , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_LIPSHITZ}, INT_MIN, INT_MAX, PARAM, "dither_method"},
84 {"shibata"              , "select shibata noise shaping dither" , 0              , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_SHIBATA }, INT_MIN, INT_MAX, PARAM, "dither_method"},
85 {"low_shibata"          , "select low shibata noise shaping dither" , 0          , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_LOW_SHIBATA }, INT_MIN, INT_MAX, PARAM, "dither_method"},
86 {"high_shibata"         , "select high shibata noise shaping dither" , 0         , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_HIGH_SHIBATA }, INT_MIN, INT_MAX, PARAM, "dither_method"},
87 {"f_weighted"           , "select f-weighted noise shaping dither" , 0           , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_F_WEIGHTED }, INT_MIN, INT_MAX, PARAM, "dither_method"},
88 {"modified_e_weighted"  , "select modified-e-weighted noise shaping dither" , 0  , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_MODIFIED_E_WEIGHTED }, INT_MIN, INT_MAX, PARAM, "dither_method"},
89 {"improved_e_weighted"  , "select improved-e-weighted noise shaping dither" , 0  , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_IMPROVED_E_WEIGHTED }, INT_MIN, INT_MAX, PARAM, "dither_method"},
90
91 {"filter_size"          , "set swr resampling filter size", OFFSET(filter_size)  , AV_OPT_TYPE_INT  , {.i64=32                    }, 0      , INT_MAX   , PARAM },
92 {"phase_shift"          , "set swr resampling phase shift", OFFSET(phase_shift)  , AV_OPT_TYPE_INT  , {.i64=10                    }, 0      , 24        , PARAM },
93 {"linear_interp"        , "enable linear interpolation" , OFFSET(linear_interp)  , AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , 1         , PARAM },
94 {"cutoff"               , "set cutoff frequency ratio"  , OFFSET(cutoff)         , AV_OPT_TYPE_DOUBLE,{.dbl=0.                    }, 0      , 1         , PARAM },
95
96 /* duplicate option in order to work with avconv */
97 {"resample_cutoff"      , "set cutoff frequency ratio"  , OFFSET(cutoff)         , AV_OPT_TYPE_DOUBLE,{.dbl=0.                    }, 0      , 1         , PARAM },
98
99 {"resampler"            , "set resampling Engine"       , OFFSET(engine)         , AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_ENGINE_NB-1, PARAM, "resampler"},
100 {"swr"                  , "select SW Resampler"         , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SWR        }, INT_MIN, INT_MAX   , PARAM, "resampler"},
101 {"soxr"                 , "select SoX Resampler"        , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SOXR       }, INT_MIN, INT_MAX   , PARAM, "resampler"},
102 {"precision"            , "set soxr resampling precision (in bits)"
103                                                         , OFFSET(precision)      , AV_OPT_TYPE_DOUBLE,{.dbl=20.0                  }, 15.0   , 33.0      , PARAM },
104 {"cheby"                , "enable soxr Chebyshev passband & higher-precision irrational ratio approximation"
105                                                         , OFFSET(cheby)          , AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , 1         , PARAM },
106 {"min_comp"             , "set minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
107                                                         , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX               }, 0      , FLT_MAX   , PARAM },
108 {"min_hard_comp"        , "set minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
109                                                         , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1                   }, 0      , INT_MAX   , PARAM },
110 {"comp_duration"        , "set duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
111                                                         , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1                     }, 0      , INT_MAX   , PARAM },
112 {"max_soft_comp"        , "set maximum factor by which data is stretched/squeezed to make it match the timestamps."
113                                                         , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0                     }, INT_MIN, INT_MAX   , PARAM },
114 {"async"                , "simplified 1 parameter audio timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum stretch/squeeze in samples per second)"
115                                                         , OFFSET(async)          , AV_OPT_TYPE_FLOAT ,{.dbl=0                     }, INT_MIN, INT_MAX   , PARAM },
116 {"first_pts"            , "Assume the first pts should be this value (in samples)."
117                                                         , OFFSET(firstpts_in_samples), AV_OPT_TYPE_INT64 ,{.i64=AV_NOPTS_VALUE    }, INT64_MIN,INT64_MAX, PARAM },
118
119 { "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" },
120     { "none",  "select none",               0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE  }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
121     { "dolby", "select Dolby",              0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
122     { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
123
124 { "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" },
125     { "cubic"           , "select cubic"                , 0                      , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC            }, INT_MIN, INT_MAX, PARAM, "filter_type" },
126     { "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" },
127     { "kaiser"          , "select Kaiser Windowed Sinc" , 0                      , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER           }, INT_MIN, INT_MAX, PARAM, "filter_type" },
128
129 { "kaiser_beta"         , "set swr Kaiser Window Beta"  , OFFSET(kaiser_beta)    , AV_OPT_TYPE_INT  , {.i64=9                     }, 2      , 16        , PARAM },
130
131 { "output_sample_bits"  , "set swr number of output sample bits", OFFSET(dither.output_sample_bits), AV_OPT_TYPE_INT  , {.i64=0   }, 0      , 64        , PARAM },
132 {0}
133 };
134
135 static const char* context_to_name(void* ptr) {
136     return "SWR";
137 }
138
139 static const AVClass av_class = {
140     .class_name                = "SWResampler",
141     .item_name                 = context_to_name,
142     .option                    = options,
143     .version                   = LIBAVUTIL_VERSION_INT,
144     .log_level_offset_offset   = OFFSET(log_level_offset),
145     .parent_log_context_offset = OFFSET(log_ctx),
146     .category                  = AV_CLASS_CATEGORY_SWRESAMPLER,
147 };
148
149 unsigned swresample_version(void)
150 {
151     av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
152     return LIBSWRESAMPLE_VERSION_INT;
153 }
154
155 const char *swresample_configuration(void)
156 {
157     return FFMPEG_CONFIGURATION;
158 }
159
160 const char *swresample_license(void)
161 {
162 #define LICENSE_PREFIX "libswresample license: "
163     return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
164 }
165
166 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
167     if(!s || s->in_convert) // s needs to be allocated but not initialized
168         return AVERROR(EINVAL);
169     s->channel_map = channel_map;
170     return 0;
171 }
172
173 const AVClass *swr_get_class(void)
174 {
175     return &av_class;
176 }
177
178 av_cold struct SwrContext *swr_alloc(void){
179     SwrContext *s= av_mallocz(sizeof(SwrContext));
180     if(s){
181         s->av_class= &av_class;
182         av_opt_set_defaults(s);
183     }
184     return s;
185 }
186
187 struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
188                                       int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
189                                       int64_t  in_ch_layout, enum AVSampleFormat  in_sample_fmt, int  in_sample_rate,
190                                       int log_offset, void *log_ctx){
191     if(!s) s= swr_alloc();
192     if(!s) return NULL;
193
194     s->log_level_offset= log_offset;
195     s->log_ctx= log_ctx;
196
197     av_opt_set_int(s, "ocl", out_ch_layout,   0);
198     av_opt_set_int(s, "osf", out_sample_fmt,  0);
199     av_opt_set_int(s, "osr", out_sample_rate, 0);
200     av_opt_set_int(s, "icl", in_ch_layout,    0);
201     av_opt_set_int(s, "isf", in_sample_fmt,   0);
202     av_opt_set_int(s, "isr", in_sample_rate,  0);
203     av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE,   0);
204     av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
205     av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
206     av_opt_set_int(s, "uch", 0, 0);
207     return s;
208 }
209
210 static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
211     a->fmt   = fmt;
212     a->bps   = av_get_bytes_per_sample(fmt);
213     a->planar= av_sample_fmt_is_planar(fmt);
214 }
215
216 static void free_temp(AudioData *a){
217     av_free(a->data);
218     memset(a, 0, sizeof(*a));
219 }
220
221 av_cold void swr_free(SwrContext **ss){
222     SwrContext *s= *ss;
223     if(s){
224         free_temp(&s->postin);
225         free_temp(&s->midbuf);
226         free_temp(&s->preout);
227         free_temp(&s->in_buffer);
228         free_temp(&s->silence);
229         free_temp(&s->drop_temp);
230         free_temp(&s->dither.noise);
231         free_temp(&s->dither.temp);
232         swri_audio_convert_free(&s-> in_convert);
233         swri_audio_convert_free(&s->out_convert);
234         swri_audio_convert_free(&s->full_convert);
235         if (s->resampler)
236             s->resampler->free(&s->resample);
237         swri_rematrix_free(s);
238     }
239
240     av_freep(ss);
241 }
242
243 av_cold int swr_init(struct SwrContext *s){
244     int ret;
245     s->in_buffer_index= 0;
246     s->in_buffer_count= 0;
247     s->resample_in_constraint= 0;
248     free_temp(&s->postin);
249     free_temp(&s->midbuf);
250     free_temp(&s->preout);
251     free_temp(&s->in_buffer);
252     free_temp(&s->silence);
253     free_temp(&s->drop_temp);
254     free_temp(&s->dither.noise);
255     free_temp(&s->dither.temp);
256     memset(s->in.ch, 0, sizeof(s->in.ch));
257     memset(s->out.ch, 0, sizeof(s->out.ch));
258     swri_audio_convert_free(&s-> in_convert);
259     swri_audio_convert_free(&s->out_convert);
260     swri_audio_convert_free(&s->full_convert);
261     swri_rematrix_free(s);
262
263     s->flushed = 0;
264
265     if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
266         av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
267         return AVERROR(EINVAL);
268     }
269     if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
270         av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
271         return AVERROR(EINVAL);
272     }
273
274     if(av_get_channel_layout_nb_channels(s-> in_ch_layout) > SWR_CH_MAX) {
275         av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
276         s->in_ch_layout = 0;
277     }
278
279     if(av_get_channel_layout_nb_channels(s->out_ch_layout) > SWR_CH_MAX) {
280         av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
281         s->out_ch_layout = 0;
282     }
283
284     switch(s->engine){
285 #if CONFIG_LIBSOXR
286         extern struct Resampler const soxr_resampler;
287         case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
288 #endif
289         case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
290         default:
291             av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
292             return AVERROR(EINVAL);
293     }
294
295     if(!s->used_ch_count)
296         s->used_ch_count= s->in.ch_count;
297
298     if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
299         av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
300         s-> in_ch_layout= 0;
301     }
302
303     if(!s-> in_ch_layout)
304         s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
305     if(!s->out_ch_layout)
306         s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);
307
308     s->rematrix= s->out_ch_layout  !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
309                  s->rematrix_custom;
310
311     if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
312         if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
313             s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
314         }else if(   av_get_planar_sample_fmt(s-> in_sample_fmt) == AV_SAMPLE_FMT_S32P
315                  && av_get_planar_sample_fmt(s->out_sample_fmt) == AV_SAMPLE_FMT_S32P
316                  && !s->rematrix
317                  && s->engine != SWR_ENGINE_SOXR){
318             s->int_sample_fmt= AV_SAMPLE_FMT_S32P;
319         }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
320             s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
321         }else{
322             av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
323             s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
324         }
325     }
326
327     if(   s->int_sample_fmt != AV_SAMPLE_FMT_S16P
328         &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
329         &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
330         &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
331         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));
332         return AVERROR(EINVAL);
333     }
334
335     set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
336     set_audiodata_fmt(&s->out, s->out_sample_fmt);
337
338     if (s->firstpts_in_samples != AV_NOPTS_VALUE) {
339         if (!s->async && s->min_compensation >= FLT_MAX/2)
340             s->async = 1;
341         s->firstpts =
342         s->outpts   = s->firstpts_in_samples * s->out_sample_rate;
343     } else
344         s->firstpts = AV_NOPTS_VALUE;
345
346     if (s->async) {
347         if (s->min_compensation >= FLT_MAX/2)
348             s->min_compensation = 0.001;
349         if (s->async > 1.0001) {
350             s->max_soft_compensation = s->async / (double) s->in_sample_rate;
351         }
352     }
353
354     if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
355         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);
356     }else
357         s->resampler->free(&s->resample);
358     if(    s->int_sample_fmt != AV_SAMPLE_FMT_S16P
359         && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
360         && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
361         && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
362         && s->resample){
363         av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
364         return -1;
365     }
366
367 #define RSC 1 //FIXME finetune
368     if(!s-> in.ch_count)
369         s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
370     if(!s->used_ch_count)
371         s->used_ch_count= s->in.ch_count;
372     if(!s->out.ch_count)
373         s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);
374
375     if(!s-> in.ch_count){
376         av_assert0(!s->in_ch_layout);
377         av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
378         return -1;
379     }
380
381     if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
382         char l1[1024], l2[1024];
383         av_get_channel_layout_string(l1, sizeof(l1), s-> in.ch_count, s-> in_ch_layout);
384         av_get_channel_layout_string(l2, sizeof(l2), s->out.ch_count, s->out_ch_layout);
385         av_log(s, AV_LOG_ERROR, "Rematrix is needed between %s and %s "
386                "but there is not enough information to do it\n", l1, l2);
387         return -1;
388     }
389
390 av_assert0(s->used_ch_count);
391 av_assert0(s->out.ch_count);
392     s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
393
394     s->in_buffer= s->in;
395     s->silence  = s->in;
396     s->drop_temp= s->out;
397
398     if(!s->resample && !s->rematrix && !s->channel_map && !s->dither.method){
399         s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
400                                                    s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
401         return 0;
402     }
403
404     s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
405                                              s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
406     s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
407                                              s->int_sample_fmt, s->out.ch_count, NULL, 0);
408
409     if (!s->in_convert || !s->out_convert)
410         return AVERROR(ENOMEM);
411
412     s->postin= s->in;
413     s->preout= s->out;
414     s->midbuf= s->in;
415
416     if(s->channel_map){
417         s->postin.ch_count=
418         s->midbuf.ch_count= s->used_ch_count;
419         if(s->resample)
420             s->in_buffer.ch_count= s->used_ch_count;
421     }
422     if(!s->resample_first){
423         s->midbuf.ch_count= s->out.ch_count;
424         if(s->resample)
425             s->in_buffer.ch_count = s->out.ch_count;
426     }
427
428     set_audiodata_fmt(&s->postin, s->int_sample_fmt);
429     set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
430     set_audiodata_fmt(&s->preout, s->int_sample_fmt);
431
432     if(s->resample){
433         set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
434     }
435
436     if ((ret = swri_dither_init(s, s->out_sample_fmt, s->int_sample_fmt)) < 0)
437         return ret;
438
439     if(s->rematrix || s->dither.method)
440         return swri_rematrix_init(s);
441
442     return 0;
443 }
444
445 int swri_realloc_audio(AudioData *a, int count){
446     int i, countb;
447     AudioData old;
448
449     if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
450         return AVERROR(EINVAL);
451
452     if(a->count >= count)
453         return 0;
454
455     count*=2;
456
457     countb= FFALIGN(count*a->bps, ALIGN);
458     old= *a;
459
460     av_assert0(a->bps);
461     av_assert0(a->ch_count);
462
463     a->data= av_mallocz(countb*a->ch_count);
464     if(!a->data)
465         return AVERROR(ENOMEM);
466     for(i=0; i<a->ch_count; i++){
467         a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
468         if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
469     }
470     if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
471     av_freep(&old.data);
472     a->count= count;
473
474     return 1;
475 }
476
477 static void copy(AudioData *out, AudioData *in,
478                  int count){
479     av_assert0(out->planar == in->planar);
480     av_assert0(out->bps == in->bps);
481     av_assert0(out->ch_count == in->ch_count);
482     if(out->planar){
483         int ch;
484         for(ch=0; ch<out->ch_count; ch++)
485             memcpy(out->ch[ch], in->ch[ch], count*out->bps);
486     }else
487         memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
488 }
489
490 static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
491     int i;
492     if(!in_arg){
493         memset(out->ch, 0, sizeof(out->ch));
494     }else if(out->planar){
495         for(i=0; i<out->ch_count; i++)
496             out->ch[i]= in_arg[i];
497     }else{
498         for(i=0; i<out->ch_count; i++)
499             out->ch[i]= in_arg[0] + i*out->bps;
500     }
501 }
502
503 static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
504     int i;
505     if(out->planar){
506         for(i=0; i<out->ch_count; i++)
507             in_arg[i]= out->ch[i];
508     }else{
509         in_arg[0]= out->ch[0];
510     }
511 }
512
513 /**
514  *
515  * out may be equal in.
516  */
517 static void buf_set(AudioData *out, AudioData *in, int count){
518     int ch;
519     if(in->planar){
520         for(ch=0; ch<out->ch_count; ch++)
521             out->ch[ch]= in->ch[ch] + count*out->bps;
522     }else{
523         for(ch=out->ch_count-1; ch>=0; ch--)
524             out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
525     }
526 }
527
528 /**
529  *
530  * @return number of samples output per channel
531  */
532 static int resample(SwrContext *s, AudioData *out_param, int out_count,
533                              const AudioData * in_param, int in_count){
534     AudioData in, out, tmp;
535     int ret_sum=0;
536     int border=0;
537     int padless = ARCH_X86 && s->engine == SWR_ENGINE_SWR ? 7 : 0;
538
539     av_assert1(s->in_buffer.ch_count == in_param->ch_count);
540     av_assert1(s->in_buffer.planar   == in_param->planar);
541     av_assert1(s->in_buffer.fmt      == in_param->fmt);
542
543     tmp=out=*out_param;
544     in =  *in_param;
545
546     do{
547         int ret, size, consumed;
548         if(!s->resample_in_constraint && s->in_buffer_count){
549             buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
550             ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
551             out_count -= ret;
552             ret_sum += ret;
553             buf_set(&out, &out, ret);
554             s->in_buffer_count -= consumed;
555             s->in_buffer_index += consumed;
556
557             if(!in_count)
558                 break;
559             if(s->in_buffer_count <= border){
560                 buf_set(&in, &in, -s->in_buffer_count);
561                 in_count += s->in_buffer_count;
562                 s->in_buffer_count=0;
563                 s->in_buffer_index=0;
564                 border = 0;
565             }
566         }
567
568         if((s->flushed || in_count > padless) && !s->in_buffer_count){
569             s->in_buffer_index=0;
570             ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, FFMAX(in_count-padless, 0), &consumed);
571             out_count -= ret;
572             ret_sum += ret;
573             buf_set(&out, &out, ret);
574             in_count -= consumed;
575             buf_set(&in, &in, consumed);
576         }
577
578         //TODO is this check sane considering the advanced copy avoidance below
579         size= s->in_buffer_index + s->in_buffer_count + in_count;
580         if(   size > s->in_buffer.count
581            && s->in_buffer_count + in_count <= s->in_buffer_index){
582             buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
583             copy(&s->in_buffer, &tmp, s->in_buffer_count);
584             s->in_buffer_index=0;
585         }else
586             if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
587                 return ret;
588
589         if(in_count){
590             int count= in_count;
591             if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
592
593             buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
594             copy(&tmp, &in, /*in_*/count);
595             s->in_buffer_count += count;
596             in_count -= count;
597             border += count;
598             buf_set(&in, &in, count);
599             s->resample_in_constraint= 0;
600             if(s->in_buffer_count != count || in_count)
601                 continue;
602             if (padless) {
603                 padless = 0;
604                 continue;
605             }
606         }
607         break;
608     }while(1);
609
610     s->resample_in_constraint= !!out_count;
611
612     return ret_sum;
613 }
614
615 static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
616                                                       AudioData *in , int  in_count){
617     AudioData *postin, *midbuf, *preout;
618     int ret/*, in_max*/;
619     AudioData preout_tmp, midbuf_tmp;
620
621     if(s->full_convert){
622         av_assert0(!s->resample);
623         swri_audio_convert(s->full_convert, out, in, in_count);
624         return out_count;
625     }
626
627 //     in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
628 //     in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
629
630     if((ret=swri_realloc_audio(&s->postin, in_count))<0)
631         return ret;
632     if(s->resample_first){
633         av_assert0(s->midbuf.ch_count == s->used_ch_count);
634         if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
635             return ret;
636     }else{
637         av_assert0(s->midbuf.ch_count ==  s->out.ch_count);
638         if((ret=swri_realloc_audio(&s->midbuf,  in_count))<0)
639             return ret;
640     }
641     if((ret=swri_realloc_audio(&s->preout, out_count))<0)
642         return ret;
643
644     postin= &s->postin;
645
646     midbuf_tmp= s->midbuf;
647     midbuf= &midbuf_tmp;
648     preout_tmp= s->preout;
649     preout= &preout_tmp;
650
651     if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
652         postin= in;
653
654     if(s->resample_first ? !s->resample : !s->rematrix)
655         midbuf= postin;
656
657     if(s->resample_first ? !s->rematrix : !s->resample)
658         preout= midbuf;
659
660     if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar
661        && !(s->out_sample_fmt==AV_SAMPLE_FMT_S32P && (s->dither.output_sample_bits&31))){
662         if(preout==in){
663             out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
664             av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
665             copy(out, in, out_count);
666             return out_count;
667         }
668         else if(preout==postin) preout= midbuf= postin= out;
669         else if(preout==midbuf) preout= midbuf= out;
670         else                    preout= out;
671     }
672
673     if(in != postin){
674         swri_audio_convert(s->in_convert, postin, in, in_count);
675     }
676
677     if(s->resample_first){
678         if(postin != midbuf)
679             out_count= resample(s, midbuf, out_count, postin, in_count);
680         if(midbuf != preout)
681             swri_rematrix(s, preout, midbuf, out_count, preout==out);
682     }else{
683         if(postin != midbuf)
684             swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
685         if(midbuf != preout)
686             out_count= resample(s, preout, out_count, midbuf, in_count);
687     }
688
689     if(preout != out && out_count){
690         AudioData *conv_src = preout;
691         if(s->dither.method){
692             int ch;
693             int dither_count= FFMAX(out_count, 1<<16);
694
695             if (preout == in) {
696                 conv_src = &s->dither.temp;
697                 if((ret=swri_realloc_audio(&s->dither.temp, dither_count))<0)
698                     return ret;
699             }
700
701             if((ret=swri_realloc_audio(&s->dither.noise, dither_count))<0)
702                 return ret;
703             if(ret)
704                 for(ch=0; ch<s->dither.noise.ch_count; ch++)
705                     swri_get_dither(s, s->dither.noise.ch[ch], s->dither.noise.count, 12345678913579<<ch, s->dither.noise.fmt);
706             av_assert0(s->dither.noise.ch_count == preout->ch_count);
707
708             if(s->dither.noise_pos + out_count > s->dither.noise.count)
709                 s->dither.noise_pos = 0;
710
711             if (s->dither.method < SWR_DITHER_NS){
712                 if (s->mix_2_1_simd) {
713                     int len1= out_count&~15;
714                     int off = len1 * preout->bps;
715
716                     if(len1)
717                         for(ch=0; ch<preout->ch_count; ch++)
718                             s->mix_2_1_simd(conv_src->ch[ch], preout->ch[ch], s->dither.noise.ch[ch] + s->dither.noise.bps * s->dither.noise_pos, s->native_simd_one, 0, 0, len1);
719                     if(out_count != len1)
720                         for(ch=0; ch<preout->ch_count; ch++)
721                             s->mix_2_1_f(conv_src->ch[ch] + off, preout->ch[ch] + off, s->dither.noise.ch[ch] + s->dither.noise.bps * s->dither.noise_pos + off + len1, s->native_one, 0, 0, out_count - len1);
722                 } else {
723                     for(ch=0; ch<preout->ch_count; ch++)
724                         s->mix_2_1_f(conv_src->ch[ch], preout->ch[ch], s->dither.noise.ch[ch] + s->dither.noise.bps * s->dither.noise_pos, s->native_one, 0, 0, out_count);
725                 }
726             } else {
727                 switch(s->int_sample_fmt) {
728                 case AV_SAMPLE_FMT_S16P :swri_noise_shaping_int16(s, conv_src, preout, &s->dither.noise, out_count); break;
729                 case AV_SAMPLE_FMT_S32P :swri_noise_shaping_int32(s, conv_src, preout, &s->dither.noise, out_count); break;
730                 case AV_SAMPLE_FMT_FLTP :swri_noise_shaping_float(s, conv_src, preout, &s->dither.noise, out_count); break;
731                 case AV_SAMPLE_FMT_DBLP :swri_noise_shaping_double(s,conv_src, preout, &s->dither.noise, out_count); break;
732                 }
733             }
734             s->dither.noise_pos += out_count;
735         }
736 //FIXME packed doesn't need more than 1 chan here!
737         swri_audio_convert(s->out_convert, out, conv_src, out_count);
738     }
739     return out_count;
740 }
741
742 int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
743                                 const uint8_t *in_arg [SWR_CH_MAX], int  in_count){
744     AudioData * in= &s->in;
745     AudioData *out= &s->out;
746
747     while(s->drop_output > 0){
748         int ret;
749         uint8_t *tmp_arg[SWR_CH_MAX];
750 #define MAX_DROP_STEP 16384
751         if((ret=swri_realloc_audio(&s->drop_temp, FFMIN(s->drop_output, MAX_DROP_STEP)))<0)
752             return ret;
753
754         reversefill_audiodata(&s->drop_temp, tmp_arg);
755         s->drop_output *= -1; //FIXME find a less hackish solution
756         ret = swr_convert(s, tmp_arg, FFMIN(-s->drop_output, MAX_DROP_STEP), in_arg, in_count); //FIXME optimize but this is as good as never called so maybe it doesn't matter
757         s->drop_output *= -1;
758         in_count = 0;
759         if(ret>0) {
760             s->drop_output -= ret;
761             continue;
762         }
763
764         if(s->drop_output || !out_arg)
765             return 0;
766     }
767
768     if(!in_arg){
769         if(s->resample){
770             if (!s->flushed)
771                 s->resampler->flush(s);
772             s->resample_in_constraint = 0;
773             s->flushed = 1;
774         }else if(!s->in_buffer_count){
775             return 0;
776         }
777     }else
778         fill_audiodata(in ,  (void*)in_arg);
779
780     fill_audiodata(out, out_arg);
781
782     if(s->resample){
783         int ret = swr_convert_internal(s, out, out_count, in, in_count);
784         if(ret>0 && !s->drop_output)
785             s->outpts += ret * (int64_t)s->in_sample_rate;
786         return ret;
787     }else{
788         AudioData tmp= *in;
789         int ret2=0;
790         int ret, size;
791         size = FFMIN(out_count, s->in_buffer_count);
792         if(size){
793             buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
794             ret= swr_convert_internal(s, out, size, &tmp, size);
795             if(ret<0)
796                 return ret;
797             ret2= ret;
798             s->in_buffer_count -= ret;
799             s->in_buffer_index += ret;
800             buf_set(out, out, ret);
801             out_count -= ret;
802             if(!s->in_buffer_count)
803                 s->in_buffer_index = 0;
804         }
805
806         if(in_count){
807             size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
808
809             if(in_count > out_count) { //FIXME move after swr_convert_internal
810                 if(   size > s->in_buffer.count
811                 && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
812                     buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
813                     copy(&s->in_buffer, &tmp, s->in_buffer_count);
814                     s->in_buffer_index=0;
815                 }else
816                     if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
817                         return ret;
818             }
819
820             if(out_count){
821                 size = FFMIN(in_count, out_count);
822                 ret= swr_convert_internal(s, out, size, in, size);
823                 if(ret<0)
824                     return ret;
825                 buf_set(in, in, ret);
826                 in_count -= ret;
827                 ret2 += ret;
828             }
829             if(in_count){
830                 buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
831                 copy(&tmp, in, in_count);
832                 s->in_buffer_count += in_count;
833             }
834         }
835         if(ret2>0 && !s->drop_output)
836             s->outpts += ret2 * (int64_t)s->in_sample_rate;
837         return ret2;
838     }
839 }
840
841 int swr_drop_output(struct SwrContext *s, int count){
842     s->drop_output += count;
843
844     if(s->drop_output <= 0)
845         return 0;
846
847     av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
848     return swr_convert(s, NULL, s->drop_output, NULL, 0);
849 }
850
851 int swr_inject_silence(struct SwrContext *s, int count){
852     int ret, i;
853     uint8_t *tmp_arg[SWR_CH_MAX];
854
855     if(count <= 0)
856         return 0;
857
858 #define MAX_SILENCE_STEP 16384
859     while (count > MAX_SILENCE_STEP) {
860         if ((ret = swr_inject_silence(s, MAX_SILENCE_STEP)) < 0)
861             return ret;
862         count -= MAX_SILENCE_STEP;
863     }
864
865     if((ret=swri_realloc_audio(&s->silence, count))<0)
866         return ret;
867
868     if(s->silence.planar) for(i=0; i<s->silence.ch_count; i++) {
869         memset(s->silence.ch[i], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps);
870     } else
871         memset(s->silence.ch[0], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps*s->silence.ch_count);
872
873     reversefill_audiodata(&s->silence, tmp_arg);
874     av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
875     ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
876     return ret;
877 }
878
879 int64_t swr_get_delay(struct SwrContext *s, int64_t base){
880     if (s->resampler && s->resample){
881         return s->resampler->get_delay(s, base);
882     }else{
883         return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
884     }
885 }
886
887 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
888     int ret;
889
890     if (!s || compensation_distance < 0)
891         return AVERROR(EINVAL);
892     if (!compensation_distance && sample_delta)
893         return AVERROR(EINVAL);
894     if (!s->resample) {
895         s->flags |= SWR_FLAG_RESAMPLE;
896         ret = swr_init(s);
897         if (ret < 0)
898             return ret;
899     }
900     if (!s->resampler->set_compensation){
901         return AVERROR(EINVAL);
902     }else{
903         return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
904     }
905 }
906
907 int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
908     if(pts == INT64_MIN)
909         return s->outpts;
910
911     if (s->firstpts == AV_NOPTS_VALUE)
912         s->outpts = s->firstpts = pts;
913
914     if(s->min_compensation >= FLT_MAX) {
915         return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
916     } else {
917         int64_t delta = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate) - s->outpts + s->drop_output*(int64_t)s->in_sample_rate;
918         double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
919
920         if(fabs(fdelta) > s->min_compensation) {
921             if(s->outpts == s->firstpts || fabs(fdelta) > s->min_hard_compensation){
922                 int ret;
923                 if(delta > 0) ret = swr_inject_silence(s,  delta / s->out_sample_rate);
924                 else          ret = swr_drop_output   (s, -delta / s-> in_sample_rate);
925                 if(ret<0){
926                     av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
927                 }
928             } else if(s->soft_compensation_duration && s->max_soft_compensation) {
929                 int duration = s->out_sample_rate * s->soft_compensation_duration;
930                 double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
931                 int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
932                 av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
933                 swr_set_compensation(s, comp, duration);
934             }
935         }
936
937         return s->outpts;
938     }
939 }