2 * Copyright (C) 2011-2013 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"},
82 {"lipshitz" , "select lipshitz noise shaping dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_LIPSHITZ}, INT_MIN, INT_MAX, PARAM, "dither_method"},
83 {"shibata" , "select shibata noise shaping dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_SHIBATA }, INT_MIN, INT_MAX, PARAM, "dither_method"},
84 {"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"},
85 {"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"},
86 {"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"},
87 {"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"},
88 {"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 {"filter_size" , "set swr resampling filter size", OFFSET(filter_size) , AV_OPT_TYPE_INT , {.i64=32 }, 0 , INT_MAX , PARAM },
91 {"phase_shift" , "set swr resampling phase shift", OFFSET(phase_shift) , AV_OPT_TYPE_INT , {.i64=10 }, 0 , 24 , PARAM },
92 {"linear_interp" , "enable linear interpolation" , OFFSET(linear_interp) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
93 {"cutoff" , "set cutoff frequency ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0. }, 0 , 1 , PARAM },
95 /* duplicate option in order to work with avconv */
96 {"resample_cutoff" , "set cutoff frequency ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0. }, 0 , 1 , PARAM },
98 {"resampler" , "set resampling Engine" , OFFSET(engine) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_ENGINE_NB-1, PARAM, "resampler"},
99 {"swr" , "select SW Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SWR }, INT_MIN, INT_MAX , PARAM, "resampler"},
100 {"soxr" , "select SoX Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SOXR }, INT_MIN, INT_MAX , PARAM, "resampler"},
101 {"precision" , "set soxr resampling precision (in bits)"
102 , OFFSET(precision) , AV_OPT_TYPE_DOUBLE,{.dbl=20.0 }, 15.0 , 33.0 , PARAM },
103 {"cheby" , "enable soxr Chebyshev passband & higher-precision irrational ratio approximation"
104 , OFFSET(cheby) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
105 {"min_comp" , "set minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
106 , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX }, 0 , FLT_MAX , PARAM },
107 {"min_hard_comp" , "set minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
108 , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1 }, 0 , INT_MAX , PARAM },
109 {"comp_duration" , "set duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
110 , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1 }, 0 , INT_MAX , PARAM },
111 {"max_soft_comp" , "set maximum factor by which data is stretched/squeezed to make it match the timestamps."
112 , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
113 {"async" , "simplified 1 parameter audio timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum stretch/squeeze in samples per second)"
114 , OFFSET(async) , AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
115 {"first_pts" , "Assume the first pts should be this value (in samples)."
116 , OFFSET(firstpts_in_samples), AV_OPT_TYPE_INT64 ,{.i64=AV_NOPTS_VALUE }, INT64_MIN,INT64_MAX, PARAM },
118 { "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" },
119 { "none", "select none", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
120 { "dolby", "select Dolby", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
121 { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
123 { "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" },
124 { "cubic" , "select cubic" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC }, INT_MIN, INT_MAX, PARAM, "filter_type" },
125 { "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" },
126 { "kaiser" , "select Kaiser Windowed Sinc" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER }, INT_MIN, INT_MAX, PARAM, "filter_type" },
128 { "kaiser_beta" , "set swr Kaiser Window Beta" , OFFSET(kaiser_beta) , AV_OPT_TYPE_INT , {.i64=9 }, 2 , 16 , PARAM },
130 { "output_sample_bits" , "" , OFFSET(dither.output_sample_bits) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 64 , 0 },
134 static const char* context_to_name(void* ptr) {
138 static const AVClass av_class = {
139 .class_name = "SWResampler",
140 .item_name = context_to_name,
142 .version = LIBAVUTIL_VERSION_INT,
143 .log_level_offset_offset = OFFSET(log_level_offset),
144 .parent_log_context_offset = OFFSET(log_ctx),
145 .category = AV_CLASS_CATEGORY_SWRESAMPLER,
148 unsigned swresample_version(void)
150 av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
151 return LIBSWRESAMPLE_VERSION_INT;
154 const char *swresample_configuration(void)
156 return FFMPEG_CONFIGURATION;
159 const char *swresample_license(void)
161 #define LICENSE_PREFIX "libswresample license: "
162 return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
165 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
166 if(!s || s->in_convert) // s needs to be allocated but not initialized
167 return AVERROR(EINVAL);
168 s->channel_map = channel_map;
172 const AVClass *swr_get_class(void)
177 av_cold struct SwrContext *swr_alloc(void){
178 SwrContext *s= av_mallocz(sizeof(SwrContext));
180 s->av_class= &av_class;
181 av_opt_set_defaults(s);
186 struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
187 int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
188 int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate,
189 int log_offset, void *log_ctx){
190 if(!s) s= swr_alloc();
193 s->log_level_offset= log_offset;
196 av_opt_set_int(s, "ocl", out_ch_layout, 0);
197 av_opt_set_int(s, "osf", out_sample_fmt, 0);
198 av_opt_set_int(s, "osr", out_sample_rate, 0);
199 av_opt_set_int(s, "icl", in_ch_layout, 0);
200 av_opt_set_int(s, "isf", in_sample_fmt, 0);
201 av_opt_set_int(s, "isr", in_sample_rate, 0);
202 av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE, 0);
203 av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
204 av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
205 av_opt_set_int(s, "uch", 0, 0);
209 static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
211 a->bps = av_get_bytes_per_sample(fmt);
212 a->planar= av_sample_fmt_is_planar(fmt);
215 static void free_temp(AudioData *a){
217 memset(a, 0, sizeof(*a));
220 av_cold void swr_free(SwrContext **ss){
223 free_temp(&s->postin);
224 free_temp(&s->midbuf);
225 free_temp(&s->preout);
226 free_temp(&s->in_buffer);
227 free_temp(&s->silence);
228 free_temp(&s->drop_temp);
229 free_temp(&s->dither.noise);
230 free_temp(&s->dither.temp);
231 swri_audio_convert_free(&s-> in_convert);
232 swri_audio_convert_free(&s->out_convert);
233 swri_audio_convert_free(&s->full_convert);
235 s->resampler->free(&s->resample);
236 swri_rematrix_free(s);
242 av_cold int swr_init(struct SwrContext *s){
244 s->in_buffer_index= 0;
245 s->in_buffer_count= 0;
246 s->resample_in_constraint= 0;
247 free_temp(&s->postin);
248 free_temp(&s->midbuf);
249 free_temp(&s->preout);
250 free_temp(&s->in_buffer);
251 free_temp(&s->silence);
252 free_temp(&s->drop_temp);
253 free_temp(&s->dither.noise);
254 free_temp(&s->dither.temp);
255 memset(s->in.ch, 0, sizeof(s->in.ch));
256 memset(s->out.ch, 0, sizeof(s->out.ch));
257 swri_audio_convert_free(&s-> in_convert);
258 swri_audio_convert_free(&s->out_convert);
259 swri_audio_convert_free(&s->full_convert);
260 swri_rematrix_free(s);
264 if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
265 av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
266 return AVERROR(EINVAL);
268 if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
269 av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
270 return AVERROR(EINVAL);
273 if(av_get_channel_layout_nb_channels(s-> in_ch_layout) > SWR_CH_MAX) {
274 av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
278 if(av_get_channel_layout_nb_channels(s->out_ch_layout) > SWR_CH_MAX) {
279 av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
280 s->out_ch_layout = 0;
285 extern struct Resampler const soxr_resampler;
286 case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
288 case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
290 av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
291 return AVERROR(EINVAL);
294 if(!s->used_ch_count)
295 s->used_ch_count= s->in.ch_count;
297 if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
298 av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
302 if(!s-> in_ch_layout)
303 s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
304 if(!s->out_ch_layout)
305 s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);
307 s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
310 if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
311 if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
312 s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
313 }else if( av_get_planar_sample_fmt(s-> in_sample_fmt) == AV_SAMPLE_FMT_S32P
314 && av_get_planar_sample_fmt(s->out_sample_fmt) == AV_SAMPLE_FMT_S32P
316 && s->engine != SWR_ENGINE_SOXR){
317 s->int_sample_fmt= AV_SAMPLE_FMT_S32P;
318 }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
319 s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
321 av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
322 s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
326 if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
327 &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
328 &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
329 &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
330 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));
331 return AVERROR(EINVAL);
334 set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
335 set_audiodata_fmt(&s->out, s->out_sample_fmt);
337 if (s->firstpts_in_samples != AV_NOPTS_VALUE) {
338 if (!s->async && s->min_compensation >= FLT_MAX/2)
341 s->outpts = s->firstpts_in_samples * s->out_sample_rate;
343 s->firstpts = AV_NOPTS_VALUE;
346 if (s->min_compensation >= FLT_MAX/2)
347 s->min_compensation = 0.001;
348 if (s->async > 1.0001) {
349 s->max_soft_compensation = s->async / (double) s->in_sample_rate;
353 if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
354 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 s->resampler->free(&s->resample);
357 if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
358 && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
359 && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
360 && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
362 av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
366 #define RSC 1 //FIXME finetune
368 s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
369 if(!s->used_ch_count)
370 s->used_ch_count= s->in.ch_count;
372 s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);
374 if(!s-> in.ch_count){
375 av_assert0(!s->in_ch_layout);
376 av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
380 if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
381 char l1[1024], l2[1024];
382 av_get_channel_layout_string(l1, sizeof(l1), s-> in.ch_count, s-> in_ch_layout);
383 av_get_channel_layout_string(l2, sizeof(l2), s->out.ch_count, s->out_ch_layout);
384 av_log(s, AV_LOG_ERROR, "Rematrix is needed between %s and %s "
385 "but there is not enough information to do it\n", l1, l2);
389 av_assert0(s->used_ch_count);
390 av_assert0(s->out.ch_count);
391 s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
395 s->drop_temp= s->out;
397 if(!s->resample && !s->rematrix && !s->channel_map && !s->dither.method){
398 s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
399 s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
403 s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
404 s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
405 s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
406 s->int_sample_fmt, s->out.ch_count, NULL, 0);
408 if (!s->in_convert || !s->out_convert)
409 return AVERROR(ENOMEM);
417 s->midbuf.ch_count= s->used_ch_count;
419 s->in_buffer.ch_count= s->used_ch_count;
421 if(!s->resample_first){
422 s->midbuf.ch_count= s->out.ch_count;
424 s->in_buffer.ch_count = s->out.ch_count;
427 set_audiodata_fmt(&s->postin, s->int_sample_fmt);
428 set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
429 set_audiodata_fmt(&s->preout, s->int_sample_fmt);
432 set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
435 if ((ret = swri_dither_init(s, s->out_sample_fmt, s->int_sample_fmt)) < 0)
438 if(s->rematrix || s->dither.method)
439 return swri_rematrix_init(s);
444 int swri_realloc_audio(AudioData *a, int count){
448 if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
449 return AVERROR(EINVAL);
451 if(a->count >= count)
456 countb= FFALIGN(count*a->bps, ALIGN);
460 av_assert0(a->ch_count);
462 a->data= av_mallocz(countb*a->ch_count);
464 return AVERROR(ENOMEM);
465 for(i=0; i<a->ch_count; i++){
466 a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
467 if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
469 if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
476 static void copy(AudioData *out, AudioData *in,
478 av_assert0(out->planar == in->planar);
479 av_assert0(out->bps == in->bps);
480 av_assert0(out->ch_count == in->ch_count);
483 for(ch=0; ch<out->ch_count; ch++)
484 memcpy(out->ch[ch], in->ch[ch], count*out->bps);
486 memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
489 static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
492 memset(out->ch, 0, sizeof(out->ch));
493 }else if(out->planar){
494 for(i=0; i<out->ch_count; i++)
495 out->ch[i]= in_arg[i];
497 for(i=0; i<out->ch_count; i++)
498 out->ch[i]= in_arg[0] + i*out->bps;
502 static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
505 for(i=0; i<out->ch_count; i++)
506 in_arg[i]= out->ch[i];
508 in_arg[0]= out->ch[0];
514 * out may be equal in.
516 static void buf_set(AudioData *out, AudioData *in, int count){
519 for(ch=0; ch<out->ch_count; ch++)
520 out->ch[ch]= in->ch[ch] + count*out->bps;
522 for(ch=out->ch_count-1; ch>=0; ch--)
523 out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
529 * @return number of samples output per channel
531 static int resample(SwrContext *s, AudioData *out_param, int out_count,
532 const AudioData * in_param, int in_count){
533 AudioData in, out, tmp;
537 av_assert1(s->in_buffer.ch_count == in_param->ch_count);
538 av_assert1(s->in_buffer.planar == in_param->planar);
539 av_assert1(s->in_buffer.fmt == in_param->fmt);
545 int ret, size, consumed;
546 if(!s->resample_in_constraint && s->in_buffer_count){
547 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
548 ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
551 buf_set(&out, &out, ret);
552 s->in_buffer_count -= consumed;
553 s->in_buffer_index += consumed;
557 if(s->in_buffer_count <= border){
558 buf_set(&in, &in, -s->in_buffer_count);
559 in_count += s->in_buffer_count;
560 s->in_buffer_count=0;
561 s->in_buffer_index=0;
566 if((s->flushed || in_count) && !s->in_buffer_count){
567 s->in_buffer_index=0;
568 ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, in_count, &consumed);
571 buf_set(&out, &out, ret);
572 in_count -= consumed;
573 buf_set(&in, &in, consumed);
576 //TODO is this check sane considering the advanced copy avoidance below
577 size= s->in_buffer_index + s->in_buffer_count + in_count;
578 if( size > s->in_buffer.count
579 && s->in_buffer_count + in_count <= s->in_buffer_index){
580 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
581 copy(&s->in_buffer, &tmp, s->in_buffer_count);
582 s->in_buffer_index=0;
584 if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
589 if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
591 buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
592 copy(&tmp, &in, /*in_*/count);
593 s->in_buffer_count += count;
596 buf_set(&in, &in, count);
597 s->resample_in_constraint= 0;
598 if(s->in_buffer_count != count || in_count)
604 s->resample_in_constraint= !!out_count;
609 static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
610 AudioData *in , int in_count){
611 AudioData *postin, *midbuf, *preout;
613 AudioData preout_tmp, midbuf_tmp;
616 av_assert0(!s->resample);
617 swri_audio_convert(s->full_convert, out, in, in_count);
621 // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
622 // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
624 if((ret=swri_realloc_audio(&s->postin, in_count))<0)
626 if(s->resample_first){
627 av_assert0(s->midbuf.ch_count == s->used_ch_count);
628 if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
631 av_assert0(s->midbuf.ch_count == s->out.ch_count);
632 if((ret=swri_realloc_audio(&s->midbuf, in_count))<0)
635 if((ret=swri_realloc_audio(&s->preout, out_count))<0)
640 midbuf_tmp= s->midbuf;
642 preout_tmp= s->preout;
645 if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
648 if(s->resample_first ? !s->resample : !s->rematrix)
651 if(s->resample_first ? !s->rematrix : !s->resample)
654 if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar){
656 out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
657 av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
658 copy(out, in, out_count);
661 else if(preout==postin) preout= midbuf= postin= out;
662 else if(preout==midbuf) preout= midbuf= out;
667 swri_audio_convert(s->in_convert, postin, in, in_count);
670 if(s->resample_first){
672 out_count= resample(s, midbuf, out_count, postin, in_count);
674 swri_rematrix(s, preout, midbuf, out_count, preout==out);
677 swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
679 out_count= resample(s, preout, out_count, midbuf, in_count);
682 if(preout != out && out_count){
683 AudioData *conv_src = preout;
684 if(s->dither.method){
686 int dither_count= FFMAX(out_count, 1<<16);
689 conv_src = &s->dither.temp;
690 if((ret=swri_realloc_audio(&s->dither.temp, dither_count))<0)
694 if((ret=swri_realloc_audio(&s->dither.noise, dither_count))<0)
697 for(ch=0; ch<s->dither.noise.ch_count; ch++)
698 swri_get_dither(s, s->dither.noise.ch[ch], s->dither.noise.count, 12345678913579<<ch, s->dither.noise.fmt);
699 av_assert0(s->dither.noise.ch_count == preout->ch_count);
701 if(s->dither.noise_pos + out_count > s->dither.noise.count)
702 s->dither.noise_pos = 0;
704 if (s->dither.method < SWR_DITHER_NS){
705 if (s->mix_2_1_simd) {
706 int len1= out_count&~15;
707 int off = len1 * preout->bps;
710 for(ch=0; ch<preout->ch_count; ch++)
711 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_one, 0, 0, len1);
712 if(out_count != len1)
713 for(ch=0; ch<preout->ch_count; ch++)
714 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);
716 for(ch=0; ch<preout->ch_count; ch++)
717 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);
720 switch(s->int_sample_fmt) {
721 case AV_SAMPLE_FMT_S16P :swri_noise_shaping_int16(s, conv_src, preout, &s->dither.noise, out_count); break;
722 case AV_SAMPLE_FMT_S32P :swri_noise_shaping_int32(s, conv_src, preout, &s->dither.noise, out_count); break;
723 case AV_SAMPLE_FMT_FLTP :swri_noise_shaping_float(s, conv_src, preout, &s->dither.noise, out_count); break;
724 case AV_SAMPLE_FMT_DBLP :swri_noise_shaping_double(s,conv_src, preout, &s->dither.noise, out_count); break;
727 s->dither.noise_pos += out_count;
729 //FIXME packed doesnt need more than 1 chan here!
730 swri_audio_convert(s->out_convert, out, conv_src, out_count);
735 int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
736 const uint8_t *in_arg [SWR_CH_MAX], int in_count){
737 AudioData * in= &s->in;
738 AudioData *out= &s->out;
740 while(s->drop_output > 0){
742 uint8_t *tmp_arg[SWR_CH_MAX];
743 #define MAX_DROP_STEP 16384
744 if((ret=swri_realloc_audio(&s->drop_temp, FFMIN(s->drop_output, MAX_DROP_STEP)))<0)
747 reversefill_audiodata(&s->drop_temp, tmp_arg);
748 s->drop_output *= -1; //FIXME find a less hackish solution
749 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 doesnt matter
750 s->drop_output *= -1;
753 s->drop_output -= ret;
757 if(s->drop_output || !out_arg)
764 s->resampler->flush(s);
765 s->resample_in_constraint = 0;
767 }else if(!s->in_buffer_count){
771 fill_audiodata(in , (void*)in_arg);
773 fill_audiodata(out, out_arg);
776 int ret = swr_convert_internal(s, out, out_count, in, in_count);
777 if(ret>0 && !s->drop_output)
778 s->outpts += ret * (int64_t)s->in_sample_rate;
784 size = FFMIN(out_count, s->in_buffer_count);
786 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
787 ret= swr_convert_internal(s, out, size, &tmp, size);
791 s->in_buffer_count -= ret;
792 s->in_buffer_index += ret;
793 buf_set(out, out, ret);
795 if(!s->in_buffer_count)
796 s->in_buffer_index = 0;
800 size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
802 if(in_count > out_count) { //FIXME move after swr_convert_internal
803 if( size > s->in_buffer.count
804 && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
805 buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
806 copy(&s->in_buffer, &tmp, s->in_buffer_count);
807 s->in_buffer_index=0;
809 if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
814 size = FFMIN(in_count, out_count);
815 ret= swr_convert_internal(s, out, size, in, size);
818 buf_set(in, in, ret);
823 buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
824 copy(&tmp, in, in_count);
825 s->in_buffer_count += in_count;
828 if(ret2>0 && !s->drop_output)
829 s->outpts += ret2 * (int64_t)s->in_sample_rate;
834 int swr_drop_output(struct SwrContext *s, int count){
835 s->drop_output += count;
837 if(s->drop_output <= 0)
840 av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
841 return swr_convert(s, NULL, s->drop_output, NULL, 0);
844 int swr_inject_silence(struct SwrContext *s, int count){
846 uint8_t *tmp_arg[SWR_CH_MAX];
851 #define MAX_SILENCE_STEP 16384
852 while (count > MAX_SILENCE_STEP) {
853 if ((ret = swr_inject_silence(s, MAX_SILENCE_STEP)) < 0)
855 count -= MAX_SILENCE_STEP;
858 if((ret=swri_realloc_audio(&s->silence, count))<0)
861 if(s->silence.planar) for(i=0; i<s->silence.ch_count; i++) {
862 memset(s->silence.ch[i], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps);
864 memset(s->silence.ch[0], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps*s->silence.ch_count);
866 reversefill_audiodata(&s->silence, tmp_arg);
867 av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
868 ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
872 int64_t swr_get_delay(struct SwrContext *s, int64_t base){
873 if (s->resampler && s->resample){
874 return s->resampler->get_delay(s, base);
876 return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
880 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
883 if (!s || compensation_distance < 0)
884 return AVERROR(EINVAL);
885 if (!compensation_distance && sample_delta)
886 return AVERROR(EINVAL);
888 s->flags |= SWR_FLAG_RESAMPLE;
893 if (!s->resampler->set_compensation){
894 return AVERROR(EINVAL);
896 return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
900 int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
904 if (s->firstpts == AV_NOPTS_VALUE)
905 s->outpts = s->firstpts = pts;
907 if(s->min_compensation >= FLT_MAX) {
908 return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
910 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;
911 double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
913 if(fabs(fdelta) > s->min_compensation) {
914 if(s->outpts == s->firstpts || fabs(fdelta) > s->min_hard_compensation){
916 if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
917 else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
919 av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
921 } else if(s->soft_compensation_duration && s->max_soft_compensation) {
922 int duration = s->out_sample_rate * s->soft_compensation_duration;
923 double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
924 int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
925 av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
926 swr_set_compensation(s, comp, duration);