FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
swresample.c
Go to the documentation of this file.
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"
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 , INT_MAX, PARAM},
55 {"in_sample_fmt" , "set input sample format" , OFFSET( in_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
56 {"osf" , "set output sample format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
57 {"out_sample_fmt" , "set output sample format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
58 {"tsf" , "set internal sample format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
59 {"internal_sample_fmt" , "set internal sample format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, 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 {
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 
167  if(!s || s->in_convert) // s needs to be allocated but not initialized
168  return AVERROR(EINVAL);
170  return 0;
171 }
172 
173 const AVClass *swr_get_class(void)
174 {
175  return &av_class;
176 }
177 
179  SwrContext *s= av_mallocz(sizeof(SwrContext));
180  if(s){
181  s->av_class= &av_class;
183  }
184  return s;
185 }
186 
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);
206  av_opt_set_int(s, "uch", 0, 0);
207  return s;
208 }
209 
211  a->fmt = fmt;
212  a->bps = av_get_bytes_per_sample(fmt);
214 }
215 
216 static void free_temp(AudioData *a){
217  av_free(a->data);
218  memset(a, 0, sizeof(*a));
219 }
220 
221 static void clear_context(SwrContext *s){
222  s->in_buffer_index= 0;
223  s->in_buffer_count= 0;
225  memset(s->in.ch, 0, sizeof(s->in.ch));
226  memset(s->out.ch, 0, sizeof(s->out.ch));
227  free_temp(&s->postin);
228  free_temp(&s->midbuf);
229  free_temp(&s->preout);
230  free_temp(&s->in_buffer);
231  free_temp(&s->silence);
232  free_temp(&s->drop_temp);
233  free_temp(&s->dither.noise);
234  free_temp(&s->dither.temp);
239 
240  s->flushed = 0;
241 }
242 
244  SwrContext *s= *ss;
245  if(s){
246  clear_context(s);
247  if (s->resampler)
248  s->resampler->free(&s->resample);
249  }
250 
251  av_freep(ss);
252 }
253 
255  int ret;
256 
257  clear_context(s);
258 
259  if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
260  av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
261  return AVERROR(EINVAL);
262  }
264  av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
265  return AVERROR(EINVAL);
266  }
267 
269  av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
270  s->in_ch_layout = 0;
271  }
272 
274  av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
275  s->out_ch_layout = 0;
276  }
277 
278  switch(s->engine){
279 #if CONFIG_LIBSOXR
280  extern struct Resampler const soxr_resampler;
281  case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
282 #endif
283  case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
284  default:
285  av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
286  return AVERROR(EINVAL);
287  }
288 
289  if(!s->used_ch_count)
290  s->used_ch_count= s->in.ch_count;
291 
292  if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
293  av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
294  s-> in_ch_layout= 0;
295  }
296 
297  if(!s-> in_ch_layout)
298  s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
299  if(!s->out_ch_layout)
301 
302  s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
303  s->rematrix_custom;
304 
308  }else if( av_get_planar_sample_fmt(s-> in_sample_fmt) == AV_SAMPLE_FMT_S32P
310  && !s->rematrix
311  && s->engine != SWR_ENGINE_SOXR){
315  }else{
316  av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
318  }
319  }
320 
325  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));
326  return AVERROR(EINVAL);
327  }
328 
329  set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
331 
333  if (!s->async && s->min_compensation >= FLT_MAX/2)
334  s->async = 1;
335  s->firstpts =
337  } else
339 
340  if (s->async) {
341  if (s->min_compensation >= FLT_MAX/2)
342  s->min_compensation = 0.001;
343  if (s->async > 1.0001) {
344  s->max_soft_compensation = s->async / (double) s->in_sample_rate;
345  }
346  }
347 
350  }else
351  s->resampler->free(&s->resample);
356  && s->resample){
357  av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
358  return -1;
359  }
360 
361 #define RSC 1 //FIXME finetune
362  if(!s-> in.ch_count)
363  s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
364  if(!s->used_ch_count)
365  s->used_ch_count= s->in.ch_count;
366  if(!s->out.ch_count)
368 
369  if(!s-> in.ch_count){
371  av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
372  return -1;
373  }
374 
375  if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
376  char l1[1024], l2[1024];
377  av_get_channel_layout_string(l1, sizeof(l1), s-> in.ch_count, s-> in_ch_layout);
378  av_get_channel_layout_string(l2, sizeof(l2), s->out.ch_count, s->out_ch_layout);
379  av_log(s, AV_LOG_ERROR, "Rematrix is needed between %s and %s "
380  "but there is not enough information to do it\n", l1, l2);
381  return -1;
382  }
383 
386  s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
387 
388  s->in_buffer= s->in;
389  s->silence = s->in;
390  s->drop_temp= s->out;
391 
392  if(!s->resample && !s->rematrix && !s->channel_map && !s->dither.method){
394  s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
395  return 0;
396  }
397 
399  s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
401  s->int_sample_fmt, s->out.ch_count, NULL, 0);
402 
403  if (!s->in_convert || !s->out_convert)
404  return AVERROR(ENOMEM);
405 
406  s->postin= s->in;
407  s->preout= s->out;
408  s->midbuf= s->in;
409 
410  if(s->channel_map){
411  s->postin.ch_count=
413  if(s->resample)
415  }
416  if(!s->resample_first){
417  s->midbuf.ch_count= s->out.ch_count;
418  if(s->resample)
419  s->in_buffer.ch_count = s->out.ch_count;
420  }
421 
425 
426  if(s->resample){
428  }
429 
430  if ((ret = swri_dither_init(s, s->out_sample_fmt, s->int_sample_fmt)) < 0)
431  return ret;
432 
433  if(s->rematrix || s->dither.method)
434  return swri_rematrix_init(s);
435 
436  return 0;
437 }
438 
440  int i, countb;
441  AudioData old;
442 
443  if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
444  return AVERROR(EINVAL);
445 
446  if(a->count >= count)
447  return 0;
448 
449  count*=2;
450 
451  countb= FFALIGN(count*a->bps, ALIGN);
452  old= *a;
453 
454  av_assert0(a->bps);
455  av_assert0(a->ch_count);
456 
457  a->data= av_mallocz(countb*a->ch_count);
458  if(!a->data)
459  return AVERROR(ENOMEM);
460  for(i=0; i<a->ch_count; i++){
461  a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
462  if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
463  }
464  if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
465  av_freep(&old.data);
466  a->count= count;
467 
468  return 1;
469 }
470 
471 static void copy(AudioData *out, AudioData *in,
472  int count){
473  av_assert0(out->planar == in->planar);
474  av_assert0(out->bps == in->bps);
475  av_assert0(out->ch_count == in->ch_count);
476  if(out->planar){
477  int ch;
478  for(ch=0; ch<out->ch_count; ch++)
479  memcpy(out->ch[ch], in->ch[ch], count*out->bps);
480  }else
481  memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
482 }
483 
484 static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
485  int i;
486  if(!in_arg){
487  memset(out->ch, 0, sizeof(out->ch));
488  }else if(out->planar){
489  for(i=0; i<out->ch_count; i++)
490  out->ch[i]= in_arg[i];
491  }else{
492  for(i=0; i<out->ch_count; i++)
493  out->ch[i]= in_arg[0] + i*out->bps;
494  }
495 }
496 
498  int i;
499  if(out->planar){
500  for(i=0; i<out->ch_count; i++)
501  in_arg[i]= out->ch[i];
502  }else{
503  in_arg[0]= out->ch[0];
504  }
505 }
506 
507 /**
508  *
509  * out may be equal in.
510  */
511 static void buf_set(AudioData *out, AudioData *in, int count){
512  int ch;
513  if(in->planar){
514  for(ch=0; ch<out->ch_count; ch++)
515  out->ch[ch]= in->ch[ch] + count*out->bps;
516  }else{
517  for(ch=out->ch_count-1; ch>=0; ch--)
518  out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
519  }
520 }
521 
522 /**
523  *
524  * @return number of samples output per channel
525  */
526 static int resample(SwrContext *s, AudioData *out_param, int out_count,
527  const AudioData * in_param, int in_count){
528  AudioData in, out, tmp;
529  int ret_sum=0;
530  int border=0;
531  int padless = ARCH_X86 && s->engine == SWR_ENGINE_SWR ? 7 : 0;
532 
533  av_assert1(s->in_buffer.ch_count == in_param->ch_count);
534  av_assert1(s->in_buffer.planar == in_param->planar);
535  av_assert1(s->in_buffer.fmt == in_param->fmt);
536 
537  tmp=out=*out_param;
538  in = *in_param;
539 
540  do{
541  int ret, size, consumed;
543  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
544  ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
545  out_count -= ret;
546  ret_sum += ret;
547  buf_set(&out, &out, ret);
548  s->in_buffer_count -= consumed;
549  s->in_buffer_index += consumed;
550 
551  if(!in_count)
552  break;
553  if(s->in_buffer_count <= border){
554  buf_set(&in, &in, -s->in_buffer_count);
555  in_count += s->in_buffer_count;
556  s->in_buffer_count=0;
557  s->in_buffer_index=0;
558  border = 0;
559  }
560  }
561 
562  if((s->flushed || in_count > padless) && !s->in_buffer_count){
563  s->in_buffer_index=0;
564  ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, FFMAX(in_count-padless, 0), &consumed);
565  out_count -= ret;
566  ret_sum += ret;
567  buf_set(&out, &out, ret);
568  in_count -= consumed;
569  buf_set(&in, &in, consumed);
570  }
571 
572  //TODO is this check sane considering the advanced copy avoidance below
573  size= s->in_buffer_index + s->in_buffer_count + in_count;
574  if( size > s->in_buffer.count
575  && s->in_buffer_count + in_count <= s->in_buffer_index){
576  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
577  copy(&s->in_buffer, &tmp, s->in_buffer_count);
578  s->in_buffer_index=0;
579  }else
580  if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
581  return ret;
582 
583  if(in_count){
584  int count= in_count;
585  if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
586 
587  buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
588  copy(&tmp, &in, /*in_*/count);
589  s->in_buffer_count += count;
590  in_count -= count;
591  border += count;
592  buf_set(&in, &in, count);
594  if(s->in_buffer_count != count || in_count)
595  continue;
596  if (padless) {
597  padless = 0;
598  continue;
599  }
600  }
601  break;
602  }while(1);
603 
604  s->resample_in_constraint= !!out_count;
605 
606  return ret_sum;
607 }
608 
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;
612  int ret/*, in_max*/;
613  AudioData preout_tmp, midbuf_tmp;
614 
615  if(s->full_convert){
616  av_assert0(!s->resample);
617  swri_audio_convert(s->full_convert, out, in, in_count);
618  return out_count;
619  }
620 
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);
623 
624  if((ret=swri_realloc_audio(&s->postin, in_count))<0)
625  return ret;
626  if(s->resample_first){
628  if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
629  return ret;
630  }else{
632  if((ret=swri_realloc_audio(&s->midbuf, in_count))<0)
633  return ret;
634  }
635  if((ret=swri_realloc_audio(&s->preout, out_count))<0)
636  return ret;
637 
638  postin= &s->postin;
639 
640  midbuf_tmp= s->midbuf;
641  midbuf= &midbuf_tmp;
642  preout_tmp= s->preout;
643  preout= &preout_tmp;
644 
645  if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
646  postin= in;
647 
648  if(s->resample_first ? !s->resample : !s->rematrix)
649  midbuf= postin;
650 
651  if(s->resample_first ? !s->rematrix : !s->resample)
652  preout= midbuf;
653 
654  if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar
656  if(preout==in){
657  out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
658  av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
659  copy(out, in, out_count);
660  return out_count;
661  }
662  else if(preout==postin) preout= midbuf= postin= out;
663  else if(preout==midbuf) preout= midbuf= out;
664  else preout= out;
665  }
666 
667  if(in != postin){
668  swri_audio_convert(s->in_convert, postin, in, in_count);
669  }
670 
671  if(s->resample_first){
672  if(postin != midbuf)
673  out_count= resample(s, midbuf, out_count, postin, in_count);
674  if(midbuf != preout)
675  swri_rematrix(s, preout, midbuf, out_count, preout==out);
676  }else{
677  if(postin != midbuf)
678  swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
679  if(midbuf != preout)
680  out_count= resample(s, preout, out_count, midbuf, in_count);
681  }
682 
683  if(preout != out && out_count){
684  AudioData *conv_src = preout;
685  if(s->dither.method){
686  int ch;
687  int dither_count= FFMAX(out_count, 1<<16);
688 
689  if (preout == in) {
690  conv_src = &s->dither.temp;
691  if((ret=swri_realloc_audio(&s->dither.temp, dither_count))<0)
692  return ret;
693  }
694 
695  if((ret=swri_realloc_audio(&s->dither.noise, dither_count))<0)
696  return ret;
697  if(ret)
698  for(ch=0; ch<s->dither.noise.ch_count; ch++)
699  swri_get_dither(s, s->dither.noise.ch[ch], s->dither.noise.count, 12345678913579<<ch, s->dither.noise.fmt);
700  av_assert0(s->dither.noise.ch_count == preout->ch_count);
701 
702  if(s->dither.noise_pos + out_count > s->dither.noise.count)
703  s->dither.noise_pos = 0;
704 
705  if (s->dither.method < SWR_DITHER_NS){
706  if (s->mix_2_1_simd) {
707  int len1= out_count&~15;
708  int off = len1 * preout->bps;
709 
710  if(len1)
711  for(ch=0; ch<preout->ch_count; ch++)
712  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);
713  if(out_count != len1)
714  for(ch=0; ch<preout->ch_count; ch++)
715  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  } else {
717  for(ch=0; ch<preout->ch_count; ch++)
718  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);
719  }
720  } else {
721  switch(s->int_sample_fmt) {
722  case AV_SAMPLE_FMT_S16P :swri_noise_shaping_int16(s, conv_src, preout, &s->dither.noise, out_count); break;
723  case AV_SAMPLE_FMT_S32P :swri_noise_shaping_int32(s, conv_src, preout, &s->dither.noise, out_count); break;
724  case AV_SAMPLE_FMT_FLTP :swri_noise_shaping_float(s, conv_src, preout, &s->dither.noise, out_count); break;
725  case AV_SAMPLE_FMT_DBLP :swri_noise_shaping_double(s,conv_src, preout, &s->dither.noise, out_count); break;
726  }
727  }
728  s->dither.noise_pos += out_count;
729  }
730 //FIXME packed doesn't need more than 1 chan here!
731  swri_audio_convert(s->out_convert, out, conv_src, out_count);
732  }
733  return out_count;
734 }
735 
737  return !!s->in_buffer.ch_count;
738 }
739 
740 int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
741  const uint8_t *in_arg [SWR_CH_MAX], int in_count){
742  AudioData * in= &s->in;
743  AudioData *out= &s->out;
744 
745  if (!swr_is_initialized(s)) {
746  av_log(s, AV_LOG_ERROR, "Context has not been initialized\n");
747  return AVERROR(EINVAL);
748  }
749 
750  while(s->drop_output > 0){
751  int ret;
752  uint8_t *tmp_arg[SWR_CH_MAX];
753 #define MAX_DROP_STEP 16384
755  return ret;
756 
757  reversefill_audiodata(&s->drop_temp, tmp_arg);
758  s->drop_output *= -1; //FIXME find a less hackish solution
759  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
760  s->drop_output *= -1;
761  in_count = 0;
762  if(ret>0) {
763  s->drop_output -= ret;
764  continue;
765  }
766 
767  if(s->drop_output || !out_arg)
768  return 0;
769  }
770 
771  if(!in_arg){
772  if(s->resample){
773  if (!s->flushed)
774  s->resampler->flush(s);
775  s->resample_in_constraint = 0;
776  s->flushed = 1;
777  }else if(!s->in_buffer_count){
778  return 0;
779  }
780  }else
781  fill_audiodata(in , (void*)in_arg);
782 
783  fill_audiodata(out, out_arg);
784 
785  if(s->resample){
786  int ret = swr_convert_internal(s, out, out_count, in, in_count);
787  if(ret>0 && !s->drop_output)
788  s->outpts += ret * (int64_t)s->in_sample_rate;
789  return ret;
790  }else{
791  AudioData tmp= *in;
792  int ret2=0;
793  int ret, size;
794  size = FFMIN(out_count, s->in_buffer_count);
795  if(size){
796  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
797  ret= swr_convert_internal(s, out, size, &tmp, size);
798  if(ret<0)
799  return ret;
800  ret2= ret;
801  s->in_buffer_count -= ret;
802  s->in_buffer_index += ret;
803  buf_set(out, out, ret);
804  out_count -= ret;
805  if(!s->in_buffer_count)
806  s->in_buffer_index = 0;
807  }
808 
809  if(in_count){
810  size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
811 
812  if(in_count > out_count) { //FIXME move after swr_convert_internal
813  if( size > s->in_buffer.count
814  && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
815  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
816  copy(&s->in_buffer, &tmp, s->in_buffer_count);
817  s->in_buffer_index=0;
818  }else
819  if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
820  return ret;
821  }
822 
823  if(out_count){
824  size = FFMIN(in_count, out_count);
825  ret= swr_convert_internal(s, out, size, in, size);
826  if(ret<0)
827  return ret;
828  buf_set(in, in, ret);
829  in_count -= ret;
830  ret2 += ret;
831  }
832  if(in_count){
833  buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
834  copy(&tmp, in, in_count);
835  s->in_buffer_count += in_count;
836  }
837  }
838  if(ret2>0 && !s->drop_output)
839  s->outpts += ret2 * (int64_t)s->in_sample_rate;
840  return ret2;
841  }
842 }
843 
844 int swr_drop_output(struct SwrContext *s, int count){
845  s->drop_output += count;
846 
847  if(s->drop_output <= 0)
848  return 0;
849 
850  av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
851  return swr_convert(s, NULL, s->drop_output, NULL, 0);
852 }
853 
855  int ret, i;
856  uint8_t *tmp_arg[SWR_CH_MAX];
857 
858  if(count <= 0)
859  return 0;
860 
861 #define MAX_SILENCE_STEP 16384
862  while (count > MAX_SILENCE_STEP) {
863  if ((ret = swr_inject_silence(s, MAX_SILENCE_STEP)) < 0)
864  return ret;
865  count -= MAX_SILENCE_STEP;
866  }
867 
868  if((ret=swri_realloc_audio(&s->silence, count))<0)
869  return ret;
870 
871  if(s->silence.planar) for(i=0; i<s->silence.ch_count; i++) {
872  memset(s->silence.ch[i], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps);
873  } else
874  memset(s->silence.ch[0], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps*s->silence.ch_count);
875 
876  reversefill_audiodata(&s->silence, tmp_arg);
877  av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
878  ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
879  return ret;
880 }
881 
882 int64_t swr_get_delay(struct SwrContext *s, int64_t base){
883  if (s->resampler && s->resample){
884  return s->resampler->get_delay(s, base);
885  }else{
886  return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
887  }
888 }
889 
890 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
891  int ret;
892 
893  if (!s || compensation_distance < 0)
894  return AVERROR(EINVAL);
895  if (!compensation_distance && sample_delta)
896  return AVERROR(EINVAL);
897  if (!s->resample) {
898  s->flags |= SWR_FLAG_RESAMPLE;
899  ret = swr_init(s);
900  if (ret < 0)
901  return ret;
902  }
903  if (!s->resampler->set_compensation){
904  return AVERROR(EINVAL);
905  }else{
906  return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
907  }
908 }
909 
910 int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
911  if(pts == INT64_MIN)
912  return s->outpts;
913 
914  if (s->firstpts == AV_NOPTS_VALUE)
915  s->outpts = s->firstpts = pts;
916 
917  if(s->min_compensation >= FLT_MAX) {
918  return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
919  } else {
920  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;
921  double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
922 
923  if(fabs(fdelta) > s->min_compensation) {
924  if(s->outpts == s->firstpts || fabs(fdelta) > s->min_hard_compensation){
925  int ret;
926  if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
927  else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
928  if(ret<0){
929  av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
930  }
933  double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
934  int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
935  av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
936  swr_set_compensation(s, comp, duration);
937  }
938  }
939 
940  return s->outpts;
941  }
942 }