FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
swresample.c
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2011-2012 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 , 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},
71 
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"},
75 
76 {"dither_scale" , "set dither scale" , OFFSET(dither_scale ), AV_OPT_TYPE_FLOAT, {.dbl=1 }, 0 , INT_MAX , PARAM},
77 
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 
83 {"filter_size" , "set swr resampling filter size", OFFSET(filter_size) , AV_OPT_TYPE_INT , {.i64=32 }, 0 , INT_MAX , PARAM },
84 {"phase_shift" , "set swr resampling phase shift", OFFSET(phase_shift) , AV_OPT_TYPE_INT , {.i64=10 }, 0 , 30 , PARAM },
85 {"linear_interp" , "enable linear interpolation" , OFFSET(linear_interp) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
86 {"cutoff" , "set cutoff frequency ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0. }, 0 , 1 , PARAM },
87 {"resampler" , "set resampling Engine" , OFFSET(engine) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_ENGINE_NB-1, PARAM, "resampler"},
88 {"swr" , "select SW Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SWR }, INT_MIN, INT_MAX , PARAM, "resampler"},
89 {"soxr" , "select SoX Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SOXR }, INT_MIN, INT_MAX , PARAM, "resampler"},
90 {"precision" , "set soxr resampling precision (in bits)"
91  , OFFSET(precision) , AV_OPT_TYPE_DOUBLE,{.dbl=20.0 }, 15.0 , 33.0 , PARAM },
92 {"cheby" , "enable soxr Chebyshev passband & higher-precision irrational ratio approximation"
93  , OFFSET(cheby) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
94 {"min_comp" , "set minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
95  , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX }, 0 , FLT_MAX , PARAM },
96 {"min_hard_comp" , "set minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
97  , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1 }, 0 , INT_MAX , PARAM },
98 {"comp_duration" , "set duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
99  , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1 }, 0 , INT_MAX , PARAM },
100 {"max_soft_comp" , "set maximum factor by which data is stretched/squeezed to make it match the timestamps."
101  , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
102 {"async" , "simplified 1 parameter audio timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum stretch/squeeze in samples per second)"
103  , OFFSET(async) , AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
104 
105 { "matrix_encoding" , "set matrixed stereo encoding" , OFFSET(matrix_encoding), AV_OPT_TYPE_INT ,{.i64 = AV_MATRIX_ENCODING_NONE}, AV_MATRIX_ENCODING_NONE, AV_MATRIX_ENCODING_NB-1, PARAM, "matrix_encoding" },
106  { "none", "select none", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
107  { "dolby", "select Dolby", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
108  { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
109 
110 { "filter_type" , "select swr filter type" , OFFSET(filter_type) , AV_OPT_TYPE_INT , { .i64 = SWR_FILTER_TYPE_KAISER }, SWR_FILTER_TYPE_CUBIC, SWR_FILTER_TYPE_KAISER, PARAM, "filter_type" },
111  { "cubic" , "select cubic" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC }, INT_MIN, INT_MAX, PARAM, "filter_type" },
112  { "blackman_nuttall", "select Blackman Nuttall Windowed Sinc", 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_BLACKMAN_NUTTALL }, INT_MIN, INT_MAX, PARAM, "filter_type" },
113  { "kaiser" , "select Kaiser Windowed Sinc" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER }, INT_MIN, INT_MAX, PARAM, "filter_type" },
114 
115 { "kaiser_beta" , "set swr Kaiser Window Beta" , OFFSET(kaiser_beta) , AV_OPT_TYPE_INT , {.i64=9 }, 2 , 16 , PARAM },
116 
117 {0}
118 };
119 
120 static const char* context_to_name(void* ptr) {
121  return "SWR";
122 }
123 
124 static const AVClass av_class = {
125  .class_name = "SWResampler",
126  .item_name = context_to_name,
127  .option = options,
128  .version = LIBAVUTIL_VERSION_INT,
129  .log_level_offset_offset = OFFSET(log_level_offset),
130  .parent_log_context_offset = OFFSET(log_ctx),
131  .category = AV_CLASS_CATEGORY_SWRESAMPLER,
132 };
133 
134 unsigned swresample_version(void)
135 {
138 }
139 
140 const char *swresample_configuration(void)
141 {
142  return FFMPEG_CONFIGURATION;
143 }
144 
145 const char *swresample_license(void)
146 {
147 #define LICENSE_PREFIX "libswresample license: "
148  return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
149 }
150 
151 int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
152  if(!s || s->in_convert) // s needs to be allocated but not initialized
153  return AVERROR(EINVAL);
155  return 0;
156 }
157 
158 const AVClass *swr_get_class(void)
159 {
160  return &av_class;
161 }
162 
164  SwrContext *s= av_mallocz(sizeof(SwrContext));
165  if(s){
166  s->av_class= &av_class;
168  }
169  return s;
170 }
171 
175  int log_offset, void *log_ctx){
176  if(!s) s= swr_alloc();
177  if(!s) return NULL;
178 
179  s->log_level_offset= log_offset;
180  s->log_ctx= log_ctx;
181 
182  av_opt_set_int(s, "ocl", out_ch_layout, 0);
183  av_opt_set_int(s, "osf", out_sample_fmt, 0);
184  av_opt_set_int(s, "osr", out_sample_rate, 0);
185  av_opt_set_int(s, "icl", in_ch_layout, 0);
186  av_opt_set_int(s, "isf", in_sample_fmt, 0);
187  av_opt_set_int(s, "isr", in_sample_rate, 0);
188  av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE, 0);
189  av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
191  av_opt_set_int(s, "uch", 0, 0);
192  return s;
193 }
194 
196  a->fmt = fmt;
197  a->bps = av_get_bytes_per_sample(fmt);
199 }
200 
201 static void free_temp(AudioData *a){
202  av_free(a->data);
203  memset(a, 0, sizeof(*a));
204 }
205 
207  SwrContext *s= *ss;
208  if(s){
209  free_temp(&s->postin);
210  free_temp(&s->midbuf);
211  free_temp(&s->preout);
212  free_temp(&s->in_buffer);
213  free_temp(&s->dither);
217  if (s->resampler)
218  s->resampler->free(&s->resample);
220  }
221 
222  av_freep(ss);
223 }
224 
225 av_cold int swr_init(struct SwrContext *s){
226  s->in_buffer_index= 0;
227  s->in_buffer_count= 0;
229  free_temp(&s->postin);
230  free_temp(&s->midbuf);
231  free_temp(&s->preout);
232  free_temp(&s->in_buffer);
233  free_temp(&s->dither);
234  memset(s->in.ch, 0, sizeof(s->in.ch));
235  memset(s->out.ch, 0, sizeof(s->out.ch));
240 
241  s->flushed = 0;
242 
243  if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
244  av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
245  return AVERROR(EINVAL);
246  }
248  av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
249  return AVERROR(EINVAL);
250  }
251 
253  av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
254  s->in_ch_layout = 0;
255  }
256 
258  av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
259  s->out_ch_layout = 0;
260  }
261 
267  }else{
268  av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
270  }
271  }
272 
277  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));
278  return AVERROR(EINVAL);
279  }
280 
281  switch(s->engine){
282 #if CONFIG_LIBSOXR
283  extern struct Resampler const soxr_resampler;
284  case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
285 #endif
286  case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
287  default:
288  av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
289  return AVERROR(EINVAL);
290  }
291 
292  set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
294 
295  if (s->async) {
296  if (s->min_compensation >= FLT_MAX/2)
297  s->min_compensation = 0.001;
298  if (s->async > 1.0001) {
299  s->max_soft_compensation = s->async / (double) s->in_sample_rate;
300  }
301  }
302 
305  }else
306  s->resampler->free(&s->resample);
311  && s->resample){
312  av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
313  return -1;
314  }
315 
316  if(!s->used_ch_count)
317  s->used_ch_count= s->in.ch_count;
318 
319  if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
320  av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
321  s-> in_ch_layout= 0;
322  }
323 
324  if(!s-> in_ch_layout)
325  s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
326  if(!s->out_ch_layout)
328 
329  s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
330  s->rematrix_custom;
331 
332 #define RSC 1 //FIXME finetune
333  if(!s-> in.ch_count)
334  s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
335  if(!s->used_ch_count)
336  s->used_ch_count= s->in.ch_count;
337  if(!s->out.ch_count)
339 
340  if(!s-> in.ch_count){
342  av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
343  return -1;
344  }
345 
346  if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
347  char l1[1024], l2[1024];
348  av_get_channel_layout_string(l1, sizeof(l1), s-> in.ch_count, s-> in_ch_layout);
349  av_get_channel_layout_string(l2, sizeof(l2), s->out.ch_count, s->out_ch_layout);
350  av_log(s, AV_LOG_ERROR, "Rematrix is needed between %s and %s "
351  "but there is not enough information to do it\n", l1, l2);
352  return -1;
353  }
354 
357  s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
358 
359  s->in_buffer= s->in;
360 
361  if(!s->resample && !s->rematrix && !s->channel_map && !s->dither_method){
363  s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
364  return 0;
365  }
366 
368  s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
370  s->int_sample_fmt, s->out.ch_count, NULL, 0);
371 
372 
373  s->postin= s->in;
374  s->preout= s->out;
375  s->midbuf= s->in;
376 
377  if(s->channel_map){
378  s->postin.ch_count=
380  if(s->resample)
382  }
383  if(!s->resample_first){
384  s->midbuf.ch_count= s->out.ch_count;
385  if(s->resample)
386  s->in_buffer.ch_count = s->out.ch_count;
387  }
388 
392 
393  if(s->resample){
395  }
396 
397  s->dither = s->preout;
398 
399  if(s->rematrix || s->dither_method)
400  return swri_rematrix_init(s);
401 
402  return 0;
403 }
404 
405 int swri_realloc_audio(AudioData *a, int count){
406  int i, countb;
407  AudioData old;
408 
409  if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
410  return AVERROR(EINVAL);
411 
412  if(a->count >= count)
413  return 0;
414 
415  count*=2;
416 
417  countb= FFALIGN(count*a->bps, ALIGN);
418  old= *a;
419 
420  av_assert0(a->bps);
421  av_assert0(a->ch_count);
422 
423  a->data= av_mallocz(countb*a->ch_count);
424  if(!a->data)
425  return AVERROR(ENOMEM);
426  for(i=0; i<a->ch_count; i++){
427  a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
428  if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
429  }
430  if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
431  av_free(old.data);
432  a->count= count;
433 
434  return 1;
435 }
436 
437 static void copy(AudioData *out, AudioData *in,
438  int count){
439  av_assert0(out->planar == in->planar);
440  av_assert0(out->bps == in->bps);
441  av_assert0(out->ch_count == in->ch_count);
442  if(out->planar){
443  int ch;
444  for(ch=0; ch<out->ch_count; ch++)
445  memcpy(out->ch[ch], in->ch[ch], count*out->bps);
446  }else
447  memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
448 }
449 
450 static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
451  int i;
452  if(!in_arg){
453  memset(out->ch, 0, sizeof(out->ch));
454  }else if(out->planar){
455  for(i=0; i<out->ch_count; i++)
456  out->ch[i]= in_arg[i];
457  }else{
458  for(i=0; i<out->ch_count; i++)
459  out->ch[i]= in_arg[0] + i*out->bps;
460  }
461 }
462 
463 static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
464  int i;
465  if(out->planar){
466  for(i=0; i<out->ch_count; i++)
467  in_arg[i]= out->ch[i];
468  }else{
469  in_arg[0]= out->ch[0];
470  }
471 }
472 
473 /**
474  *
475  * out may be equal in.
476  */
477 static void buf_set(AudioData *out, AudioData *in, int count){
478  int ch;
479  if(in->planar){
480  for(ch=0; ch<out->ch_count; ch++)
481  out->ch[ch]= in->ch[ch] + count*out->bps;
482  }else{
483  for(ch=out->ch_count-1; ch>=0; ch--)
484  out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
485  }
486 }
487 
488 /**
489  *
490  * @return number of samples output per channel
491  */
492 static int resample(SwrContext *s, AudioData *out_param, int out_count,
493  const AudioData * in_param, int in_count){
494  AudioData in, out, tmp;
495  int ret_sum=0;
496  int border=0;
497 
498  av_assert1(s->in_buffer.ch_count == in_param->ch_count);
499  av_assert1(s->in_buffer.planar == in_param->planar);
500  av_assert1(s->in_buffer.fmt == in_param->fmt);
501 
502  tmp=out=*out_param;
503  in = *in_param;
504 
505  do{
506  int ret, size, consumed;
508  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
509  ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
510  out_count -= ret;
511  ret_sum += ret;
512  buf_set(&out, &out, ret);
513  s->in_buffer_count -= consumed;
514  s->in_buffer_index += consumed;
515 
516  if(!in_count)
517  break;
518  if(s->in_buffer_count <= border){
519  buf_set(&in, &in, -s->in_buffer_count);
520  in_count += s->in_buffer_count;
521  s->in_buffer_count=0;
522  s->in_buffer_index=0;
523  border = 0;
524  }
525  }
526 
527  if((s->flushed || in_count) && !s->in_buffer_count){
528  s->in_buffer_index=0;
529  ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, in_count, &consumed);
530  out_count -= ret;
531  ret_sum += ret;
532  buf_set(&out, &out, ret);
533  in_count -= consumed;
534  buf_set(&in, &in, consumed);
535  }
536 
537  //TODO is this check sane considering the advanced copy avoidance below
538  size= s->in_buffer_index + s->in_buffer_count + in_count;
539  if( size > s->in_buffer.count
540  && s->in_buffer_count + in_count <= s->in_buffer_index){
541  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
542  copy(&s->in_buffer, &tmp, s->in_buffer_count);
543  s->in_buffer_index=0;
544  }else
545  if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
546  return ret;
547 
548  if(in_count){
549  int count= in_count;
550  if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
551 
552  buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
553  copy(&tmp, &in, /*in_*/count);
554  s->in_buffer_count += count;
555  in_count -= count;
556  border += count;
557  buf_set(&in, &in, count);
559  if(s->in_buffer_count != count || in_count)
560  continue;
561  }
562  break;
563  }while(1);
564 
565  s->resample_in_constraint= !!out_count;
566 
567  return ret_sum;
568 }
569 
570 static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
571  AudioData *in , int in_count){
572  AudioData *postin, *midbuf, *preout;
573  int ret/*, in_max*/;
574  AudioData preout_tmp, midbuf_tmp;
575 
576  if(s->full_convert){
577  av_assert0(!s->resample);
578  swri_audio_convert(s->full_convert, out, in, in_count);
579  return out_count;
580  }
581 
582 // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
583 // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
584 
585  if((ret=swri_realloc_audio(&s->postin, in_count))<0)
586  return ret;
587  if(s->resample_first){
589  if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
590  return ret;
591  }else{
593  if((ret=swri_realloc_audio(&s->midbuf, in_count))<0)
594  return ret;
595  }
596  if((ret=swri_realloc_audio(&s->preout, out_count))<0)
597  return ret;
598 
599  postin= &s->postin;
600 
601  midbuf_tmp= s->midbuf;
602  midbuf= &midbuf_tmp;
603  preout_tmp= s->preout;
604  preout= &preout_tmp;
605 
606  if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
607  postin= in;
608 
609  if(s->resample_first ? !s->resample : !s->rematrix)
610  midbuf= postin;
611 
612  if(s->resample_first ? !s->rematrix : !s->resample)
613  preout= midbuf;
614 
615  if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar){
616  if(preout==in){
617  out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
618  av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
619  copy(out, in, out_count);
620  return out_count;
621  }
622  else if(preout==postin) preout= midbuf= postin= out;
623  else if(preout==midbuf) preout= midbuf= out;
624  else preout= out;
625  }
626 
627  if(in != postin){
628  swri_audio_convert(s->in_convert, postin, in, in_count);
629  }
630 
631  if(s->resample_first){
632  if(postin != midbuf)
633  out_count= resample(s, midbuf, out_count, postin, in_count);
634  if(midbuf != preout)
635  swri_rematrix(s, preout, midbuf, out_count, preout==out);
636  }else{
637  if(postin != midbuf)
638  swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
639  if(midbuf != preout)
640  out_count= resample(s, preout, out_count, midbuf, in_count);
641  }
642 
643  if(preout != out && out_count){
644  if(s->dither_method){
645  int ch;
646  int dither_count= FFMAX(out_count, 1<<16);
647  av_assert0(preout != in);
648 
649  if((ret=swri_realloc_audio(&s->dither, dither_count))<0)
650  return ret;
651  if(ret)
652  for(ch=0; ch<s->dither.ch_count; ch++)
653  swri_get_dither(s, s->dither.ch[ch], s->dither.count, 12345678913579<<ch, s->out_sample_fmt, s->int_sample_fmt);
654  av_assert0(s->dither.ch_count == preout->ch_count);
655 
656  if(s->dither_pos + out_count > s->dither.count)
657  s->dither_pos = 0;
658 
659  for(ch=0; ch<preout->ch_count; ch++)
660  s->mix_2_1_f(preout->ch[ch], preout->ch[ch], s->dither.ch[ch] + s->dither.bps * s->dither_pos, s->native_one, 0, 0, out_count);
661 
662  s->dither_pos += out_count;
663  }
664 //FIXME packed doesnt need more than 1 chan here!
665  swri_audio_convert(s->out_convert, out, preout, out_count);
666  }
667  return out_count;
668 }
669 
670 int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
671  const uint8_t *in_arg [SWR_CH_MAX], int in_count){
672  AudioData * in= &s->in;
673  AudioData *out= &s->out;
674 
675  if(s->drop_output > 0){
676  int ret;
677  AudioData tmp = s->out;
678  uint8_t *tmp_arg[SWR_CH_MAX];
679  tmp.count = 0;
680  tmp.data = NULL;
681  if((ret=swri_realloc_audio(&tmp, s->drop_output))<0)
682  return ret;
683 
684  reversefill_audiodata(&tmp, tmp_arg);
685  s->drop_output *= -1; //FIXME find a less hackish solution
686  ret = swr_convert(s, tmp_arg, -s->drop_output, in_arg, in_count); //FIXME optimize but this is as good as never called so maybe it doesnt matter
687  s->drop_output *= -1;
688  if(ret>0)
689  s->drop_output -= ret;
690 
691  av_freep(&tmp.data);
692  if(s->drop_output || !out_arg)
693  return 0;
694  in_count = 0;
695  }
696 
697  if(!in_arg){
698  if(s->resample){
699  if (!s->flushed)
700  s->resampler->flush(s);
701  s->resample_in_constraint = 0;
702  s->flushed = 1;
703  }else if(!s->in_buffer_count){
704  return 0;
705  }
706  }else
707  fill_audiodata(in , (void*)in_arg);
708 
709  fill_audiodata(out, out_arg);
710 
711  if(s->resample){
712  int ret = swr_convert_internal(s, out, out_count, in, in_count);
713  if(ret>0 && !s->drop_output)
714  s->outpts += ret * (int64_t)s->in_sample_rate;
715  return ret;
716  }else{
717  AudioData tmp= *in;
718  int ret2=0;
719  int ret, size;
720  size = FFMIN(out_count, s->in_buffer_count);
721  if(size){
722  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
723  ret= swr_convert_internal(s, out, size, &tmp, size);
724  if(ret<0)
725  return ret;
726  ret2= ret;
727  s->in_buffer_count -= ret;
728  s->in_buffer_index += ret;
729  buf_set(out, out, ret);
730  out_count -= ret;
731  if(!s->in_buffer_count)
732  s->in_buffer_index = 0;
733  }
734 
735  if(in_count){
736  size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
737 
738  if(in_count > out_count) { //FIXME move after swr_convert_internal
739  if( size > s->in_buffer.count
740  && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
741  buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
742  copy(&s->in_buffer, &tmp, s->in_buffer_count);
743  s->in_buffer_index=0;
744  }else
745  if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
746  return ret;
747  }
748 
749  if(out_count){
750  size = FFMIN(in_count, out_count);
751  ret= swr_convert_internal(s, out, size, in, size);
752  if(ret<0)
753  return ret;
754  buf_set(in, in, ret);
755  in_count -= ret;
756  ret2 += ret;
757  }
758  if(in_count){
759  buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
760  copy(&tmp, in, in_count);
761  s->in_buffer_count += in_count;
762  }
763  }
764  if(ret2>0 && !s->drop_output)
765  s->outpts += ret2 * (int64_t)s->in_sample_rate;
766  return ret2;
767  }
768 }
769 
770 int swr_drop_output(struct SwrContext *s, int count){
771  s->drop_output += count;
772 
773  if(s->drop_output <= 0)
774  return 0;
775 
776  av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
777  return swr_convert(s, NULL, s->drop_output, NULL, 0);
778 }
779 
780 int swr_inject_silence(struct SwrContext *s, int count){
781  int ret, i;
782  AudioData silence = s->in;
783  uint8_t *tmp_arg[SWR_CH_MAX];
784 
785  if(count <= 0)
786  return 0;
787 
788  silence.count = 0;
789  silence.data = NULL;
790  if((ret=swri_realloc_audio(&silence, count))<0)
791  return ret;
792 
793  if(silence.planar) for(i=0; i<silence.ch_count; i++) {
794  memset(silence.ch[i], silence.bps==1 ? 0x80 : 0, count*silence.bps);
795  } else
796  memset(silence.ch[0], silence.bps==1 ? 0x80 : 0, count*silence.bps*silence.ch_count);
797 
798  reversefill_audiodata(&silence, tmp_arg);
799  av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
800  ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
801  av_freep(&silence.data);
802  return ret;
803 }
804 
805 int64_t swr_get_delay(struct SwrContext *s, int64_t base){
806  if (s->resampler && s->resample){
807  return s->resampler->get_delay(s, base);
808  }else{
809  return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
810  }
811 }
812 
813 int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
814  int ret;
815 
816  if (!s || compensation_distance < 0)
817  return AVERROR(EINVAL);
818  if (!compensation_distance && sample_delta)
819  return AVERROR(EINVAL);
820  if (!s->resample) {
821  s->flags |= SWR_FLAG_RESAMPLE;
822  ret = swr_init(s);
823  if (ret < 0)
824  return ret;
825  }
826  if (!s->resampler->set_compensation){
827  return AVERROR(EINVAL);
828  }else{
829  return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
830  }
831 }
832 
833 int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
834  if(pts == INT64_MIN)
835  return s->outpts;
836  if(s->min_compensation >= FLT_MAX) {
837  return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
838  } else {
839  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;
840  double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
841 
842  if(fabs(fdelta) > s->min_compensation) {
843  if(!s->outpts || fabs(fdelta) > s->min_hard_compensation){
844  int ret;
845  if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
846  else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
847  if(ret<0){
848  av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
849  }
852  double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
853  int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
854  av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
855  swr_set_compensation(s, comp, duration);
856  }
857  }
858 
859  return s->outpts;
860  }
861 }