You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

947 lines
42KB

  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. #include "libavutil/opt.h"
  21. #include "swresample_internal.h"
  22. #include "audioconvert.h"
  23. #include "libavutil/avassert.h"
  24. #include "libavutil/channel_layout.h"
  25. #include <float.h>
  26. #define C30DB M_SQRT2
  27. #define C15DB 1.189207115
  28. #define C__0DB 1.0
  29. #define C_15DB 0.840896415
  30. #define C_30DB M_SQRT1_2
  31. #define C_45DB 0.594603558
  32. #define C_60DB 0.5
  33. #define ALIGN 32
  34. //TODO split options array out?
  35. #define OFFSET(x) offsetof(SwrContext,x)
  36. #define PARAM AV_OPT_FLAG_AUDIO_PARAM
  37. static const AVOption options[]={
  38. {"ich" , "set input channel count" , OFFSET( in.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  39. {"in_channel_count" , "set input channel count" , OFFSET( in.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  40. {"och" , "set output channel count" , OFFSET(out.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  41. {"out_channel_count" , "set output channel count" , OFFSET(out.ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  42. {"uch" , "set used channel count" , OFFSET(used_ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  43. {"used_channel_count" , "set used channel count" , OFFSET(used_ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  44. {"isr" , "set input sample rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  45. {"in_sample_rate" , "set input sample rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  46. {"osr" , "set output sample rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  47. {"out_sample_rate" , "set output sample rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  48. {"isf" , "set input sample format" , OFFSET( in_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
  49. {"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},
  50. {"osf" , "set output sample format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
  51. {"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},
  52. {"tsf" , "set internal sample format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1 , INT_MAX, PARAM},
  53. {"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},
  54. {"icl" , "set input channel layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  55. {"in_channel_layout" , "set input channel layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  56. {"ocl" , "set output channel layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  57. {"out_channel_layout" , "set output channel layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  58. {"clev" , "set center mix level" , OFFSET(clev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  59. {"center_mix_level" , "set center mix level" , OFFSET(clev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  60. {"slev" , "set surround mix level" , OFFSET(slev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  61. {"surround_mix_level" , "set surround mix Level" , OFFSET(slev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  62. {"lfe_mix_level" , "set LFE mix level" , OFFSET(lfe_mix_level ), AV_OPT_TYPE_FLOAT, {.dbl=0 }, -32 , 32 , PARAM},
  63. {"rmvol" , "set rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
  64. {"rematrix_volume" , "set rematrix volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
  65. {"rematrix_maxval" , "set rematrix maxval" , OFFSET(rematrix_maxval), AV_OPT_TYPE_FLOAT, {.dbl=0.0 }, 0 , 1000 , PARAM},
  66. {"flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
  67. {"swr_flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
  68. {"res" , "force resampling" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_FLAG_RESAMPLE }, INT_MIN, INT_MAX , PARAM, "flags"},
  69. {"dither_scale" , "set dither scale" , OFFSET(dither.scale ), AV_OPT_TYPE_FLOAT, {.dbl=1 }, 0 , INT_MAX , PARAM},
  70. {"dither_method" , "set dither method" , OFFSET(dither.method ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_DITHER_NB-1, PARAM, "dither_method"},
  71. {"rectangular" , "select rectangular dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX , PARAM, "dither_method"},
  72. {"triangular" , "select triangular dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX , PARAM, "dither_method"},
  73. {"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"},
  74. {"lipshitz" , "select lipshitz noise shaping dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_LIPSHITZ}, INT_MIN, INT_MAX, PARAM, "dither_method"},
  75. {"shibata" , "select shibata noise shaping dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_SHIBATA }, INT_MIN, INT_MAX, PARAM, "dither_method"},
  76. {"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"},
  77. {"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"},
  78. {"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"},
  79. {"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"},
  80. {"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"},
  81. {"filter_size" , "set swr resampling filter size", OFFSET(filter_size) , AV_OPT_TYPE_INT , {.i64=32 }, 0 , INT_MAX , PARAM },
  82. {"phase_shift" , "set swr resampling phase shift", OFFSET(phase_shift) , AV_OPT_TYPE_INT , {.i64=10 }, 0 , 24 , PARAM },
  83. {"linear_interp" , "enable linear interpolation" , OFFSET(linear_interp) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
  84. {"cutoff" , "set cutoff frequency ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0. }, 0 , 1 , PARAM },
  85. /* duplicate option in order to work with avconv */
  86. {"resample_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. {"first_pts" , "Assume the first pts should be this value (in samples)."
  105. , OFFSET(firstpts_in_samples), AV_OPT_TYPE_INT64 ,{.i64=AV_NOPTS_VALUE }, INT64_MIN,INT64_MAX, PARAM },
  106. { "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" },
  107. { "none", "select none", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  108. { "dolby", "select Dolby", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  109. { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  110. { "filter_type" , "select swr filter type" , OFFSET(filter_type) , AV_OPT_TYPE_INT , { .i64 = SWR_FILTER_TYPE_KAISER }, SWR_FILTER_TYPE_CUBIC, SWR_FILTER_TYPE_KAISER, PARAM, "filter_type" },
  111. { "cubic" , "select cubic" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  112. { "blackman_nuttall", "select Blackman Nuttall Windowed Sinc", 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_BLACKMAN_NUTTALL }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  113. { "kaiser" , "select Kaiser Windowed Sinc" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  114. { "kaiser_beta" , "set swr Kaiser Window Beta" , OFFSET(kaiser_beta) , AV_OPT_TYPE_INT , {.i64=9 }, 2 , 16 , PARAM },
  115. { "output_sample_bits" , "set swr number of output sample bits", OFFSET(dither.output_sample_bits), AV_OPT_TYPE_INT , {.i64=0 }, 0 , 64 , PARAM },
  116. {0}
  117. };
  118. static const char* context_to_name(void* ptr) {
  119. return "SWR";
  120. }
  121. static const AVClass av_class = {
  122. .class_name = "SWResampler",
  123. .item_name = context_to_name,
  124. .option = options,
  125. .version = LIBAVUTIL_VERSION_INT,
  126. .log_level_offset_offset = OFFSET(log_level_offset),
  127. .parent_log_context_offset = OFFSET(log_ctx),
  128. .category = AV_CLASS_CATEGORY_SWRESAMPLER,
  129. };
  130. unsigned swresample_version(void)
  131. {
  132. av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
  133. return LIBSWRESAMPLE_VERSION_INT;
  134. }
  135. const char *swresample_configuration(void)
  136. {
  137. return FFMPEG_CONFIGURATION;
  138. }
  139. const char *swresample_license(void)
  140. {
  141. #define LICENSE_PREFIX "libswresample license: "
  142. return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
  143. }
  144. int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
  145. if(!s || s->in_convert) // s needs to be allocated but not initialized
  146. return AVERROR(EINVAL);
  147. s->channel_map = channel_map;
  148. return 0;
  149. }
  150. const AVClass *swr_get_class(void)
  151. {
  152. return &av_class;
  153. }
  154. av_cold struct SwrContext *swr_alloc(void){
  155. SwrContext *s= av_mallocz(sizeof(SwrContext));
  156. if(s){
  157. s->av_class= &av_class;
  158. av_opt_set_defaults(s);
  159. }
  160. return s;
  161. }
  162. struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
  163. int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
  164. int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate,
  165. int log_offset, void *log_ctx){
  166. if(!s) s= swr_alloc();
  167. if(!s) return NULL;
  168. s->log_level_offset= log_offset;
  169. s->log_ctx= log_ctx;
  170. av_opt_set_int(s, "ocl", out_ch_layout, 0);
  171. av_opt_set_int(s, "osf", out_sample_fmt, 0);
  172. av_opt_set_int(s, "osr", out_sample_rate, 0);
  173. av_opt_set_int(s, "icl", in_ch_layout, 0);
  174. av_opt_set_int(s, "isf", in_sample_fmt, 0);
  175. av_opt_set_int(s, "isr", in_sample_rate, 0);
  176. av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE, 0);
  177. av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
  178. av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
  179. av_opt_set_int(s, "uch", 0, 0);
  180. return s;
  181. }
  182. static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
  183. a->fmt = fmt;
  184. a->bps = av_get_bytes_per_sample(fmt);
  185. a->planar= av_sample_fmt_is_planar(fmt);
  186. }
  187. static void free_temp(AudioData *a){
  188. av_free(a->data);
  189. memset(a, 0, sizeof(*a));
  190. }
  191. static void clear_context(SwrContext *s){
  192. s->in_buffer_index= 0;
  193. s->in_buffer_count= 0;
  194. s->resample_in_constraint= 0;
  195. memset(s->in.ch, 0, sizeof(s->in.ch));
  196. memset(s->out.ch, 0, sizeof(s->out.ch));
  197. free_temp(&s->postin);
  198. free_temp(&s->midbuf);
  199. free_temp(&s->preout);
  200. free_temp(&s->in_buffer);
  201. free_temp(&s->silence);
  202. free_temp(&s->drop_temp);
  203. free_temp(&s->dither.noise);
  204. free_temp(&s->dither.temp);
  205. swri_audio_convert_free(&s-> in_convert);
  206. swri_audio_convert_free(&s->out_convert);
  207. swri_audio_convert_free(&s->full_convert);
  208. swri_rematrix_free(s);
  209. s->flushed = 0;
  210. }
  211. av_cold void swr_free(SwrContext **ss){
  212. SwrContext *s= *ss;
  213. if(s){
  214. clear_context(s);
  215. if (s->resampler)
  216. s->resampler->free(&s->resample);
  217. }
  218. av_freep(ss);
  219. }
  220. av_cold void swr_close(SwrContext *s){
  221. clear_context(s);
  222. }
  223. av_cold int swr_init(struct SwrContext *s){
  224. int ret;
  225. clear_context(s);
  226. if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
  227. av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
  228. return AVERROR(EINVAL);
  229. }
  230. if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
  231. av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
  232. return AVERROR(EINVAL);
  233. }
  234. if(av_get_channel_layout_nb_channels(s-> in_ch_layout) > SWR_CH_MAX) {
  235. av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
  236. s->in_ch_layout = 0;
  237. }
  238. if(av_get_channel_layout_nb_channels(s->out_ch_layout) > SWR_CH_MAX) {
  239. av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
  240. s->out_ch_layout = 0;
  241. }
  242. switch(s->engine){
  243. #if CONFIG_LIBSOXR
  244. extern struct Resampler const soxr_resampler;
  245. case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
  246. #endif
  247. case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
  248. default:
  249. av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
  250. return AVERROR(EINVAL);
  251. }
  252. if(!s->used_ch_count)
  253. s->used_ch_count= s->in.ch_count;
  254. if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
  255. av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
  256. s-> in_ch_layout= 0;
  257. }
  258. if(!s-> in_ch_layout)
  259. s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
  260. if(!s->out_ch_layout)
  261. s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);
  262. s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
  263. s->rematrix_custom;
  264. if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
  265. if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
  266. s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
  267. }else if( av_get_planar_sample_fmt(s-> in_sample_fmt) == AV_SAMPLE_FMT_S32P
  268. && av_get_planar_sample_fmt(s->out_sample_fmt) == AV_SAMPLE_FMT_S32P
  269. && !s->rematrix
  270. && s->engine != SWR_ENGINE_SOXR){
  271. s->int_sample_fmt= AV_SAMPLE_FMT_S32P;
  272. }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
  273. s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
  274. }else{
  275. av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
  276. s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
  277. }
  278. }
  279. if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
  280. &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
  281. &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
  282. &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
  283. 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));
  284. return AVERROR(EINVAL);
  285. }
  286. set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
  287. set_audiodata_fmt(&s->out, s->out_sample_fmt);
  288. if (s->firstpts_in_samples != AV_NOPTS_VALUE) {
  289. if (!s->async && s->min_compensation >= FLT_MAX/2)
  290. s->async = 1;
  291. s->firstpts =
  292. s->outpts = s->firstpts_in_samples * s->out_sample_rate;
  293. } else
  294. s->firstpts = AV_NOPTS_VALUE;
  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. if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
  303. 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);
  304. }else
  305. s->resampler->free(&s->resample);
  306. if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
  307. && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
  308. && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
  309. && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
  310. && s->resample){
  311. av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
  312. return -1;
  313. }
  314. #define RSC 1 //FIXME finetune
  315. if(!s-> in.ch_count)
  316. s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
  317. if(!s->used_ch_count)
  318. s->used_ch_count= s->in.ch_count;
  319. if(!s->out.ch_count)
  320. s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);
  321. if(!s-> in.ch_count){
  322. av_assert0(!s->in_ch_layout);
  323. av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
  324. return -1;
  325. }
  326. if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
  327. char l1[1024], l2[1024];
  328. av_get_channel_layout_string(l1, sizeof(l1), s-> in.ch_count, s-> in_ch_layout);
  329. av_get_channel_layout_string(l2, sizeof(l2), s->out.ch_count, s->out_ch_layout);
  330. av_log(s, AV_LOG_ERROR, "Rematrix is needed between %s and %s "
  331. "but there is not enough information to do it\n", l1, l2);
  332. return -1;
  333. }
  334. av_assert0(s->used_ch_count);
  335. av_assert0(s->out.ch_count);
  336. s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
  337. s->in_buffer= s->in;
  338. s->silence = s->in;
  339. s->drop_temp= s->out;
  340. if(!s->resample && !s->rematrix && !s->channel_map && !s->dither.method){
  341. s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
  342. s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
  343. return 0;
  344. }
  345. s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
  346. s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
  347. s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
  348. s->int_sample_fmt, s->out.ch_count, NULL, 0);
  349. if (!s->in_convert || !s->out_convert)
  350. return AVERROR(ENOMEM);
  351. s->postin= s->in;
  352. s->preout= s->out;
  353. s->midbuf= s->in;
  354. if(s->channel_map){
  355. s->postin.ch_count=
  356. s->midbuf.ch_count= s->used_ch_count;
  357. if(s->resample)
  358. s->in_buffer.ch_count= s->used_ch_count;
  359. }
  360. if(!s->resample_first){
  361. s->midbuf.ch_count= s->out.ch_count;
  362. if(s->resample)
  363. s->in_buffer.ch_count = s->out.ch_count;
  364. }
  365. set_audiodata_fmt(&s->postin, s->int_sample_fmt);
  366. set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
  367. set_audiodata_fmt(&s->preout, s->int_sample_fmt);
  368. if(s->resample){
  369. set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
  370. }
  371. if ((ret = swri_dither_init(s, s->out_sample_fmt, s->int_sample_fmt)) < 0)
  372. return ret;
  373. if(s->rematrix || s->dither.method)
  374. return swri_rematrix_init(s);
  375. return 0;
  376. }
  377. int swri_realloc_audio(AudioData *a, int count){
  378. int i, countb;
  379. AudioData old;
  380. if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
  381. return AVERROR(EINVAL);
  382. if(a->count >= count)
  383. return 0;
  384. count*=2;
  385. countb= FFALIGN(count*a->bps, ALIGN);
  386. old= *a;
  387. av_assert0(a->bps);
  388. av_assert0(a->ch_count);
  389. a->data= av_mallocz(countb*a->ch_count);
  390. if(!a->data)
  391. return AVERROR(ENOMEM);
  392. for(i=0; i<a->ch_count; i++){
  393. a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
  394. if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
  395. }
  396. if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
  397. av_freep(&old.data);
  398. a->count= count;
  399. return 1;
  400. }
  401. static void copy(AudioData *out, AudioData *in,
  402. int count){
  403. av_assert0(out->planar == in->planar);
  404. av_assert0(out->bps == in->bps);
  405. av_assert0(out->ch_count == in->ch_count);
  406. if(out->planar){
  407. int ch;
  408. for(ch=0; ch<out->ch_count; ch++)
  409. memcpy(out->ch[ch], in->ch[ch], count*out->bps);
  410. }else
  411. memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
  412. }
  413. static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  414. int i;
  415. if(!in_arg){
  416. memset(out->ch, 0, sizeof(out->ch));
  417. }else if(out->planar){
  418. for(i=0; i<out->ch_count; i++)
  419. out->ch[i]= in_arg[i];
  420. }else{
  421. for(i=0; i<out->ch_count; i++)
  422. out->ch[i]= in_arg[0] + i*out->bps;
  423. }
  424. }
  425. static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  426. int i;
  427. if(out->planar){
  428. for(i=0; i<out->ch_count; i++)
  429. in_arg[i]= out->ch[i];
  430. }else{
  431. in_arg[0]= out->ch[0];
  432. }
  433. }
  434. /**
  435. *
  436. * out may be equal in.
  437. */
  438. static void buf_set(AudioData *out, AudioData *in, int count){
  439. int ch;
  440. if(in->planar){
  441. for(ch=0; ch<out->ch_count; ch++)
  442. out->ch[ch]= in->ch[ch] + count*out->bps;
  443. }else{
  444. for(ch=out->ch_count-1; ch>=0; ch--)
  445. out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
  446. }
  447. }
  448. /**
  449. *
  450. * @return number of samples output per channel
  451. */
  452. static int resample(SwrContext *s, AudioData *out_param, int out_count,
  453. const AudioData * in_param, int in_count){
  454. AudioData in, out, tmp;
  455. int ret_sum=0;
  456. int border=0;
  457. int padless = ARCH_X86 && s->engine == SWR_ENGINE_SWR ? 7 : 0;
  458. av_assert1(s->in_buffer.ch_count == in_param->ch_count);
  459. av_assert1(s->in_buffer.planar == in_param->planar);
  460. av_assert1(s->in_buffer.fmt == in_param->fmt);
  461. tmp=out=*out_param;
  462. in = *in_param;
  463. do{
  464. int ret, size, consumed;
  465. if(!s->resample_in_constraint && s->in_buffer_count){
  466. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  467. ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
  468. out_count -= ret;
  469. ret_sum += ret;
  470. buf_set(&out, &out, ret);
  471. s->in_buffer_count -= consumed;
  472. s->in_buffer_index += consumed;
  473. if(!in_count)
  474. break;
  475. if(s->in_buffer_count <= border){
  476. buf_set(&in, &in, -s->in_buffer_count);
  477. in_count += s->in_buffer_count;
  478. s->in_buffer_count=0;
  479. s->in_buffer_index=0;
  480. border = 0;
  481. }
  482. }
  483. if((s->flushed || in_count > padless) && !s->in_buffer_count){
  484. s->in_buffer_index=0;
  485. ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, FFMAX(in_count-padless, 0), &consumed);
  486. out_count -= ret;
  487. ret_sum += ret;
  488. buf_set(&out, &out, ret);
  489. in_count -= consumed;
  490. buf_set(&in, &in, consumed);
  491. }
  492. //TODO is this check sane considering the advanced copy avoidance below
  493. size= s->in_buffer_index + s->in_buffer_count + in_count;
  494. if( size > s->in_buffer.count
  495. && s->in_buffer_count + in_count <= s->in_buffer_index){
  496. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  497. copy(&s->in_buffer, &tmp, s->in_buffer_count);
  498. s->in_buffer_index=0;
  499. }else
  500. if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
  501. return ret;
  502. if(in_count){
  503. int count= in_count;
  504. if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
  505. buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
  506. copy(&tmp, &in, /*in_*/count);
  507. s->in_buffer_count += count;
  508. in_count -= count;
  509. border += count;
  510. buf_set(&in, &in, count);
  511. s->resample_in_constraint= 0;
  512. if(s->in_buffer_count != count || in_count)
  513. continue;
  514. if (padless) {
  515. padless = 0;
  516. continue;
  517. }
  518. }
  519. break;
  520. }while(1);
  521. s->resample_in_constraint= !!out_count;
  522. return ret_sum;
  523. }
  524. static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
  525. AudioData *in , int in_count){
  526. AudioData *postin, *midbuf, *preout;
  527. int ret/*, in_max*/;
  528. AudioData preout_tmp, midbuf_tmp;
  529. if(s->full_convert){
  530. av_assert0(!s->resample);
  531. swri_audio_convert(s->full_convert, out, in, in_count);
  532. return out_count;
  533. }
  534. // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
  535. // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
  536. if((ret=swri_realloc_audio(&s->postin, in_count))<0)
  537. return ret;
  538. if(s->resample_first){
  539. av_assert0(s->midbuf.ch_count == s->used_ch_count);
  540. if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
  541. return ret;
  542. }else{
  543. av_assert0(s->midbuf.ch_count == s->out.ch_count);
  544. if((ret=swri_realloc_audio(&s->midbuf, in_count))<0)
  545. return ret;
  546. }
  547. if((ret=swri_realloc_audio(&s->preout, out_count))<0)
  548. return ret;
  549. postin= &s->postin;
  550. midbuf_tmp= s->midbuf;
  551. midbuf= &midbuf_tmp;
  552. preout_tmp= s->preout;
  553. preout= &preout_tmp;
  554. if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
  555. postin= in;
  556. if(s->resample_first ? !s->resample : !s->rematrix)
  557. midbuf= postin;
  558. if(s->resample_first ? !s->rematrix : !s->resample)
  559. preout= midbuf;
  560. if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar
  561. && !(s->out_sample_fmt==AV_SAMPLE_FMT_S32P && (s->dither.output_sample_bits&31))){
  562. if(preout==in){
  563. out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
  564. av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
  565. copy(out, in, out_count);
  566. return out_count;
  567. }
  568. else if(preout==postin) preout= midbuf= postin= out;
  569. else if(preout==midbuf) preout= midbuf= out;
  570. else preout= out;
  571. }
  572. if(in != postin){
  573. swri_audio_convert(s->in_convert, postin, in, in_count);
  574. }
  575. if(s->resample_first){
  576. if(postin != midbuf)
  577. out_count= resample(s, midbuf, out_count, postin, in_count);
  578. if(midbuf != preout)
  579. swri_rematrix(s, preout, midbuf, out_count, preout==out);
  580. }else{
  581. if(postin != midbuf)
  582. swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
  583. if(midbuf != preout)
  584. out_count= resample(s, preout, out_count, midbuf, in_count);
  585. }
  586. if(preout != out && out_count){
  587. AudioData *conv_src = preout;
  588. if(s->dither.method){
  589. int ch;
  590. int dither_count= FFMAX(out_count, 1<<16);
  591. if (preout == in) {
  592. conv_src = &s->dither.temp;
  593. if((ret=swri_realloc_audio(&s->dither.temp, dither_count))<0)
  594. return ret;
  595. }
  596. if((ret=swri_realloc_audio(&s->dither.noise, dither_count))<0)
  597. return ret;
  598. if(ret)
  599. for(ch=0; ch<s->dither.noise.ch_count; ch++)
  600. swri_get_dither(s, s->dither.noise.ch[ch], s->dither.noise.count, 12345678913579<<ch, s->dither.noise.fmt);
  601. av_assert0(s->dither.noise.ch_count == preout->ch_count);
  602. if(s->dither.noise_pos + out_count > s->dither.noise.count)
  603. s->dither.noise_pos = 0;
  604. if (s->dither.method < SWR_DITHER_NS){
  605. if (s->mix_2_1_simd) {
  606. int len1= out_count&~15;
  607. int off = len1 * preout->bps;
  608. if(len1)
  609. for(ch=0; ch<preout->ch_count; ch++)
  610. 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);
  611. if(out_count != len1)
  612. for(ch=0; ch<preout->ch_count; ch++)
  613. 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);
  614. } else {
  615. for(ch=0; ch<preout->ch_count; ch++)
  616. 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);
  617. }
  618. } else {
  619. switch(s->int_sample_fmt) {
  620. case AV_SAMPLE_FMT_S16P :swri_noise_shaping_int16(s, conv_src, preout, &s->dither.noise, out_count); break;
  621. case AV_SAMPLE_FMT_S32P :swri_noise_shaping_int32(s, conv_src, preout, &s->dither.noise, out_count); break;
  622. case AV_SAMPLE_FMT_FLTP :swri_noise_shaping_float(s, conv_src, preout, &s->dither.noise, out_count); break;
  623. case AV_SAMPLE_FMT_DBLP :swri_noise_shaping_double(s,conv_src, preout, &s->dither.noise, out_count); break;
  624. }
  625. }
  626. s->dither.noise_pos += out_count;
  627. }
  628. //FIXME packed doesn't need more than 1 chan here!
  629. swri_audio_convert(s->out_convert, out, conv_src, out_count);
  630. }
  631. return out_count;
  632. }
  633. int swr_is_initialized(struct SwrContext *s) {
  634. return !!s->in_buffer.ch_count;
  635. }
  636. int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
  637. const uint8_t *in_arg [SWR_CH_MAX], int in_count){
  638. AudioData * in= &s->in;
  639. AudioData *out= &s->out;
  640. if (!swr_is_initialized(s)) {
  641. av_log(s, AV_LOG_ERROR, "Context has not been initialized\n");
  642. return AVERROR(EINVAL);
  643. }
  644. while(s->drop_output > 0){
  645. int ret;
  646. uint8_t *tmp_arg[SWR_CH_MAX];
  647. #define MAX_DROP_STEP 16384
  648. if((ret=swri_realloc_audio(&s->drop_temp, FFMIN(s->drop_output, MAX_DROP_STEP)))<0)
  649. return ret;
  650. reversefill_audiodata(&s->drop_temp, tmp_arg);
  651. s->drop_output *= -1; //FIXME find a less hackish solution
  652. 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
  653. s->drop_output *= -1;
  654. in_count = 0;
  655. if(ret>0) {
  656. s->drop_output -= ret;
  657. continue;
  658. }
  659. if(s->drop_output || !out_arg)
  660. return 0;
  661. }
  662. if(!in_arg){
  663. if(s->resample){
  664. if (!s->flushed)
  665. s->resampler->flush(s);
  666. s->resample_in_constraint = 0;
  667. s->flushed = 1;
  668. }else if(!s->in_buffer_count){
  669. return 0;
  670. }
  671. }else
  672. fill_audiodata(in , (void*)in_arg);
  673. fill_audiodata(out, out_arg);
  674. if(s->resample){
  675. int ret = swr_convert_internal(s, out, out_count, in, in_count);
  676. if(ret>0 && !s->drop_output)
  677. s->outpts += ret * (int64_t)s->in_sample_rate;
  678. return ret;
  679. }else{
  680. AudioData tmp= *in;
  681. int ret2=0;
  682. int ret, size;
  683. size = FFMIN(out_count, s->in_buffer_count);
  684. if(size){
  685. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  686. ret= swr_convert_internal(s, out, size, &tmp, size);
  687. if(ret<0)
  688. return ret;
  689. ret2= ret;
  690. s->in_buffer_count -= ret;
  691. s->in_buffer_index += ret;
  692. buf_set(out, out, ret);
  693. out_count -= ret;
  694. if(!s->in_buffer_count)
  695. s->in_buffer_index = 0;
  696. }
  697. if(in_count){
  698. size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
  699. if(in_count > out_count) { //FIXME move after swr_convert_internal
  700. if( size > s->in_buffer.count
  701. && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
  702. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  703. copy(&s->in_buffer, &tmp, s->in_buffer_count);
  704. s->in_buffer_index=0;
  705. }else
  706. if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
  707. return ret;
  708. }
  709. if(out_count){
  710. size = FFMIN(in_count, out_count);
  711. ret= swr_convert_internal(s, out, size, in, size);
  712. if(ret<0)
  713. return ret;
  714. buf_set(in, in, ret);
  715. in_count -= ret;
  716. ret2 += ret;
  717. }
  718. if(in_count){
  719. buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
  720. copy(&tmp, in, in_count);
  721. s->in_buffer_count += in_count;
  722. }
  723. }
  724. if(ret2>0 && !s->drop_output)
  725. s->outpts += ret2 * (int64_t)s->in_sample_rate;
  726. return ret2;
  727. }
  728. }
  729. int swr_drop_output(struct SwrContext *s, int count){
  730. s->drop_output += count;
  731. if(s->drop_output <= 0)
  732. return 0;
  733. av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
  734. return swr_convert(s, NULL, s->drop_output, NULL, 0);
  735. }
  736. int swr_inject_silence(struct SwrContext *s, int count){
  737. int ret, i;
  738. uint8_t *tmp_arg[SWR_CH_MAX];
  739. if(count <= 0)
  740. return 0;
  741. #define MAX_SILENCE_STEP 16384
  742. while (count > MAX_SILENCE_STEP) {
  743. if ((ret = swr_inject_silence(s, MAX_SILENCE_STEP)) < 0)
  744. return ret;
  745. count -= MAX_SILENCE_STEP;
  746. }
  747. if((ret=swri_realloc_audio(&s->silence, count))<0)
  748. return ret;
  749. if(s->silence.planar) for(i=0; i<s->silence.ch_count; i++) {
  750. memset(s->silence.ch[i], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps);
  751. } else
  752. memset(s->silence.ch[0], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps*s->silence.ch_count);
  753. reversefill_audiodata(&s->silence, tmp_arg);
  754. av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
  755. ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
  756. return ret;
  757. }
  758. int64_t swr_get_delay(struct SwrContext *s, int64_t base){
  759. if (s->resampler && s->resample){
  760. return s->resampler->get_delay(s, base);
  761. }else{
  762. return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
  763. }
  764. }
  765. int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
  766. int ret;
  767. if (!s || compensation_distance < 0)
  768. return AVERROR(EINVAL);
  769. if (!compensation_distance && sample_delta)
  770. return AVERROR(EINVAL);
  771. if (!s->resample) {
  772. s->flags |= SWR_FLAG_RESAMPLE;
  773. ret = swr_init(s);
  774. if (ret < 0)
  775. return ret;
  776. }
  777. if (!s->resampler->set_compensation){
  778. return AVERROR(EINVAL);
  779. }else{
  780. return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
  781. }
  782. }
  783. int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
  784. if(pts == INT64_MIN)
  785. return s->outpts;
  786. if (s->firstpts == AV_NOPTS_VALUE)
  787. s->outpts = s->firstpts = pts;
  788. if(s->min_compensation >= FLT_MAX) {
  789. return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
  790. } else {
  791. 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;
  792. double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
  793. if(fabs(fdelta) > s->min_compensation) {
  794. if(s->outpts == s->firstpts || fabs(fdelta) > s->min_hard_compensation){
  795. int ret;
  796. if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
  797. else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
  798. if(ret<0){
  799. av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
  800. }
  801. } else if(s->soft_compensation_duration && s->max_soft_compensation) {
  802. int duration = s->out_sample_rate * s->soft_compensation_duration;
  803. double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
  804. int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
  805. av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
  806. swr_set_compensation(s, comp, duration);
  807. }
  808. }
  809. return s->outpts;
  810. }
  811. }