swresample.c 34 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805
  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. #include "libavutil/opt.h"
  21. #include "swresample_internal.h"
  22. #include "audioconvert.h"
  23. #include "libavutil/avassert.h"
  24. #include "libavutil/audioconvert.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" , "Input Channel Count" , OFFSET( in.ch_count ), AV_OPT_TYPE_INT , {.i64=2 }, 0 , SWR_CH_MAX, PARAM},
  39. {"in_channel_count" , "Input Channel Count" , OFFSET( in.ch_count ), AV_OPT_TYPE_INT , {.i64=2 }, 0 , SWR_CH_MAX, PARAM},
  40. {"och" , "Output Channel Count" , OFFSET(out.ch_count ), AV_OPT_TYPE_INT , {.i64=2 }, 0 , SWR_CH_MAX, PARAM},
  41. {"out_channel_count" , "Output Channel Count" , OFFSET(out.ch_count ), AV_OPT_TYPE_INT , {.i64=2 }, 0 , SWR_CH_MAX, PARAM},
  42. {"uch" , "Used Channel Count" , OFFSET(used_ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  43. {"used_channel_count" , "Used Channel Count" , OFFSET(used_ch_count ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_CH_MAX, PARAM},
  44. {"isr" , "Input Sample Rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  45. {"in_sample_rate" , "Input Sample Rate" , OFFSET( in_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  46. {"osr" , "Output Sample Rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  47. {"out_sample_rate" , "Output Sample Rate" , OFFSET(out_sample_rate), AV_OPT_TYPE_INT , {.i64=0 }, 0 , INT_MAX , PARAM},
  48. {"isf" , "Input Sample Format" , OFFSET( in_sample_fmt ), AV_OPT_TYPE_INT , {.i64=AV_SAMPLE_FMT_NONE }, -1 , AV_SAMPLE_FMT_NB-1+256, PARAM},
  49. {"in_sample_fmt" , "Input Sample Format" , OFFSET( in_sample_fmt ), AV_OPT_TYPE_INT , {.i64=AV_SAMPLE_FMT_NONE }, -1 , AV_SAMPLE_FMT_NB-1+256, PARAM},
  50. {"osf" , "Output Sample Format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_INT , {.i64=AV_SAMPLE_FMT_NONE }, -1 , AV_SAMPLE_FMT_NB-1+256, PARAM},
  51. {"out_sample_fmt" , "Output Sample Format" , OFFSET(out_sample_fmt ), AV_OPT_TYPE_INT , {.i64=AV_SAMPLE_FMT_NONE }, -1 , AV_SAMPLE_FMT_NB-1+256, PARAM},
  52. {"tsf" , "Internal Sample Format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_INT , {.i64=AV_SAMPLE_FMT_NONE }, -1 , AV_SAMPLE_FMT_FLTP, PARAM},
  53. {"internal_sample_fmt" , "Internal Sample Format" , OFFSET(int_sample_fmt ), AV_OPT_TYPE_INT , {.i64=AV_SAMPLE_FMT_NONE }, -1 , AV_SAMPLE_FMT_FLTP, PARAM},
  54. {"icl" , "Input Channel Layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  55. {"in_channel_layout" , "Input Channel Layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  56. {"ocl" , "Output Channel Layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  57. {"out_channel_layout" , "Output Channel Layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  58. {"clev" , "Center Mix Level" , OFFSET(clev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  59. {"center_mix_level" , "Center Mix Level" , OFFSET(clev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  60. {"slev" , "Sourround Mix Level" , OFFSET(slev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  61. {"surround_mix_level" , "Sourround Mix Level" , OFFSET(slev ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB }, -32 , 32 , PARAM},
  62. {"lfe_mix_level" , "LFE Mix Level" , OFFSET(lfe_mix_level ), AV_OPT_TYPE_FLOAT, {.dbl=0 }, -32 , 32 , PARAM},
  63. {"rmvol" , "Rematrix Volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
  64. {"rematrix_volume" , "Rematrix Volume" , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0 }, -1000 , 1000 , PARAM},
  65. {"flags" , NULL , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
  66. {"swr_flags" , NULL , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
  67. {"res" , "Force Resampling" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_FLAG_RESAMPLE }, INT_MIN, INT_MAX , PARAM, "flags"},
  68. {"dither_scale" , "Dither Scale" , OFFSET(dither_scale ), AV_OPT_TYPE_FLOAT, {.dbl=1 }, 0 , INT_MAX , PARAM},
  69. {"dither_method" , "Dither Method" , OFFSET(dither_method ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_DITHER_NB-1, PARAM, "dither_method"},
  70. {"rectangular" , "Rectangular Dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX , PARAM, "dither_method"},
  71. {"triangular" , "Triangular Dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX , PARAM, "dither_method"},
  72. {"triangular_hp" , "Triangular Dither With High Pass" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR_HIGHPASS }, INT_MIN, INT_MAX, PARAM, "dither_method"},
  73. {"filter_size" , "Resampling Filter Size" , OFFSET(filter_size) , AV_OPT_TYPE_INT , {.i64=16 }, 0 , INT_MAX , PARAM },
  74. {"phase_shift" , "Resampling Phase Shift" , OFFSET(phase_shift) , AV_OPT_TYPE_INT , {.i64=10 }, 0 , 30 , PARAM },
  75. {"linear_interp" , "Use Linear Interpolation" , OFFSET(linear_interp) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
  76. {"cutoff" , "Cutoff Frequency Ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0.8 }, 0 , 1 , PARAM },
  77. {"min_comp" , "Minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
  78. , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX }, 0 , FLT_MAX , PARAM },
  79. {"min_hard_comp" , "Minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
  80. , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1 }, 0 , INT_MAX , PARAM },
  81. {"comp_duration" , "Duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
  82. , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1 }, 0 , INT_MAX , PARAM },
  83. {"max_soft_comp" , "Maximum factor by which data is stretched/squeezed to make it match the timestamps."
  84. , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
  85. { "matrix_encoding" , "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" },
  86. { "none", "None", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  87. { "dolby", "Dolby", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  88. { "dplii", "Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  89. { "filter_type" , "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" },
  90. { "cubic" , "Cubic" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  91. { "blackman_nuttall", "Blackman Nuttall Windowed Sinc", 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_BLACKMAN_NUTTALL }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  92. { "kaiser" , "Kaiser Windowed Sinc" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  93. { "kaiser_beta" , "Kaiser Window Beta" ,OFFSET(kaiser_beta) , AV_OPT_TYPE_INT , {.i64=9 }, 2 , 16 , PARAM },
  94. {0}
  95. };
  96. static const char* context_to_name(void* ptr) {
  97. return "SWR";
  98. }
  99. static const AVClass av_class = {
  100. .class_name = "SWResampler",
  101. .item_name = context_to_name,
  102. .option = options,
  103. .version = LIBAVUTIL_VERSION_INT,
  104. .log_level_offset_offset = OFFSET(log_level_offset),
  105. .parent_log_context_offset = OFFSET(log_ctx),
  106. .category = AV_CLASS_CATEGORY_SWRESAMPLER,
  107. };
  108. unsigned swresample_version(void)
  109. {
  110. av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
  111. return LIBSWRESAMPLE_VERSION_INT;
  112. }
  113. const char *swresample_configuration(void)
  114. {
  115. return FFMPEG_CONFIGURATION;
  116. }
  117. const char *swresample_license(void)
  118. {
  119. #define LICENSE_PREFIX "libswresample license: "
  120. return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
  121. }
  122. int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
  123. if(!s || s->in_convert) // s needs to be allocated but not initialized
  124. return AVERROR(EINVAL);
  125. s->channel_map = channel_map;
  126. return 0;
  127. }
  128. const AVClass *swr_get_class(void)
  129. {
  130. return &av_class;
  131. }
  132. av_cold struct SwrContext *swr_alloc(void){
  133. SwrContext *s= av_mallocz(sizeof(SwrContext));
  134. if(s){
  135. s->av_class= &av_class;
  136. av_opt_set_defaults(s);
  137. }
  138. return s;
  139. }
  140. struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
  141. int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
  142. int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate,
  143. int log_offset, void *log_ctx){
  144. if(!s) s= swr_alloc();
  145. if(!s) return NULL;
  146. s->log_level_offset= log_offset;
  147. s->log_ctx= log_ctx;
  148. av_opt_set_int(s, "ocl", out_ch_layout, 0);
  149. av_opt_set_int(s, "osf", out_sample_fmt, 0);
  150. av_opt_set_int(s, "osr", out_sample_rate, 0);
  151. av_opt_set_int(s, "icl", in_ch_layout, 0);
  152. av_opt_set_int(s, "isf", in_sample_fmt, 0);
  153. av_opt_set_int(s, "isr", in_sample_rate, 0);
  154. av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE, 0);
  155. av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
  156. av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
  157. av_opt_set_int(s, "uch", 0, 0);
  158. return s;
  159. }
  160. static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
  161. a->fmt = fmt;
  162. a->bps = av_get_bytes_per_sample(fmt);
  163. a->planar= av_sample_fmt_is_planar(fmt);
  164. }
  165. static void free_temp(AudioData *a){
  166. av_free(a->data);
  167. memset(a, 0, sizeof(*a));
  168. }
  169. av_cold void swr_free(SwrContext **ss){
  170. SwrContext *s= *ss;
  171. if(s){
  172. free_temp(&s->postin);
  173. free_temp(&s->midbuf);
  174. free_temp(&s->preout);
  175. free_temp(&s->in_buffer);
  176. free_temp(&s->dither);
  177. swri_audio_convert_free(&s-> in_convert);
  178. swri_audio_convert_free(&s->out_convert);
  179. swri_audio_convert_free(&s->full_convert);
  180. swri_resample_free(&s->resample);
  181. swri_rematrix_free(s);
  182. }
  183. av_freep(ss);
  184. }
  185. av_cold int swr_init(struct SwrContext *s){
  186. s->in_buffer_index= 0;
  187. s->in_buffer_count= 0;
  188. s->resample_in_constraint= 0;
  189. free_temp(&s->postin);
  190. free_temp(&s->midbuf);
  191. free_temp(&s->preout);
  192. free_temp(&s->in_buffer);
  193. free_temp(&s->dither);
  194. memset(s->in.ch, 0, sizeof(s->in.ch));
  195. memset(s->out.ch, 0, sizeof(s->out.ch));
  196. swri_audio_convert_free(&s-> in_convert);
  197. swri_audio_convert_free(&s->out_convert);
  198. swri_audio_convert_free(&s->full_convert);
  199. swri_rematrix_free(s);
  200. s->flushed = 0;
  201. if(av_get_channel_layout_nb_channels(s-> in_ch_layout) > SWR_CH_MAX) {
  202. av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
  203. s->in_ch_layout = 0;
  204. }
  205. if(av_get_channel_layout_nb_channels(s->out_ch_layout) > SWR_CH_MAX) {
  206. av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
  207. s->out_ch_layout = 0;
  208. }
  209. if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
  210. av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
  211. return AVERROR(EINVAL);
  212. }
  213. if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
  214. av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
  215. return AVERROR(EINVAL);
  216. }
  217. if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
  218. if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
  219. s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
  220. }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
  221. s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
  222. }else{
  223. av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
  224. s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
  225. }
  226. }
  227. if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
  228. &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
  229. &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
  230. &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
  231. 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));
  232. return AVERROR(EINVAL);
  233. }
  234. set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
  235. set_audiodata_fmt(&s->out, s->out_sample_fmt);
  236. if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
  237. s->resample = swri_resample_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);
  238. }else
  239. swri_resample_free(&s->resample);
  240. if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
  241. && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
  242. && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
  243. && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
  244. && s->resample){
  245. av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
  246. return -1;
  247. }
  248. if(!s->used_ch_count)
  249. s->used_ch_count= s->in.ch_count;
  250. if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
  251. av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
  252. s-> in_ch_layout= 0;
  253. }
  254. if(!s-> in_ch_layout)
  255. s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
  256. if(!s->out_ch_layout)
  257. s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);
  258. s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
  259. s->rematrix_custom;
  260. #define RSC 1 //FIXME finetune
  261. if(!s-> in.ch_count)
  262. s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
  263. if(!s->used_ch_count)
  264. s->used_ch_count= s->in.ch_count;
  265. if(!s->out.ch_count)
  266. s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);
  267. if(!s-> in.ch_count){
  268. av_assert0(!s->in_ch_layout);
  269. av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
  270. return -1;
  271. }
  272. if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
  273. av_log(s, AV_LOG_ERROR, "Rematrix is needed but there is not enough information to do it\n");
  274. return -1;
  275. }
  276. av_assert0(s->used_ch_count);
  277. av_assert0(s->out.ch_count);
  278. s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
  279. s->in_buffer= s->in;
  280. if(!s->resample && !s->rematrix && !s->channel_map && !s->dither_method){
  281. s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
  282. s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
  283. return 0;
  284. }
  285. s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
  286. s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
  287. s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
  288. s->int_sample_fmt, s->out.ch_count, NULL, 0);
  289. s->postin= s->in;
  290. s->preout= s->out;
  291. s->midbuf= s->in;
  292. if(s->channel_map){
  293. s->postin.ch_count=
  294. s->midbuf.ch_count= s->used_ch_count;
  295. if(s->resample)
  296. s->in_buffer.ch_count= s->used_ch_count;
  297. }
  298. if(!s->resample_first){
  299. s->midbuf.ch_count= s->out.ch_count;
  300. if(s->resample)
  301. s->in_buffer.ch_count = s->out.ch_count;
  302. }
  303. set_audiodata_fmt(&s->postin, s->int_sample_fmt);
  304. set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
  305. set_audiodata_fmt(&s->preout, s->int_sample_fmt);
  306. if(s->resample){
  307. set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
  308. }
  309. s->dither = s->preout;
  310. if(s->rematrix || s->dither_method)
  311. return swri_rematrix_init(s);
  312. return 0;
  313. }
  314. static int realloc_audio(AudioData *a, int count){
  315. int i, countb;
  316. AudioData old;
  317. if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
  318. return AVERROR(EINVAL);
  319. if(a->count >= count)
  320. return 0;
  321. count*=2;
  322. countb= FFALIGN(count*a->bps, ALIGN);
  323. old= *a;
  324. av_assert0(a->bps);
  325. av_assert0(a->ch_count);
  326. a->data= av_mallocz(countb*a->ch_count);
  327. if(!a->data)
  328. return AVERROR(ENOMEM);
  329. for(i=0; i<a->ch_count; i++){
  330. a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
  331. if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
  332. }
  333. if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
  334. av_free(old.data);
  335. a->count= count;
  336. return 1;
  337. }
  338. static void copy(AudioData *out, AudioData *in,
  339. int count){
  340. av_assert0(out->planar == in->planar);
  341. av_assert0(out->bps == in->bps);
  342. av_assert0(out->ch_count == in->ch_count);
  343. if(out->planar){
  344. int ch;
  345. for(ch=0; ch<out->ch_count; ch++)
  346. memcpy(out->ch[ch], in->ch[ch], count*out->bps);
  347. }else
  348. memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
  349. }
  350. static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  351. int i;
  352. if(!in_arg){
  353. memset(out->ch, 0, sizeof(out->ch));
  354. }else if(out->planar){
  355. for(i=0; i<out->ch_count; i++)
  356. out->ch[i]= in_arg[i];
  357. }else{
  358. for(i=0; i<out->ch_count; i++)
  359. out->ch[i]= in_arg[0] + i*out->bps;
  360. }
  361. }
  362. static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  363. int i;
  364. if(out->planar){
  365. for(i=0; i<out->ch_count; i++)
  366. in_arg[i]= out->ch[i];
  367. }else{
  368. in_arg[0]= out->ch[0];
  369. }
  370. }
  371. /**
  372. *
  373. * out may be equal in.
  374. */
  375. static void buf_set(AudioData *out, AudioData *in, int count){
  376. int ch;
  377. if(in->planar){
  378. for(ch=0; ch<out->ch_count; ch++)
  379. out->ch[ch]= in->ch[ch] + count*out->bps;
  380. }else{
  381. for(ch=out->ch_count-1; ch>=0; ch--)
  382. out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
  383. }
  384. }
  385. /**
  386. *
  387. * @return number of samples output per channel
  388. */
  389. static int resample(SwrContext *s, AudioData *out_param, int out_count,
  390. const AudioData * in_param, int in_count){
  391. AudioData in, out, tmp;
  392. int ret_sum=0;
  393. int border=0;
  394. av_assert1(s->in_buffer.ch_count == in_param->ch_count);
  395. av_assert1(s->in_buffer.planar == in_param->planar);
  396. av_assert1(s->in_buffer.fmt == in_param->fmt);
  397. tmp=out=*out_param;
  398. in = *in_param;
  399. do{
  400. int ret, size, consumed;
  401. if(!s->resample_in_constraint && s->in_buffer_count){
  402. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  403. ret= swri_multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
  404. out_count -= ret;
  405. ret_sum += ret;
  406. buf_set(&out, &out, ret);
  407. s->in_buffer_count -= consumed;
  408. s->in_buffer_index += consumed;
  409. if(!in_count)
  410. break;
  411. if(s->in_buffer_count <= border){
  412. buf_set(&in, &in, -s->in_buffer_count);
  413. in_count += s->in_buffer_count;
  414. s->in_buffer_count=0;
  415. s->in_buffer_index=0;
  416. border = 0;
  417. }
  418. }
  419. if(in_count && !s->in_buffer_count){
  420. s->in_buffer_index=0;
  421. ret= swri_multiple_resample(s->resample, &out, out_count, &in, in_count, &consumed);
  422. out_count -= ret;
  423. ret_sum += ret;
  424. buf_set(&out, &out, ret);
  425. in_count -= consumed;
  426. buf_set(&in, &in, consumed);
  427. }
  428. //TODO is this check sane considering the advanced copy avoidance below
  429. size= s->in_buffer_index + s->in_buffer_count + in_count;
  430. if( size > s->in_buffer.count
  431. && s->in_buffer_count + in_count <= s->in_buffer_index){
  432. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  433. copy(&s->in_buffer, &tmp, s->in_buffer_count);
  434. s->in_buffer_index=0;
  435. }else
  436. if((ret=realloc_audio(&s->in_buffer, size)) < 0)
  437. return ret;
  438. if(in_count){
  439. int count= in_count;
  440. if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
  441. buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
  442. copy(&tmp, &in, /*in_*/count);
  443. s->in_buffer_count += count;
  444. in_count -= count;
  445. border += count;
  446. buf_set(&in, &in, count);
  447. s->resample_in_constraint= 0;
  448. if(s->in_buffer_count != count || in_count)
  449. continue;
  450. }
  451. break;
  452. }while(1);
  453. s->resample_in_constraint= !!out_count;
  454. return ret_sum;
  455. }
  456. static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
  457. AudioData *in , int in_count){
  458. AudioData *postin, *midbuf, *preout;
  459. int ret/*, in_max*/;
  460. AudioData preout_tmp, midbuf_tmp;
  461. if(s->full_convert){
  462. av_assert0(!s->resample);
  463. swri_audio_convert(s->full_convert, out, in, in_count);
  464. return out_count;
  465. }
  466. // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
  467. // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
  468. if((ret=realloc_audio(&s->postin, in_count))<0)
  469. return ret;
  470. if(s->resample_first){
  471. av_assert0(s->midbuf.ch_count == s->used_ch_count);
  472. if((ret=realloc_audio(&s->midbuf, out_count))<0)
  473. return ret;
  474. }else{
  475. av_assert0(s->midbuf.ch_count == s->out.ch_count);
  476. if((ret=realloc_audio(&s->midbuf, in_count))<0)
  477. return ret;
  478. }
  479. if((ret=realloc_audio(&s->preout, out_count))<0)
  480. return ret;
  481. postin= &s->postin;
  482. midbuf_tmp= s->midbuf;
  483. midbuf= &midbuf_tmp;
  484. preout_tmp= s->preout;
  485. preout= &preout_tmp;
  486. if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
  487. postin= in;
  488. if(s->resample_first ? !s->resample : !s->rematrix)
  489. midbuf= postin;
  490. if(s->resample_first ? !s->rematrix : !s->resample)
  491. preout= midbuf;
  492. if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar){
  493. if(preout==in){
  494. out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
  495. av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
  496. copy(out, in, out_count);
  497. return out_count;
  498. }
  499. else if(preout==postin) preout= midbuf= postin= out;
  500. else if(preout==midbuf) preout= midbuf= out;
  501. else preout= out;
  502. }
  503. if(in != postin){
  504. swri_audio_convert(s->in_convert, postin, in, in_count);
  505. }
  506. if(s->resample_first){
  507. if(postin != midbuf)
  508. out_count= resample(s, midbuf, out_count, postin, in_count);
  509. if(midbuf != preout)
  510. swri_rematrix(s, preout, midbuf, out_count, preout==out);
  511. }else{
  512. if(postin != midbuf)
  513. swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
  514. if(midbuf != preout)
  515. out_count= resample(s, preout, out_count, midbuf, in_count);
  516. }
  517. if(preout != out && out_count){
  518. if(s->dither_method){
  519. int ch;
  520. int dither_count= FFMAX(out_count, 1<<16);
  521. av_assert0(preout != in);
  522. if((ret=realloc_audio(&s->dither, dither_count))<0)
  523. return ret;
  524. if(ret)
  525. for(ch=0; ch<s->dither.ch_count; ch++)
  526. swri_get_dither(s, s->dither.ch[ch], s->dither.count, 12345678913579<<ch, s->out_sample_fmt, s->int_sample_fmt);
  527. av_assert0(s->dither.ch_count == preout->ch_count);
  528. if(s->dither_pos + out_count > s->dither.count)
  529. s->dither_pos = 0;
  530. for(ch=0; ch<preout->ch_count; ch++)
  531. 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);
  532. s->dither_pos += out_count;
  533. }
  534. //FIXME packed doesnt need more than 1 chan here!
  535. swri_audio_convert(s->out_convert, out, preout, out_count);
  536. }
  537. return out_count;
  538. }
  539. int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
  540. const uint8_t *in_arg [SWR_CH_MAX], int in_count){
  541. AudioData * in= &s->in;
  542. AudioData *out= &s->out;
  543. if(s->drop_output > 0){
  544. int ret;
  545. AudioData tmp = s->out;
  546. uint8_t *tmp_arg[SWR_CH_MAX];
  547. tmp.count = 0;
  548. tmp.data = NULL;
  549. if((ret=realloc_audio(&tmp, s->drop_output))<0)
  550. return ret;
  551. reversefill_audiodata(&tmp, tmp_arg);
  552. s->drop_output *= -1; //FIXME find a less hackish solution
  553. 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
  554. s->drop_output *= -1;
  555. if(ret>0)
  556. s->drop_output -= ret;
  557. av_freep(&tmp.data);
  558. if(s->drop_output || !out_arg)
  559. return 0;
  560. in_count = 0;
  561. }
  562. if(!in_arg){
  563. if(s->in_buffer_count){
  564. if (s->resample && !s->flushed) {
  565. AudioData *a= &s->in_buffer;
  566. int i, j, ret;
  567. if((ret=realloc_audio(a, s->in_buffer_index + 2*s->in_buffer_count)) < 0)
  568. return ret;
  569. av_assert0(a->planar);
  570. for(i=0; i<a->ch_count; i++){
  571. for(j=0; j<s->in_buffer_count; j++){
  572. memcpy(a->ch[i] + (s->in_buffer_index+s->in_buffer_count+j )*a->bps,
  573. a->ch[i] + (s->in_buffer_index+s->in_buffer_count-j-1)*a->bps, a->bps);
  574. }
  575. }
  576. s->in_buffer_count += (s->in_buffer_count+1)/2;
  577. s->resample_in_constraint = 0;
  578. s->flushed = 1;
  579. }
  580. }else{
  581. return 0;
  582. }
  583. }else
  584. fill_audiodata(in , (void*)in_arg);
  585. fill_audiodata(out, out_arg);
  586. if(s->resample){
  587. int ret = swr_convert_internal(s, out, out_count, in, in_count);
  588. if(ret>0 && !s->drop_output)
  589. s->outpts += ret * (int64_t)s->in_sample_rate;
  590. return ret;
  591. }else{
  592. AudioData tmp= *in;
  593. int ret2=0;
  594. int ret, size;
  595. size = FFMIN(out_count, s->in_buffer_count);
  596. if(size){
  597. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  598. ret= swr_convert_internal(s, out, size, &tmp, size);
  599. if(ret<0)
  600. return ret;
  601. ret2= ret;
  602. s->in_buffer_count -= ret;
  603. s->in_buffer_index += ret;
  604. buf_set(out, out, ret);
  605. out_count -= ret;
  606. if(!s->in_buffer_count)
  607. s->in_buffer_index = 0;
  608. }
  609. if(in_count){
  610. size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
  611. if(in_count > out_count) { //FIXME move after swr_convert_internal
  612. if( size > s->in_buffer.count
  613. && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
  614. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  615. copy(&s->in_buffer, &tmp, s->in_buffer_count);
  616. s->in_buffer_index=0;
  617. }else
  618. if((ret=realloc_audio(&s->in_buffer, size)) < 0)
  619. return ret;
  620. }
  621. if(out_count){
  622. size = FFMIN(in_count, out_count);
  623. ret= swr_convert_internal(s, out, size, in, size);
  624. if(ret<0)
  625. return ret;
  626. buf_set(in, in, ret);
  627. in_count -= ret;
  628. ret2 += ret;
  629. }
  630. if(in_count){
  631. buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
  632. copy(&tmp, in, in_count);
  633. s->in_buffer_count += in_count;
  634. }
  635. }
  636. if(ret2>0 && !s->drop_output)
  637. s->outpts += ret2 * (int64_t)s->in_sample_rate;
  638. return ret2;
  639. }
  640. }
  641. int swr_drop_output(struct SwrContext *s, int count){
  642. s->drop_output += count;
  643. if(s->drop_output <= 0)
  644. return 0;
  645. av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
  646. return swr_convert(s, NULL, s->drop_output, NULL, 0);
  647. }
  648. int swr_inject_silence(struct SwrContext *s, int count){
  649. int ret, i;
  650. AudioData silence = s->in;
  651. uint8_t *tmp_arg[SWR_CH_MAX];
  652. if(count <= 0)
  653. return 0;
  654. silence.count = 0;
  655. silence.data = NULL;
  656. if((ret=realloc_audio(&silence, count))<0)
  657. return ret;
  658. if(silence.planar) for(i=0; i<silence.ch_count; i++) {
  659. memset(silence.ch[i], silence.bps==1 ? 0x80 : 0, count*silence.bps);
  660. } else
  661. memset(silence.ch[0], silence.bps==1 ? 0x80 : 0, count*silence.bps*silence.ch_count);
  662. reversefill_audiodata(&silence, tmp_arg);
  663. av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
  664. ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
  665. av_freep(&silence.data);
  666. return ret;
  667. }
  668. int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
  669. if(pts == INT64_MIN)
  670. return s->outpts;
  671. if(s->min_compensation >= FLT_MAX) {
  672. return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
  673. } else {
  674. 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;
  675. double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
  676. if(fabs(fdelta) > s->min_compensation) {
  677. if(!s->outpts || fabs(fdelta) > s->min_hard_compensation){
  678. int ret;
  679. if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
  680. else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
  681. if(ret<0){
  682. av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
  683. }
  684. } else if(s->soft_compensation_duration && s->max_soft_compensation) {
  685. int duration = s->out_sample_rate * s->soft_compensation_duration;
  686. double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
  687. int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
  688. av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
  689. swr_set_compensation(s, comp, duration);
  690. }
  691. }
  692. return s->outpts;
  693. }
  694. }