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.

848 lines
36KB

  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/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 , AV_SAMPLE_FMT_NB-1, PARAM},
  49. {"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},
  50. {"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},
  51. {"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},
  52. {"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},
  53. {"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},
  54. {"icl" , "set input channel layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  55. {"in_channel_layout" , "set input channel layout" , OFFSET( in_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  56. {"ocl" , "set output channel layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_INT64, {.i64=0 }, 0 , INT64_MAX , PARAM, "channel_layout"},
  57. {"out_channel_layout" , "set output channel layout" , OFFSET(out_ch_layout ), AV_OPT_TYPE_INT64, {.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. {"flags" , "set flags" , OFFSET(flags ), AV_OPT_TYPE_FLAGS, {.i64=0 }, 0 , UINT_MAX , PARAM, "flags"},
  66. {"swr_flags" , "set flags" , 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" , "set dither scale" , OFFSET(dither_scale ), AV_OPT_TYPE_FLOAT, {.dbl=1 }, 0 , INT_MAX , PARAM},
  69. {"dither_method" , "set dither method" , OFFSET(dither_method ), AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_DITHER_NB-1, PARAM, "dither_method"},
  70. {"rectangular" , "select rectangular dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX , PARAM, "dither_method"},
  71. {"triangular" , "select triangular dither" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX , PARAM, "dither_method"},
  72. {"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"},
  73. {"filter_size" , "set swr resampling filter size", OFFSET(filter_size) , AV_OPT_TYPE_INT , {.i64=16 }, 0 , INT_MAX , PARAM },
  74. {"phase_shift" , "set swr resampling phase shift", OFFSET(phase_shift) , AV_OPT_TYPE_INT , {.i64=10 }, 0 , 30 , PARAM },
  75. {"linear_interp" , "enable linear interpolation" , OFFSET(linear_interp) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
  76. {"cutoff" , "set cutoff frequency ratio" , OFFSET(cutoff) , AV_OPT_TYPE_DOUBLE,{.dbl=0. }, 0 , 1 , PARAM },
  77. {"resampler" , "set resampling Engine" , OFFSET(engine) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , SWR_ENGINE_NB-1, PARAM, "resampler"},
  78. {"swr" , "select SW Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SWR }, INT_MIN, INT_MAX , PARAM, "resampler"},
  79. {"soxr" , "select SoX Resampler" , 0 , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SOXR }, INT_MIN, INT_MAX , PARAM, "resampler"},
  80. {"precision" , "set soxr resampling precision (in bits)"
  81. , OFFSET(precision) , AV_OPT_TYPE_DOUBLE,{.dbl=20.0 }, 15.0 , 33.0 , PARAM },
  82. {"cheby" , "enable soxr Chebyshev passband & higher-precision irrational ratio approximation"
  83. , OFFSET(cheby) , AV_OPT_TYPE_INT , {.i64=0 }, 0 , 1 , PARAM },
  84. {"min_comp" , "set minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
  85. , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX }, 0 , FLT_MAX , PARAM },
  86. {"min_hard_comp" , "set minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
  87. , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1 }, 0 , INT_MAX , PARAM },
  88. {"comp_duration" , "set duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
  89. , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1 }, 0 , INT_MAX , PARAM },
  90. {"max_soft_comp" , "set maximum factor by which data is stretched/squeezed to make it match the timestamps."
  91. , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
  92. {"async" , "simplified 1 parameter audio timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum stretch/squeeze in samples per second)"
  93. , OFFSET(async) , AV_OPT_TYPE_FLOAT ,{.dbl=0 }, INT_MIN, INT_MAX , PARAM },
  94. { "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" },
  95. { "none", "select none", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  96. { "dolby", "select Dolby", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  97. { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
  98. { "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" },
  99. { "cubic" , "select cubic" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  100. { "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" },
  101. { "kaiser" , "select Kaiser Windowed Sinc" , 0 , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER }, INT_MIN, INT_MAX, PARAM, "filter_type" },
  102. { "kaiser_beta" , "set swr Kaiser Window Beta" , OFFSET(kaiser_beta) , AV_OPT_TYPE_INT , {.i64=9 }, 2 , 16 , PARAM },
  103. {0}
  104. };
  105. static const char* context_to_name(void* ptr) {
  106. return "SWR";
  107. }
  108. static const AVClass av_class = {
  109. .class_name = "SWResampler",
  110. .item_name = context_to_name,
  111. .option = options,
  112. .version = LIBAVUTIL_VERSION_INT,
  113. .log_level_offset_offset = OFFSET(log_level_offset),
  114. .parent_log_context_offset = OFFSET(log_ctx),
  115. .category = AV_CLASS_CATEGORY_SWRESAMPLER,
  116. };
  117. unsigned swresample_version(void)
  118. {
  119. av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
  120. return LIBSWRESAMPLE_VERSION_INT;
  121. }
  122. const char *swresample_configuration(void)
  123. {
  124. return FFMPEG_CONFIGURATION;
  125. }
  126. const char *swresample_license(void)
  127. {
  128. #define LICENSE_PREFIX "libswresample license: "
  129. return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
  130. }
  131. int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
  132. if(!s || s->in_convert) // s needs to be allocated but not initialized
  133. return AVERROR(EINVAL);
  134. s->channel_map = channel_map;
  135. return 0;
  136. }
  137. const AVClass *swr_get_class(void)
  138. {
  139. return &av_class;
  140. }
  141. av_cold struct SwrContext *swr_alloc(void){
  142. SwrContext *s= av_mallocz(sizeof(SwrContext));
  143. if(s){
  144. s->av_class= &av_class;
  145. av_opt_set_defaults(s);
  146. }
  147. return s;
  148. }
  149. struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
  150. int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
  151. int64_t in_ch_layout, enum AVSampleFormat in_sample_fmt, int in_sample_rate,
  152. int log_offset, void *log_ctx){
  153. if(!s) s= swr_alloc();
  154. if(!s) return NULL;
  155. s->log_level_offset= log_offset;
  156. s->log_ctx= log_ctx;
  157. av_opt_set_int(s, "ocl", out_ch_layout, 0);
  158. av_opt_set_int(s, "osf", out_sample_fmt, 0);
  159. av_opt_set_int(s, "osr", out_sample_rate, 0);
  160. av_opt_set_int(s, "icl", in_ch_layout, 0);
  161. av_opt_set_int(s, "isf", in_sample_fmt, 0);
  162. av_opt_set_int(s, "isr", in_sample_rate, 0);
  163. av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE, 0);
  164. av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
  165. av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
  166. av_opt_set_int(s, "uch", 0, 0);
  167. return s;
  168. }
  169. static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
  170. a->fmt = fmt;
  171. a->bps = av_get_bytes_per_sample(fmt);
  172. a->planar= av_sample_fmt_is_planar(fmt);
  173. }
  174. static void free_temp(AudioData *a){
  175. av_free(a->data);
  176. memset(a, 0, sizeof(*a));
  177. }
  178. av_cold void swr_free(SwrContext **ss){
  179. SwrContext *s= *ss;
  180. if(s){
  181. free_temp(&s->postin);
  182. free_temp(&s->midbuf);
  183. free_temp(&s->preout);
  184. free_temp(&s->in_buffer);
  185. free_temp(&s->dither);
  186. swri_audio_convert_free(&s-> in_convert);
  187. swri_audio_convert_free(&s->out_convert);
  188. swri_audio_convert_free(&s->full_convert);
  189. if (s->resampler)
  190. s->resampler->free(&s->resample);
  191. swri_rematrix_free(s);
  192. }
  193. av_freep(ss);
  194. }
  195. av_cold int swr_init(struct SwrContext *s){
  196. s->in_buffer_index= 0;
  197. s->in_buffer_count= 0;
  198. s->resample_in_constraint= 0;
  199. free_temp(&s->postin);
  200. free_temp(&s->midbuf);
  201. free_temp(&s->preout);
  202. free_temp(&s->in_buffer);
  203. free_temp(&s->dither);
  204. memset(s->in.ch, 0, sizeof(s->in.ch));
  205. memset(s->out.ch, 0, sizeof(s->out.ch));
  206. swri_audio_convert_free(&s-> in_convert);
  207. swri_audio_convert_free(&s->out_convert);
  208. swri_audio_convert_free(&s->full_convert);
  209. swri_rematrix_free(s);
  210. s->flushed = 0;
  211. if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
  212. av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
  213. return AVERROR(EINVAL);
  214. }
  215. if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
  216. av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
  217. return AVERROR(EINVAL);
  218. }
  219. if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
  220. if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
  221. s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
  222. }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
  223. s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
  224. }else{
  225. av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
  226. s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
  227. }
  228. }
  229. if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
  230. &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
  231. &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
  232. &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
  233. 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));
  234. return AVERROR(EINVAL);
  235. }
  236. switch(s->engine){
  237. #if CONFIG_LIBSOXR
  238. extern struct Resampler const soxr_resampler;
  239. case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
  240. #endif
  241. case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
  242. default:
  243. av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
  244. return AVERROR(EINVAL);
  245. }
  246. set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
  247. set_audiodata_fmt(&s->out, s->out_sample_fmt);
  248. if (s->async) {
  249. if (s->min_compensation >= FLT_MAX/2)
  250. s->min_compensation = 0.001;
  251. if (s->async > 1.0001) {
  252. s->max_soft_compensation = s->async / (double) s->in_sample_rate;
  253. }
  254. }
  255. if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
  256. 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);
  257. }else
  258. s->resampler->free(&s->resample);
  259. if( s->int_sample_fmt != AV_SAMPLE_FMT_S16P
  260. && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
  261. && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
  262. && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
  263. && s->resample){
  264. av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
  265. return -1;
  266. }
  267. if(!s->used_ch_count)
  268. s->used_ch_count= s->in.ch_count;
  269. if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
  270. av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
  271. s-> in_ch_layout= 0;
  272. }
  273. if(!s-> in_ch_layout)
  274. s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
  275. if(!s->out_ch_layout)
  276. s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);
  277. s->rematrix= s->out_ch_layout !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
  278. s->rematrix_custom;
  279. #define RSC 1 //FIXME finetune
  280. if(!s-> in.ch_count)
  281. s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
  282. if(!s->used_ch_count)
  283. s->used_ch_count= s->in.ch_count;
  284. if(!s->out.ch_count)
  285. s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);
  286. if(!s-> in.ch_count){
  287. av_assert0(!s->in_ch_layout);
  288. av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
  289. return -1;
  290. }
  291. if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
  292. av_log(s, AV_LOG_ERROR, "Rematrix is needed but there is not enough information to do it\n");
  293. return -1;
  294. }
  295. av_assert0(s->used_ch_count);
  296. av_assert0(s->out.ch_count);
  297. s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;
  298. s->in_buffer= s->in;
  299. if(!s->resample && !s->rematrix && !s->channel_map && !s->dither_method){
  300. s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
  301. s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
  302. return 0;
  303. }
  304. s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
  305. s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
  306. s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
  307. s->int_sample_fmt, s->out.ch_count, NULL, 0);
  308. s->postin= s->in;
  309. s->preout= s->out;
  310. s->midbuf= s->in;
  311. if(s->channel_map){
  312. s->postin.ch_count=
  313. s->midbuf.ch_count= s->used_ch_count;
  314. if(s->resample)
  315. s->in_buffer.ch_count= s->used_ch_count;
  316. }
  317. if(!s->resample_first){
  318. s->midbuf.ch_count= s->out.ch_count;
  319. if(s->resample)
  320. s->in_buffer.ch_count = s->out.ch_count;
  321. }
  322. set_audiodata_fmt(&s->postin, s->int_sample_fmt);
  323. set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
  324. set_audiodata_fmt(&s->preout, s->int_sample_fmt);
  325. if(s->resample){
  326. set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
  327. }
  328. s->dither = s->preout;
  329. if(s->rematrix || s->dither_method)
  330. return swri_rematrix_init(s);
  331. return 0;
  332. }
  333. int swri_realloc_audio(AudioData *a, int count){
  334. int i, countb;
  335. AudioData old;
  336. if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
  337. return AVERROR(EINVAL);
  338. if(a->count >= count)
  339. return 0;
  340. count*=2;
  341. countb= FFALIGN(count*a->bps, ALIGN);
  342. old= *a;
  343. av_assert0(a->bps);
  344. av_assert0(a->ch_count);
  345. a->data= av_mallocz(countb*a->ch_count);
  346. if(!a->data)
  347. return AVERROR(ENOMEM);
  348. for(i=0; i<a->ch_count; i++){
  349. a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
  350. if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
  351. }
  352. if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
  353. av_free(old.data);
  354. a->count= count;
  355. return 1;
  356. }
  357. static void copy(AudioData *out, AudioData *in,
  358. int count){
  359. av_assert0(out->planar == in->planar);
  360. av_assert0(out->bps == in->bps);
  361. av_assert0(out->ch_count == in->ch_count);
  362. if(out->planar){
  363. int ch;
  364. for(ch=0; ch<out->ch_count; ch++)
  365. memcpy(out->ch[ch], in->ch[ch], count*out->bps);
  366. }else
  367. memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
  368. }
  369. static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  370. int i;
  371. if(!in_arg){
  372. memset(out->ch, 0, sizeof(out->ch));
  373. }else if(out->planar){
  374. for(i=0; i<out->ch_count; i++)
  375. out->ch[i]= in_arg[i];
  376. }else{
  377. for(i=0; i<out->ch_count; i++)
  378. out->ch[i]= in_arg[0] + i*out->bps;
  379. }
  380. }
  381. static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  382. int i;
  383. if(out->planar){
  384. for(i=0; i<out->ch_count; i++)
  385. in_arg[i]= out->ch[i];
  386. }else{
  387. in_arg[0]= out->ch[0];
  388. }
  389. }
  390. /**
  391. *
  392. * out may be equal in.
  393. */
  394. static void buf_set(AudioData *out, AudioData *in, int count){
  395. int ch;
  396. if(in->planar){
  397. for(ch=0; ch<out->ch_count; ch++)
  398. out->ch[ch]= in->ch[ch] + count*out->bps;
  399. }else{
  400. for(ch=out->ch_count-1; ch>=0; ch--)
  401. out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
  402. }
  403. }
  404. /**
  405. *
  406. * @return number of samples output per channel
  407. */
  408. static int resample(SwrContext *s, AudioData *out_param, int out_count,
  409. const AudioData * in_param, int in_count){
  410. AudioData in, out, tmp;
  411. int ret_sum=0;
  412. int border=0;
  413. av_assert1(s->in_buffer.ch_count == in_param->ch_count);
  414. av_assert1(s->in_buffer.planar == in_param->planar);
  415. av_assert1(s->in_buffer.fmt == in_param->fmt);
  416. tmp=out=*out_param;
  417. in = *in_param;
  418. do{
  419. int ret, size, consumed;
  420. if(!s->resample_in_constraint && s->in_buffer_count){
  421. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  422. ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
  423. out_count -= ret;
  424. ret_sum += ret;
  425. buf_set(&out, &out, ret);
  426. s->in_buffer_count -= consumed;
  427. s->in_buffer_index += consumed;
  428. if(!in_count)
  429. break;
  430. if(s->in_buffer_count <= border){
  431. buf_set(&in, &in, -s->in_buffer_count);
  432. in_count += s->in_buffer_count;
  433. s->in_buffer_count=0;
  434. s->in_buffer_index=0;
  435. border = 0;
  436. }
  437. }
  438. if((s->flushed || in_count) && !s->in_buffer_count){
  439. s->in_buffer_index=0;
  440. ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, in_count, &consumed);
  441. out_count -= ret;
  442. ret_sum += ret;
  443. buf_set(&out, &out, ret);
  444. in_count -= consumed;
  445. buf_set(&in, &in, consumed);
  446. }
  447. //TODO is this check sane considering the advanced copy avoidance below
  448. size= s->in_buffer_index + s->in_buffer_count + in_count;
  449. if( size > s->in_buffer.count
  450. && s->in_buffer_count + in_count <= s->in_buffer_index){
  451. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  452. copy(&s->in_buffer, &tmp, s->in_buffer_count);
  453. s->in_buffer_index=0;
  454. }else
  455. if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
  456. return ret;
  457. if(in_count){
  458. int count= in_count;
  459. if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;
  460. buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
  461. copy(&tmp, &in, /*in_*/count);
  462. s->in_buffer_count += count;
  463. in_count -= count;
  464. border += count;
  465. buf_set(&in, &in, count);
  466. s->resample_in_constraint= 0;
  467. if(s->in_buffer_count != count || in_count)
  468. continue;
  469. }
  470. break;
  471. }while(1);
  472. s->resample_in_constraint= !!out_count;
  473. return ret_sum;
  474. }
  475. static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
  476. AudioData *in , int in_count){
  477. AudioData *postin, *midbuf, *preout;
  478. int ret/*, in_max*/;
  479. AudioData preout_tmp, midbuf_tmp;
  480. if(s->full_convert){
  481. av_assert0(!s->resample);
  482. swri_audio_convert(s->full_convert, out, in, in_count);
  483. return out_count;
  484. }
  485. // in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
  486. // in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);
  487. if((ret=swri_realloc_audio(&s->postin, in_count))<0)
  488. return ret;
  489. if(s->resample_first){
  490. av_assert0(s->midbuf.ch_count == s->used_ch_count);
  491. if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
  492. return ret;
  493. }else{
  494. av_assert0(s->midbuf.ch_count == s->out.ch_count);
  495. if((ret=swri_realloc_audio(&s->midbuf, in_count))<0)
  496. return ret;
  497. }
  498. if((ret=swri_realloc_audio(&s->preout, out_count))<0)
  499. return ret;
  500. postin= &s->postin;
  501. midbuf_tmp= s->midbuf;
  502. midbuf= &midbuf_tmp;
  503. preout_tmp= s->preout;
  504. preout= &preout_tmp;
  505. if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
  506. postin= in;
  507. if(s->resample_first ? !s->resample : !s->rematrix)
  508. midbuf= postin;
  509. if(s->resample_first ? !s->rematrix : !s->resample)
  510. preout= midbuf;
  511. if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar){
  512. if(preout==in){
  513. out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
  514. av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
  515. copy(out, in, out_count);
  516. return out_count;
  517. }
  518. else if(preout==postin) preout= midbuf= postin= out;
  519. else if(preout==midbuf) preout= midbuf= out;
  520. else preout= out;
  521. }
  522. if(in != postin){
  523. swri_audio_convert(s->in_convert, postin, in, in_count);
  524. }
  525. if(s->resample_first){
  526. if(postin != midbuf)
  527. out_count= resample(s, midbuf, out_count, postin, in_count);
  528. if(midbuf != preout)
  529. swri_rematrix(s, preout, midbuf, out_count, preout==out);
  530. }else{
  531. if(postin != midbuf)
  532. swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
  533. if(midbuf != preout)
  534. out_count= resample(s, preout, out_count, midbuf, in_count);
  535. }
  536. if(preout != out && out_count){
  537. if(s->dither_method){
  538. int ch;
  539. int dither_count= FFMAX(out_count, 1<<16);
  540. av_assert0(preout != in);
  541. if((ret=swri_realloc_audio(&s->dither, dither_count))<0)
  542. return ret;
  543. if(ret)
  544. for(ch=0; ch<s->dither.ch_count; ch++)
  545. swri_get_dither(s, s->dither.ch[ch], s->dither.count, 12345678913579<<ch, s->out_sample_fmt, s->int_sample_fmt);
  546. av_assert0(s->dither.ch_count == preout->ch_count);
  547. if(s->dither_pos + out_count > s->dither.count)
  548. s->dither_pos = 0;
  549. for(ch=0; ch<preout->ch_count; ch++)
  550. 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);
  551. s->dither_pos += out_count;
  552. }
  553. //FIXME packed doesnt need more than 1 chan here!
  554. swri_audio_convert(s->out_convert, out, preout, out_count);
  555. }
  556. return out_count;
  557. }
  558. int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
  559. const uint8_t *in_arg [SWR_CH_MAX], int in_count){
  560. AudioData * in= &s->in;
  561. AudioData *out= &s->out;
  562. if(s->drop_output > 0){
  563. int ret;
  564. AudioData tmp = s->out;
  565. uint8_t *tmp_arg[SWR_CH_MAX];
  566. tmp.count = 0;
  567. tmp.data = NULL;
  568. if((ret=swri_realloc_audio(&tmp, s->drop_output))<0)
  569. return ret;
  570. reversefill_audiodata(&tmp, tmp_arg);
  571. s->drop_output *= -1; //FIXME find a less hackish solution
  572. 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
  573. s->drop_output *= -1;
  574. if(ret>0)
  575. s->drop_output -= ret;
  576. av_freep(&tmp.data);
  577. if(s->drop_output || !out_arg)
  578. return 0;
  579. in_count = 0;
  580. }
  581. if(!in_arg){
  582. if(s->resample){
  583. if (!s->flushed)
  584. s->resampler->flush(s);
  585. s->resample_in_constraint = 0;
  586. s->flushed = 1;
  587. }else if(!s->in_buffer_count){
  588. return 0;
  589. }
  590. }else
  591. fill_audiodata(in , (void*)in_arg);
  592. fill_audiodata(out, out_arg);
  593. if(s->resample){
  594. int ret = swr_convert_internal(s, out, out_count, in, in_count);
  595. if(ret>0 && !s->drop_output)
  596. s->outpts += ret * (int64_t)s->in_sample_rate;
  597. return ret;
  598. }else{
  599. AudioData tmp= *in;
  600. int ret2=0;
  601. int ret, size;
  602. size = FFMIN(out_count, s->in_buffer_count);
  603. if(size){
  604. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  605. ret= swr_convert_internal(s, out, size, &tmp, size);
  606. if(ret<0)
  607. return ret;
  608. ret2= ret;
  609. s->in_buffer_count -= ret;
  610. s->in_buffer_index += ret;
  611. buf_set(out, out, ret);
  612. out_count -= ret;
  613. if(!s->in_buffer_count)
  614. s->in_buffer_index = 0;
  615. }
  616. if(in_count){
  617. size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;
  618. if(in_count > out_count) { //FIXME move after swr_convert_internal
  619. if( size > s->in_buffer.count
  620. && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
  621. buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
  622. copy(&s->in_buffer, &tmp, s->in_buffer_count);
  623. s->in_buffer_index=0;
  624. }else
  625. if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
  626. return ret;
  627. }
  628. if(out_count){
  629. size = FFMIN(in_count, out_count);
  630. ret= swr_convert_internal(s, out, size, in, size);
  631. if(ret<0)
  632. return ret;
  633. buf_set(in, in, ret);
  634. in_count -= ret;
  635. ret2 += ret;
  636. }
  637. if(in_count){
  638. buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
  639. copy(&tmp, in, in_count);
  640. s->in_buffer_count += in_count;
  641. }
  642. }
  643. if(ret2>0 && !s->drop_output)
  644. s->outpts += ret2 * (int64_t)s->in_sample_rate;
  645. return ret2;
  646. }
  647. }
  648. int swr_drop_output(struct SwrContext *s, int count){
  649. s->drop_output += count;
  650. if(s->drop_output <= 0)
  651. return 0;
  652. av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
  653. return swr_convert(s, NULL, s->drop_output, NULL, 0);
  654. }
  655. int swr_inject_silence(struct SwrContext *s, int count){
  656. int ret, i;
  657. AudioData silence = s->in;
  658. uint8_t *tmp_arg[SWR_CH_MAX];
  659. if(count <= 0)
  660. return 0;
  661. silence.count = 0;
  662. silence.data = NULL;
  663. if((ret=swri_realloc_audio(&silence, count))<0)
  664. return ret;
  665. if(silence.planar) for(i=0; i<silence.ch_count; i++) {
  666. memset(silence.ch[i], silence.bps==1 ? 0x80 : 0, count*silence.bps);
  667. } else
  668. memset(silence.ch[0], silence.bps==1 ? 0x80 : 0, count*silence.bps*silence.ch_count);
  669. reversefill_audiodata(&silence, tmp_arg);
  670. av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
  671. ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
  672. av_freep(&silence.data);
  673. return ret;
  674. }
  675. int64_t swr_get_delay(struct SwrContext *s, int64_t base){
  676. if (s->resampler && s->resample){
  677. return s->resampler->get_delay(s, base);
  678. }else{
  679. return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
  680. }
  681. }
  682. int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
  683. int ret;
  684. if (!s || compensation_distance < 0)
  685. return AVERROR(EINVAL);
  686. if (!compensation_distance && sample_delta)
  687. return AVERROR(EINVAL);
  688. if (!s->resample) {
  689. s->flags |= SWR_FLAG_RESAMPLE;
  690. ret = swr_init(s);
  691. if (ret < 0)
  692. return ret;
  693. }
  694. if (!s->resampler->set_compensation){
  695. return AVERROR(EINVAL);
  696. }else{
  697. return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
  698. }
  699. }
  700. int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
  701. if(pts == INT64_MIN)
  702. return s->outpts;
  703. if(s->min_compensation >= FLT_MAX) {
  704. return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
  705. } else {
  706. int64_t delta = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate) - s->outpts;
  707. double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);
  708. if(fabs(fdelta) > s->min_compensation) {
  709. if(!s->outpts || fabs(fdelta) > s->min_hard_compensation){
  710. int ret;
  711. if(delta > 0) ret = swr_inject_silence(s, delta / s->out_sample_rate);
  712. else ret = swr_drop_output (s, -delta / s-> in_sample_rate);
  713. if(ret<0){
  714. av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
  715. }
  716. } else if(s->soft_compensation_duration && s->max_soft_compensation) {
  717. int duration = s->out_sample_rate * s->soft_compensation_duration;
  718. double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
  719. int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
  720. av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
  721. swr_set_compensation(s, comp, duration);
  722. }
  723. }
  724. return s->outpts;
  725. }
  726. }