swresample.c 41.9 KB
Newer Older
Michael Niedermayer's avatar
Michael Niedermayer committed
1
/*
2
 * Copyright (C) 2011-2013 Michael Niedermayer (michaelni@gmx.at)
Michael Niedermayer's avatar
Michael Niedermayer committed
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 *
 * This file is part of libswresample
 *
 * libswresample is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * libswresample is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with libswresample; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavutil/opt.h"
#include "swresample_internal.h"
#include "audioconvert.h"
#include "libavutil/avassert.h"
25
#include "libavutil/channel_layout.h"
Michael Niedermayer's avatar
Michael Niedermayer committed
26

27 28
#include <float.h>

Michael Niedermayer's avatar
Michael Niedermayer committed
29 30 31 32 33 34 35 36
#define  C30DB  M_SQRT2
#define  C15DB  1.189207115
#define C__0DB  1.0
#define C_15DB  0.840896415
#define C_30DB  M_SQRT1_2
#define C_45DB  0.594603558
#define C_60DB  0.5

37
#define ALIGN 32
Michael Niedermayer's avatar
Michael Niedermayer committed
38 39 40

//TODO split options array out?
#define OFFSET(x) offsetof(SwrContext,x)
41 42
#define PARAM AV_OPT_FLAG_AUDIO_PARAM

Michael Niedermayer's avatar
Michael Niedermayer committed
43
static const AVOption options[]={
44 45 46 47
{"ich"                  , "set input channel count"     , OFFSET( in.ch_count   ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_CH_MAX, PARAM},
{"in_channel_count"     , "set input channel count"     , OFFSET( in.ch_count   ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_CH_MAX, PARAM},
{"och"                  , "set output channel count"    , OFFSET(out.ch_count   ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_CH_MAX, PARAM},
{"out_channel_count"    , "set output channel count"    , OFFSET(out.ch_count   ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_CH_MAX, PARAM},
48 49 50 51 52 53
{"uch"                  , "set used channel count"      , OFFSET(used_ch_count  ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_CH_MAX, PARAM},
{"used_channel_count"   , "set used channel count"      , OFFSET(used_ch_count  ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_CH_MAX, PARAM},
{"isr"                  , "set input sample rate"       , OFFSET( in_sample_rate), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , INT_MAX   , PARAM},
{"in_sample_rate"       , "set input sample rate"       , OFFSET( in_sample_rate), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , INT_MAX   , PARAM},
{"osr"                  , "set output sample rate"      , OFFSET(out_sample_rate), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , INT_MAX   , PARAM},
{"out_sample_rate"      , "set output sample rate"      , OFFSET(out_sample_rate), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , INT_MAX   , PARAM},
54 55 56 57 58 59
{"isf"                  , "set input sample format"     , OFFSET( in_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1   , INT_MAX, PARAM},
{"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},
{"osf"                  , "set output sample format"    , OFFSET(out_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1   , INT_MAX, PARAM},
{"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},
{"tsf"                  , "set internal sample format"  , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1   , INT_MAX, PARAM},
{"internal_sample_fmt"  , "set internal sample format"  , OFFSET(int_sample_fmt ), AV_OPT_TYPE_SAMPLE_FMT , {.i64=AV_SAMPLE_FMT_NONE}, -1   , INT_MAX, PARAM},
60 61 62 63
{"icl"                  , "set input channel layout"    , OFFSET( in_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
{"in_channel_layout"    , "set input channel layout"    , OFFSET( in_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
{"ocl"                  , "set output channel layout"   , OFFSET(out_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
{"out_channel_layout"   , "set output channel layout"   , OFFSET(out_ch_layout  ), AV_OPT_TYPE_CHANNEL_LAYOUT, {.i64=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
64 65 66 67 68 69 70
{"clev"                 , "set center mix level"        , OFFSET(clev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
{"center_mix_level"     , "set center mix level"        , OFFSET(clev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
{"slev"                 , "set surround mix level"      , OFFSET(slev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
{"surround_mix_level"   , "set surround mix Level"      , OFFSET(slev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
{"lfe_mix_level"        , "set LFE mix level"           , OFFSET(lfe_mix_level  ), AV_OPT_TYPE_FLOAT, {.dbl=0                     }, -32    , 32        , PARAM},
{"rmvol"                , "set rematrix volume"         , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0                   }, -1000  , 1000      , PARAM},
{"rematrix_volume"      , "set rematrix volume"         , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0                   }, -1000  , 1000      , PARAM},
71
{"rematrix_maxval"      , "set rematrix maxval"         , OFFSET(rematrix_maxval), AV_OPT_TYPE_FLOAT, {.dbl=0.0                   }, 0      , 1000      , PARAM},
72 73 74 75 76

{"flags"                , "set flags"                   , OFFSET(flags          ), AV_OPT_TYPE_FLAGS, {.i64=0                     }, 0      , UINT_MAX  , PARAM, "flags"},
{"swr_flags"            , "set flags"                   , OFFSET(flags          ), AV_OPT_TYPE_FLAGS, {.i64=0                     }, 0      , UINT_MAX  , PARAM, "flags"},
{"res"                  , "force resampling"            , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_FLAG_RESAMPLE     }, INT_MIN, INT_MAX   , PARAM, "flags"},

77
{"dither_scale"         , "set dither scale"            , OFFSET(dither.scale   ), AV_OPT_TYPE_FLOAT, {.dbl=1                     }, 0      , INT_MAX   , PARAM},
78

79
{"dither_method"        , "set dither method"           , OFFSET(dither.method  ), AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_DITHER_NB-1, PARAM, "dither_method"},
80 81
{"rectangular"          , "select rectangular dither"   , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX   , PARAM, "dither_method"},
{"triangular"           , "select triangular dither"    , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX   , PARAM, "dither_method"},
82 83 84 85 86 87 88 89
{"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"},
{"lipshitz"             , "select lipshitz noise shaping dither" , 0             , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_LIPSHITZ}, INT_MIN, INT_MAX, PARAM, "dither_method"},
{"shibata"              , "select shibata noise shaping dither" , 0              , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_SHIBATA }, INT_MIN, INT_MAX, PARAM, "dither_method"},
{"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"},
{"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"},
{"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"},
{"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"},
{"improved_e_weighted"  , "select improved-e-weighted noise shaping dither" , 0  , AV_OPT_TYPE_CONST, {.i64=SWR_DITHER_NS_IMPROVED_E_WEIGHTED }, INT_MIN, INT_MAX, PARAM, "dither_method"},
90

91
{"filter_size"          , "set swr resampling filter size", OFFSET(filter_size)  , AV_OPT_TYPE_INT  , {.i64=32                    }, 0      , INT_MAX   , PARAM },
92
{"phase_shift"          , "set swr resampling phase shift", OFFSET(phase_shift)  , AV_OPT_TYPE_INT  , {.i64=10                    }, 0      , 24        , PARAM },
93
{"linear_interp"        , "enable linear interpolation" , OFFSET(linear_interp)  , AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , 1         , PARAM },
94
{"cutoff"               , "set cutoff frequency ratio"  , OFFSET(cutoff)         , AV_OPT_TYPE_DOUBLE,{.dbl=0.                    }, 0      , 1         , PARAM },
95 96 97 98

/* duplicate option in order to work with avconv */
{"resample_cutoff"      , "set cutoff frequency ratio"  , OFFSET(cutoff)         , AV_OPT_TYPE_DOUBLE,{.dbl=0.                    }, 0      , 1         , PARAM },

99 100
{"resampler"            , "set resampling Engine"       , OFFSET(engine)         , AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , SWR_ENGINE_NB-1, PARAM, "resampler"},
{"swr"                  , "select SW Resampler"         , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SWR        }, INT_MIN, INT_MAX   , PARAM, "resampler"},
101
{"soxr"                 , "select SoX Resampler"        , 0                      , AV_OPT_TYPE_CONST, {.i64=SWR_ENGINE_SOXR       }, INT_MIN, INT_MAX   , PARAM, "resampler"},
Rob Sykes's avatar
Rob Sykes committed
102 103 104 105
{"precision"            , "set soxr resampling precision (in bits)"
                                                        , OFFSET(precision)      , AV_OPT_TYPE_DOUBLE,{.dbl=20.0                  }, 15.0   , 33.0      , PARAM },
{"cheby"                , "enable soxr Chebyshev passband & higher-precision irrational ratio approximation"
                                                        , OFFSET(cheby)          , AV_OPT_TYPE_INT  , {.i64=0                     }, 0      , 1         , PARAM },
106
{"min_comp"             , "set minimum difference between timestamps and audio data (in seconds) below which no timestamp compensation of either kind is applied"
107
                                                        , OFFSET(min_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=FLT_MAX               }, 0      , FLT_MAX   , PARAM },
108 109 110 111 112 113
{"min_hard_comp"        , "set minimum difference between timestamps and audio data (in seconds) to trigger padding/trimming the data."
                                                        , OFFSET(min_hard_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0.1                   }, 0      , INT_MAX   , PARAM },
{"comp_duration"        , "set duration (in seconds) over which data is stretched/squeezed to make it match the timestamps."
                                                        , OFFSET(soft_compensation_duration),AV_OPT_TYPE_FLOAT ,{.dbl=1                     }, 0      , INT_MAX   , PARAM },
{"max_soft_comp"        , "set maximum factor by which data is stretched/squeezed to make it match the timestamps."
                                                        , OFFSET(max_soft_compensation),AV_OPT_TYPE_FLOAT ,{.dbl=0                     }, INT_MIN, INT_MAX   , PARAM },
114 115
{"async"                , "simplified 1 parameter audio timestamp matching, 0(disabled), 1(filling and trimming), >1(maximum stretch/squeeze in samples per second)"
                                                        , OFFSET(async)          , AV_OPT_TYPE_FLOAT ,{.dbl=0                     }, INT_MIN, INT_MAX   , PARAM },
116 117
{"first_pts"            , "Assume the first pts should be this value (in samples)."
                                                        , OFFSET(firstpts_in_samples), AV_OPT_TYPE_INT64 ,{.i64=AV_NOPTS_VALUE    }, INT64_MIN,INT64_MAX, PARAM },
118 119 120 121 122 123

{ "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" },
    { "none",  "select none",               0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_NONE  }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
    { "dolby", "select Dolby",              0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DOLBY }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },
    { "dplii", "select Dolby Pro Logic II", 0, AV_OPT_TYPE_CONST, { .i64 = AV_MATRIX_ENCODING_DPLII }, INT_MIN, INT_MAX, PARAM, "matrix_encoding" },

Rob Sykes's avatar
Rob Sykes committed
124
{ "filter_type"         , "select swr filter type"      , OFFSET(filter_type)    , AV_OPT_TYPE_INT  , { .i64 = SWR_FILTER_TYPE_KAISER }, SWR_FILTER_TYPE_CUBIC, SWR_FILTER_TYPE_KAISER, PARAM, "filter_type" },
125 126 127 128
    { "cubic"           , "select cubic"                , 0                      , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_CUBIC            }, INT_MIN, INT_MAX, PARAM, "filter_type" },
    { "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" },
    { "kaiser"          , "select Kaiser Windowed Sinc" , 0                      , AV_OPT_TYPE_CONST, { .i64 = SWR_FILTER_TYPE_KAISER           }, INT_MIN, INT_MAX, PARAM, "filter_type" },

Rob Sykes's avatar
Rob Sykes committed
129
{ "kaiser_beta"         , "set swr Kaiser Window Beta"  , OFFSET(kaiser_beta)    , AV_OPT_TYPE_INT  , {.i64=9                     }, 2      , 16        , PARAM },
130

131
{ "output_sample_bits"  , "set swr number of output sample bits", OFFSET(dither.output_sample_bits), AV_OPT_TYPE_INT  , {.i64=0   }, 0      , 64        , PARAM },
Michael Niedermayer's avatar
Michael Niedermayer committed
132 133 134 135 136 137 138
{0}
};

static const char* context_to_name(void* ptr) {
    return "SWR";
}

139
static const AVClass av_class = {
140
    .class_name                = "SWResampler",
141 142 143 144 145
    .item_name                 = context_to_name,
    .option                    = options,
    .version                   = LIBAVUTIL_VERSION_INT,
    .log_level_offset_offset   = OFFSET(log_level_offset),
    .parent_log_context_offset = OFFSET(log_ctx),
Paul B Mahol's avatar
Paul B Mahol committed
146
    .category                  = AV_CLASS_CATEGORY_SWRESAMPLER,
147
};
Michael Niedermayer's avatar
Michael Niedermayer committed
148

149 150
unsigned swresample_version(void)
{
151
    av_assert0(LIBSWRESAMPLE_VERSION_MICRO >= 100);
152
    return LIBSWRESAMPLE_VERSION_INT;
153 154 155 156 157 158 159 160 161 162 163 164 165
}

const char *swresample_configuration(void)
{
    return FFMPEG_CONFIGURATION;
}

const char *swresample_license(void)
{
#define LICENSE_PREFIX "libswresample license: "
    return LICENSE_PREFIX FFMPEG_LICENSE + sizeof(LICENSE_PREFIX) - 1;
}

166 167 168 169 170 171 172
int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
    if(!s || s->in_convert) // s needs to be allocated but not initialized
        return AVERROR(EINVAL);
    s->channel_map = channel_map;
    return 0;
}

173 174 175 176 177
const AVClass *swr_get_class(void)
{
    return &av_class;
}

178
av_cold struct SwrContext *swr_alloc(void){
Michael Niedermayer's avatar
Michael Niedermayer committed
179 180 181
    SwrContext *s= av_mallocz(sizeof(SwrContext));
    if(s){
        s->av_class= &av_class;
182
        av_opt_set_defaults(s);
Michael Niedermayer's avatar
Michael Niedermayer committed
183 184 185 186
    }
    return s;
}

187 188 189
struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
                                      int64_t out_ch_layout, enum AVSampleFormat out_sample_fmt, int out_sample_rate,
                                      int64_t  in_ch_layout, enum AVSampleFormat  in_sample_fmt, int  in_sample_rate,
190
                                      int log_offset, void *log_ctx){
Michael Niedermayer's avatar
Michael Niedermayer committed
191 192 193 194 195 196
    if(!s) s= swr_alloc();
    if(!s) return NULL;

    s->log_level_offset= log_offset;
    s->log_ctx= log_ctx;

197 198 199 200 201 202
    av_opt_set_int(s, "ocl", out_ch_layout,   0);
    av_opt_set_int(s, "osf", out_sample_fmt,  0);
    av_opt_set_int(s, "osr", out_sample_rate, 0);
    av_opt_set_int(s, "icl", in_ch_layout,    0);
    av_opt_set_int(s, "isf", in_sample_fmt,   0);
    av_opt_set_int(s, "isr", in_sample_rate,  0);
203
    av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE,   0);
204 205
    av_opt_set_int(s, "ich", av_get_channel_layout_nb_channels(s-> in_ch_layout), 0);
    av_opt_set_int(s, "och", av_get_channel_layout_nb_channels(s->out_ch_layout), 0);
Clément Bœsch's avatar
Clément Bœsch committed
206
    av_opt_set_int(s, "uch", 0, 0);
Michael Niedermayer's avatar
Michael Niedermayer committed
207 208 209
    return s;
}

210
static void set_audiodata_fmt(AudioData *a, enum AVSampleFormat fmt){
211
    a->fmt   = fmt;
212 213 214
    a->bps   = av_get_bytes_per_sample(fmt);
    a->planar= av_sample_fmt_is_planar(fmt);
}
Michael Niedermayer's avatar
Michael Niedermayer committed
215 216 217 218 219 220

static void free_temp(AudioData *a){
    av_free(a->data);
    memset(a, 0, sizeof(*a));
}

221
static void clear_context(SwrContext *s){
Michael Niedermayer's avatar
Michael Niedermayer committed
222 223 224
    s->in_buffer_index= 0;
    s->in_buffer_count= 0;
    s->resample_in_constraint= 0;
225 226
    memset(s->in.ch, 0, sizeof(s->in.ch));
    memset(s->out.ch, 0, sizeof(s->out.ch));
Michael Niedermayer's avatar
Michael Niedermayer committed
227 228 229 230
    free_temp(&s->postin);
    free_temp(&s->midbuf);
    free_temp(&s->preout);
    free_temp(&s->in_buffer);
231
    free_temp(&s->silence);
232
    free_temp(&s->drop_temp);
233
    free_temp(&s->dither.noise);
234
    free_temp(&s->dither.temp);
235 236 237
    swri_audio_convert_free(&s-> in_convert);
    swri_audio_convert_free(&s->out_convert);
    swri_audio_convert_free(&s->full_convert);
238
    swri_rematrix_free(s);
Michael Niedermayer's avatar
Michael Niedermayer committed
239

240
    s->flushed = 0;
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257
}

av_cold void swr_free(SwrContext **ss){
    SwrContext *s= *ss;
    if(s){
        clear_context(s);
        if (s->resampler)
            s->resampler->free(&s->resample);
    }

    av_freep(ss);
}

av_cold int swr_init(struct SwrContext *s){
    int ret;

    clear_context(s);
258

259
    if(s-> in_sample_fmt >= AV_SAMPLE_FMT_NB){
260
        av_log(s, AV_LOG_ERROR, "Requested input sample format %d is invalid\n", s->in_sample_fmt);
261 262 263
        return AVERROR(EINVAL);
    }
    if(s->out_sample_fmt >= AV_SAMPLE_FMT_NB){
264
        av_log(s, AV_LOG_ERROR, "Requested output sample format %d is invalid\n", s->out_sample_fmt);
265 266 267
        return AVERROR(EINVAL);
    }

268 269 270 271 272 273 274 275 276 277
    if(av_get_channel_layout_nb_channels(s-> in_ch_layout) > SWR_CH_MAX) {
        av_log(s, AV_LOG_WARNING, "Input channel layout 0x%"PRIx64" is invalid or unsupported.\n", s-> in_ch_layout);
        s->in_ch_layout = 0;
    }

    if(av_get_channel_layout_nb_channels(s->out_ch_layout) > SWR_CH_MAX) {
        av_log(s, AV_LOG_WARNING, "Output channel layout 0x%"PRIx64" is invalid or unsupported.\n", s->out_ch_layout);
        s->out_ch_layout = 0;
    }

278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304
    switch(s->engine){
#if CONFIG_LIBSOXR
        extern struct Resampler const soxr_resampler;
        case SWR_ENGINE_SOXR: s->resampler = &soxr_resampler; break;
#endif
        case SWR_ENGINE_SWR : s->resampler = &swri_resampler; break;
        default:
            av_log(s, AV_LOG_ERROR, "Requested resampling engine is unavailable\n");
            return AVERROR(EINVAL);
    }

    if(!s->used_ch_count)
        s->used_ch_count= s->in.ch_count;

    if(s->used_ch_count && s-> in_ch_layout && s->used_ch_count != av_get_channel_layout_nb_channels(s-> in_ch_layout)){
        av_log(s, AV_LOG_WARNING, "Input channel layout has a different number of channels than the number of used channels, ignoring layout\n");
        s-> in_ch_layout= 0;
    }

    if(!s-> in_ch_layout)
        s-> in_ch_layout= av_get_default_channel_layout(s->used_ch_count);
    if(!s->out_ch_layout)
        s->out_ch_layout= av_get_default_channel_layout(s->out.ch_count);

    s->rematrix= s->out_ch_layout  !=s->in_ch_layout || s->rematrix_volume!=1.0 ||
                 s->rematrix_custom;

305 306 307
    if(s->int_sample_fmt == AV_SAMPLE_FMT_NONE){
        if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_S16P){
            s->int_sample_fmt= AV_SAMPLE_FMT_S16P;
308 309 310 311 312
        }else if(   av_get_planar_sample_fmt(s-> in_sample_fmt) == AV_SAMPLE_FMT_S32P
                 && av_get_planar_sample_fmt(s->out_sample_fmt) == AV_SAMPLE_FMT_S32P
                 && !s->rematrix
                 && s->engine != SWR_ENGINE_SOXR){
            s->int_sample_fmt= AV_SAMPLE_FMT_S32P;
313 314 315
        }else if(av_get_planar_sample_fmt(s->in_sample_fmt) <= AV_SAMPLE_FMT_FLTP){
            s->int_sample_fmt= AV_SAMPLE_FMT_FLTP;
        }else{
Clément Bœsch's avatar
Clément Bœsch committed
316
            av_log(s, AV_LOG_DEBUG, "Using double precision mode\n");
317 318 319
            s->int_sample_fmt= AV_SAMPLE_FMT_DBLP;
        }
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
320

321 322
    if(   s->int_sample_fmt != AV_SAMPLE_FMT_S16P
        &&s->int_sample_fmt != AV_SAMPLE_FMT_S32P
323 324 325
        &&s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
        &&s->int_sample_fmt != AV_SAMPLE_FMT_DBLP){
        av_log(s, AV_LOG_ERROR, "Requested sample format %s is not supported internally, S16/S32/FLT/DBL is supported\n", av_get_sample_fmt_name(s->int_sample_fmt));
326 327
        return AVERROR(EINVAL);
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
328

329 330 331
    set_audiodata_fmt(&s-> in, s-> in_sample_fmt);
    set_audiodata_fmt(&s->out, s->out_sample_fmt);

332 333 334 335 336
    if (s->firstpts_in_samples != AV_NOPTS_VALUE) {
        if (!s->async && s->min_compensation >= FLT_MAX/2)
            s->async = 1;
        s->firstpts =
        s->outpts   = s->firstpts_in_samples * s->out_sample_rate;
337 338
    } else
        s->firstpts = AV_NOPTS_VALUE;
339

340 341 342 343 344 345 346 347
    if (s->async) {
        if (s->min_compensation >= FLT_MAX/2)
            s->min_compensation = 0.001;
        if (s->async > 1.0001) {
            s->max_soft_compensation = s->async / (double) s->in_sample_rate;
        }
    }

Michael Niedermayer's avatar
Michael Niedermayer committed
348
    if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
349
        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);
Michael Niedermayer's avatar
Michael Niedermayer committed
350
    }else
351
        s->resampler->free(&s->resample);
352 353 354
    if(    s->int_sample_fmt != AV_SAMPLE_FMT_S16P
        && s->int_sample_fmt != AV_SAMPLE_FMT_S32P
        && s->int_sample_fmt != AV_SAMPLE_FMT_FLTP
355
        && s->int_sample_fmt != AV_SAMPLE_FMT_DBLP
356
        && s->resample){
357
        av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt/dbl\n");
Michael Niedermayer's avatar
Michael Niedermayer committed
358 359 360 361 362 363
        return -1;
    }

#define RSC 1 //FIXME finetune
    if(!s-> in.ch_count)
        s-> in.ch_count= av_get_channel_layout_nb_channels(s-> in_ch_layout);
364 365
    if(!s->used_ch_count)
        s->used_ch_count= s->in.ch_count;
Michael Niedermayer's avatar
Michael Niedermayer committed
366 367 368
    if(!s->out.ch_count)
        s->out.ch_count= av_get_channel_layout_nb_channels(s->out_ch_layout);

369 370 371 372 373 374
    if(!s-> in.ch_count){
        av_assert0(!s->in_ch_layout);
        av_log(s, AV_LOG_ERROR, "Input channel count and layout are unset\n");
        return -1;
    }

375
    if ((!s->out_ch_layout || !s->in_ch_layout) && s->used_ch_count != s->out.ch_count && !s->rematrix_custom) {
376 377 378 379 380
        char l1[1024], l2[1024];
        av_get_channel_layout_string(l1, sizeof(l1), s-> in.ch_count, s-> in_ch_layout);
        av_get_channel_layout_string(l2, sizeof(l2), s->out.ch_count, s->out_ch_layout);
        av_log(s, AV_LOG_ERROR, "Rematrix is needed between %s and %s "
               "but there is not enough information to do it\n", l1, l2);
381 382 383
        return -1;
    }

384
av_assert0(s->used_ch_count);
Michael Niedermayer's avatar
Michael Niedermayer committed
385 386 387
av_assert0(s->out.ch_count);
    s->resample_first= RSC*s->out.ch_count/s->in.ch_count - RSC < s->out_sample_rate/(float)s-> in_sample_rate - 1.0;

388
    s->in_buffer= s->in;
389
    s->silence  = s->in;
390
    s->drop_temp= s->out;
Michael Niedermayer's avatar
Michael Niedermayer committed
391

392
    if(!s->resample && !s->rematrix && !s->channel_map && !s->dither.method){
393 394
        s->full_convert = swri_audio_convert_alloc(s->out_sample_fmt,
                                                   s-> in_sample_fmt, s-> in.ch_count, NULL, 0);
395 396 397
        return 0;
    }

398 399 400 401
    s->in_convert = swri_audio_convert_alloc(s->int_sample_fmt,
                                             s-> in_sample_fmt, s->used_ch_count, s->channel_map, 0);
    s->out_convert= swri_audio_convert_alloc(s->out_sample_fmt,
                                             s->int_sample_fmt, s->out.ch_count, NULL, 0);
Michael Niedermayer's avatar
Michael Niedermayer committed
402

403 404
    if (!s->in_convert || !s->out_convert)
        return AVERROR(ENOMEM);
Michael Niedermayer's avatar
Michael Niedermayer committed
405 406 407 408

    s->postin= s->in;
    s->preout= s->out;
    s->midbuf= s->in;
409

410 411
    if(s->channel_map){
        s->postin.ch_count=
412 413 414
        s->midbuf.ch_count= s->used_ch_count;
        if(s->resample)
            s->in_buffer.ch_count= s->used_ch_count;
415
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
416 417
    if(!s->resample_first){
        s->midbuf.ch_count= s->out.ch_count;
418 419
        if(s->resample)
            s->in_buffer.ch_count = s->out.ch_count;
Michael Niedermayer's avatar
Michael Niedermayer committed
420 421
    }

422 423 424
    set_audiodata_fmt(&s->postin, s->int_sample_fmt);
    set_audiodata_fmt(&s->midbuf, s->int_sample_fmt);
    set_audiodata_fmt(&s->preout, s->int_sample_fmt);
Michael Niedermayer's avatar
Michael Niedermayer committed
425

426
    if(s->resample){
427
        set_audiodata_fmt(&s->in_buffer, s->int_sample_fmt);
428
    }
Michael Niedermayer's avatar
Michael Niedermayer committed
429

430 431
    if ((ret = swri_dither_init(s, s->out_sample_fmt, s->int_sample_fmt)) < 0)
        return ret;
432

433
    if(s->rematrix || s->dither.method)
434
        return swri_rematrix_init(s);
Michael Niedermayer's avatar
Michael Niedermayer committed
435 436 437 438

    return 0;
}

439
int swri_realloc_audio(AudioData *a, int count){
Michael Niedermayer's avatar
Michael Niedermayer committed
440 441 442
    int i, countb;
    AudioData old;

443 444 445
    if(count < 0 || count > INT_MAX/2/a->bps/a->ch_count)
        return AVERROR(EINVAL);

Michael Niedermayer's avatar
Michael Niedermayer committed
446 447 448 449 450
    if(a->count >= count)
        return 0;

    count*=2;

451
    countb= FFALIGN(count*a->bps, ALIGN);
Michael Niedermayer's avatar
Michael Niedermayer committed
452 453 454 455 456
    old= *a;

    av_assert0(a->bps);
    av_assert0(a->ch_count);

457
    a->data= av_mallocz(countb*a->ch_count);
Michael Niedermayer's avatar
Michael Niedermayer committed
458 459 460 461 462 463
    if(!a->data)
        return AVERROR(ENOMEM);
    for(i=0; i<a->ch_count; i++){
        a->ch[i]= a->data + i*(a->planar ? countb : a->bps);
        if(a->planar) memcpy(a->ch[i], old.ch[i], a->count*a->bps);
    }
464
    if(!a->planar) memcpy(a->ch[0], old.ch[0], a->count*a->ch_count*a->bps);
465
    av_freep(&old.data);
Michael Niedermayer's avatar
Michael Niedermayer committed
466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483
    a->count= count;

    return 1;
}

static void copy(AudioData *out, AudioData *in,
                 int count){
    av_assert0(out->planar == in->planar);
    av_assert0(out->bps == in->bps);
    av_assert0(out->ch_count == in->ch_count);
    if(out->planar){
        int ch;
        for(ch=0; ch<out->ch_count; ch++)
            memcpy(out->ch[ch], in->ch[ch], count*out->bps);
    }else
        memcpy(out->ch[0], in->ch[0], count*out->ch_count*out->bps);
}

484 485
static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
    int i;
486 487 488
    if(!in_arg){
        memset(out->ch, 0, sizeof(out->ch));
    }else if(out->planar){
489 490 491 492 493 494 495 496
        for(i=0; i<out->ch_count; i++)
            out->ch[i]= in_arg[i];
    }else{
        for(i=0; i<out->ch_count; i++)
            out->ch[i]= in_arg[0] + i*out->bps;
    }
}

497 498 499 500 501 502 503 504 505 506
static void reversefill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
    int i;
    if(out->planar){
        for(i=0; i<out->ch_count; i++)
            in_arg[i]= out->ch[i];
    }else{
        in_arg[0]= out->ch[0];
    }
}

507 508 509 510 511
/**
 *
 * out may be equal in.
 */
static void buf_set(AudioData *out, AudioData *in, int count){
512
    int ch;
513 514 515
    if(in->planar){
        for(ch=0; ch<out->ch_count; ch++)
            out->ch[ch]= in->ch[ch] + count*out->bps;
516
    }else{
517
        for(ch=out->ch_count-1; ch>=0; ch--)
518 519
            out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
    }
520 521 522 523 524 525 526 527 528 529 530
}

/**
 *
 * @return number of samples output per channel
 */
static int resample(SwrContext *s, AudioData *out_param, int out_count,
                             const AudioData * in_param, int in_count){
    AudioData in, out, tmp;
    int ret_sum=0;
    int border=0;
531
    int padless = ARCH_X86 && s->engine == SWR_ENGINE_SWR ? 7 : 0;
532

533 534 535 536
    av_assert1(s->in_buffer.ch_count == in_param->ch_count);
    av_assert1(s->in_buffer.planar   == in_param->planar);
    av_assert1(s->in_buffer.fmt      == in_param->fmt);

537 538 539 540 541 542 543
    tmp=out=*out_param;
    in =  *in_param;

    do{
        int ret, size, consumed;
        if(!s->resample_in_constraint && s->in_buffer_count){
            buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
544
            ret= s->resampler->multiple_resample(s->resample, &out, out_count, &tmp, s->in_buffer_count, &consumed);
545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
            out_count -= ret;
            ret_sum += ret;
            buf_set(&out, &out, ret);
            s->in_buffer_count -= consumed;
            s->in_buffer_index += consumed;

            if(!in_count)
                break;
            if(s->in_buffer_count <= border){
                buf_set(&in, &in, -s->in_buffer_count);
                in_count += s->in_buffer_count;
                s->in_buffer_count=0;
                s->in_buffer_index=0;
                border = 0;
            }
        }

562
        if((s->flushed || in_count > padless) && !s->in_buffer_count){
563
            s->in_buffer_index=0;
564
            ret= s->resampler->multiple_resample(s->resample, &out, out_count, &in, FFMAX(in_count-padless, 0), &consumed);
565 566 567 568 569 570 571 572 573 574 575 576 577 578 579
            out_count -= ret;
            ret_sum += ret;
            buf_set(&out, &out, ret);
            in_count -= consumed;
            buf_set(&in, &in, consumed);
        }

        //TODO is this check sane considering the advanced copy avoidance below
        size= s->in_buffer_index + s->in_buffer_count + in_count;
        if(   size > s->in_buffer.count
           && s->in_buffer_count + in_count <= s->in_buffer_index){
            buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
            copy(&s->in_buffer, &tmp, s->in_buffer_count);
            s->in_buffer_index=0;
        }else
580
            if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
                return ret;

        if(in_count){
            int count= in_count;
            if(s->in_buffer_count && s->in_buffer_count+2 < count && out_count) count= s->in_buffer_count+2;

            buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
            copy(&tmp, &in, /*in_*/count);
            s->in_buffer_count += count;
            in_count -= count;
            border += count;
            buf_set(&in, &in, count);
            s->resample_in_constraint= 0;
            if(s->in_buffer_count != count || in_count)
                continue;
596 597 598 599
            if (padless) {
                padless = 0;
                continue;
            }
600 601 602 603 604 605 606 607 608
        }
        break;
    }while(1);

    s->resample_in_constraint= !!out_count;

    return ret_sum;
}

609 610
static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_count,
                                                      AudioData *in , int  in_count){
Michael Niedermayer's avatar
Michael Niedermayer committed
611
    AudioData *postin, *midbuf, *preout;
612
    int ret/*, in_max*/;
Michael Niedermayer's avatar
Michael Niedermayer committed
613 614
    AudioData preout_tmp, midbuf_tmp;

615 616
    if(s->full_convert){
        av_assert0(!s->resample);
617
        swri_audio_convert(s->full_convert, out, in, in_count);
618 619 620
        return out_count;
    }

Michael Niedermayer's avatar
Michael Niedermayer committed
621 622 623
//     in_max= out_count*(int64_t)s->in_sample_rate / s->out_sample_rate + resample_filter_taps;
//     in_count= FFMIN(in_count, in_in + 2 - s->hist_buffer_count);

624
    if((ret=swri_realloc_audio(&s->postin, in_count))<0)
Michael Niedermayer's avatar
Michael Niedermayer committed
625 626
        return ret;
    if(s->resample_first){
627
        av_assert0(s->midbuf.ch_count == s->used_ch_count);
628
        if((ret=swri_realloc_audio(&s->midbuf, out_count))<0)
Michael Niedermayer's avatar
Michael Niedermayer committed
629 630 631
            return ret;
    }else{
        av_assert0(s->midbuf.ch_count ==  s->out.ch_count);
632
        if((ret=swri_realloc_audio(&s->midbuf,  in_count))<0)
Michael Niedermayer's avatar
Michael Niedermayer committed
633 634
            return ret;
    }
635
    if((ret=swri_realloc_audio(&s->preout, out_count))<0)
Michael Niedermayer's avatar
Michael Niedermayer committed
636 637 638 639 640 641 642 643 644
        return ret;

    postin= &s->postin;

    midbuf_tmp= s->midbuf;
    midbuf= &midbuf_tmp;
    preout_tmp= s->preout;
    preout= &preout_tmp;

645
    if(s->int_sample_fmt == s-> in_sample_fmt && s->in.planar && !s->channel_map)
Michael Niedermayer's avatar
Michael Niedermayer committed
646 647 648 649 650 651 652 653
        postin= in;

    if(s->resample_first ? !s->resample : !s->rematrix)
        midbuf= postin;

    if(s->resample_first ? !s->rematrix : !s->resample)
        preout= midbuf;

654 655
    if(s->int_sample_fmt == s->out_sample_fmt && s->out.planar
       && !(s->out_sample_fmt==AV_SAMPLE_FMT_S32P && (s->dither.output_sample_bits&31))){
Michael Niedermayer's avatar
Michael Niedermayer committed
656
        if(preout==in){
Lou Logan's avatar
Lou Logan committed
657
            out_count= FFMIN(out_count, in_count); //TODO check at the end if this is needed or redundant
Michael Niedermayer's avatar
Michael Niedermayer committed
658 659 660 661 662 663 664 665 666 667
            av_assert0(s->in.planar); //we only support planar internally so it has to be, we support copying non planar though
            copy(out, in, out_count);
            return out_count;
        }
        else if(preout==postin) preout= midbuf= postin= out;
        else if(preout==midbuf) preout= midbuf= out;
        else                    preout= out;
    }

    if(in != postin){
668
        swri_audio_convert(s->in_convert, postin, in, in_count);
Michael Niedermayer's avatar
Michael Niedermayer committed
669 670 671 672 673 674
    }

    if(s->resample_first){
        if(postin != midbuf)
            out_count= resample(s, midbuf, out_count, postin, in_count);
        if(midbuf != preout)
675
            swri_rematrix(s, preout, midbuf, out_count, preout==out);
Michael Niedermayer's avatar
Michael Niedermayer committed
676 677
    }else{
        if(postin != midbuf)
678
            swri_rematrix(s, midbuf, postin, in_count, midbuf==out);
Michael Niedermayer's avatar
Michael Niedermayer committed
679 680 681 682
        if(midbuf != preout)
            out_count= resample(s, preout, out_count, midbuf, in_count);
    }

683
    if(preout != out && out_count){
684
        AudioData *conv_src = preout;
685
        if(s->dither.method){
686
            int ch;
687
            int dither_count= FFMAX(out_count, 1<<16);
688 689 690 691 692 693

            if (preout == in) {
                conv_src = &s->dither.temp;
                if((ret=swri_realloc_audio(&s->dither.temp, dither_count))<0)
                    return ret;
            }
694

695
            if((ret=swri_realloc_audio(&s->dither.noise, dither_count))<0)
696 697
                return ret;
            if(ret)
698
                for(ch=0; ch<s->dither.noise.ch_count; ch++)
699
                    swri_get_dither(s, s->dither.noise.ch[ch], s->dither.noise.count, 12345678913579<<ch, s->dither.noise.fmt);
700
            av_assert0(s->dither.noise.ch_count == preout->ch_count);
701

702 703
            if(s->dither.noise_pos + out_count > s->dither.noise.count)
                s->dither.noise_pos = 0;
704

705
            if (s->dither.method < SWR_DITHER_NS){
706 707 708 709 710 711
                if (s->mix_2_1_simd) {
                    int len1= out_count&~15;
                    int off = len1 * preout->bps;

                    if(len1)
                        for(ch=0; ch<preout->ch_count; ch++)
712
                            s->mix_2_1_simd(conv_src->ch[ch], preout->ch[ch], s->dither.noise.ch[ch] + s->dither.noise.bps * s->dither.noise_pos, s->native_simd_one, 0, 0, len1);
713 714
                    if(out_count != len1)
                        for(ch=0; ch<preout->ch_count; ch++)
715
                            s->mix_2_1_f(conv_src->ch[ch] + off, preout->ch[ch] + off, s->dither.noise.ch[ch] + s->dither.noise.bps * s->dither.noise_pos + off + len1, s->native_one, 0, 0, out_count - len1);
716
                } else {
717
                    for(ch=0; ch<preout->ch_count; ch++)
718
                        s->mix_2_1_f(conv_src->ch[ch], preout->ch[ch], s->dither.noise.ch[ch] + s->dither.noise.bps * s->dither.noise_pos, s->native_one, 0, 0, out_count);
719
                }
720
            } else {
721
                switch(s->int_sample_fmt) {
722 723 724 725
                case AV_SAMPLE_FMT_S16P :swri_noise_shaping_int16(s, conv_src, preout, &s->dither.noise, out_count); break;
                case AV_SAMPLE_FMT_S32P :swri_noise_shaping_int32(s, conv_src, preout, &s->dither.noise, out_count); break;
                case AV_SAMPLE_FMT_FLTP :swri_noise_shaping_float(s, conv_src, preout, &s->dither.noise, out_count); break;
                case AV_SAMPLE_FMT_DBLP :swri_noise_shaping_double(s,conv_src, preout, &s->dither.noise, out_count); break;
726
                }
727
            }
728
            s->dither.noise_pos += out_count;
729
        }
730
//FIXME packed doesn't need more than 1 chan here!
731
        swri_audio_convert(s->out_convert, out, conv_src, out_count);
Michael Niedermayer's avatar
Michael Niedermayer committed
732 733 734 735
    }
    return out_count;
}

736 737 738 739
int swr_is_initialized(struct SwrContext *s) {
    return !!s->in_buffer.ch_count;
}

740 741 742 743 744
int swr_convert(struct SwrContext *s, uint8_t *out_arg[SWR_CH_MAX], int out_count,
                                const uint8_t *in_arg [SWR_CH_MAX], int  in_count){
    AudioData * in= &s->in;
    AudioData *out= &s->out;

745 746 747 748 749
    if (!swr_is_initialized(s)) {
        av_log(s, AV_LOG_ERROR, "Context has not been initialized\n");
        return AVERROR(EINVAL);
    }

750
    while(s->drop_output > 0){
751 752
        int ret;
        uint8_t *tmp_arg[SWR_CH_MAX];
753 754
#define MAX_DROP_STEP 16384
        if((ret=swri_realloc_audio(&s->drop_temp, FFMIN(s->drop_output, MAX_DROP_STEP)))<0)
755 756
            return ret;

757
        reversefill_audiodata(&s->drop_temp, tmp_arg);
758
        s->drop_output *= -1; //FIXME find a less hackish solution
759
        ret = swr_convert(s, tmp_arg, FFMIN(-s->drop_output, MAX_DROP_STEP), in_arg, in_count); //FIXME optimize but this is as good as never called so maybe it doesn't matter
760
        s->drop_output *= -1;
761 762
        in_count = 0;
        if(ret>0) {
763
            s->drop_output -= ret;
764 765
            continue;
        }
766 767 768 769 770

        if(s->drop_output || !out_arg)
            return 0;
    }

771
    if(!in_arg){
772 773 774 775 776 777
        if(s->resample){
            if (!s->flushed)
                s->resampler->flush(s);
            s->resample_in_constraint = 0;
            s->flushed = 1;
        }else if(!s->in_buffer_count){
778 779 780 781 782 783 784 785
            return 0;
        }
    }else
        fill_audiodata(in ,  (void*)in_arg);

    fill_audiodata(out, out_arg);

    if(s->resample){
786 787 788 789
        int ret = swr_convert_internal(s, out, out_count, in, in_count);
        if(ret>0 && !s->drop_output)
            s->outpts += ret * (int64_t)s->in_sample_rate;
        return ret;
790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818
    }else{
        AudioData tmp= *in;
        int ret2=0;
        int ret, size;
        size = FFMIN(out_count, s->in_buffer_count);
        if(size){
            buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
            ret= swr_convert_internal(s, out, size, &tmp, size);
            if(ret<0)
                return ret;
            ret2= ret;
            s->in_buffer_count -= ret;
            s->in_buffer_index += ret;
            buf_set(out, out, ret);
            out_count -= ret;
            if(!s->in_buffer_count)
                s->in_buffer_index = 0;
        }

        if(in_count){
            size= s->in_buffer_index + s->in_buffer_count + in_count - out_count;

            if(in_count > out_count) { //FIXME move after swr_convert_internal
                if(   size > s->in_buffer.count
                && s->in_buffer_count + in_count - out_count <= s->in_buffer_index){
                    buf_set(&tmp, &s->in_buffer, s->in_buffer_index);
                    copy(&s->in_buffer, &tmp, s->in_buffer_count);
                    s->in_buffer_index=0;
                }else
819
                    if((ret=swri_realloc_audio(&s->in_buffer, size)) < 0)
820 821 822 823 824 825 826 827 828 829 830 831 832
                        return ret;
            }

            if(out_count){
                size = FFMIN(in_count, out_count);
                ret= swr_convert_internal(s, out, size, in, size);
                if(ret<0)
                    return ret;
                buf_set(in, in, ret);
                in_count -= ret;
                ret2 += ret;
            }
            if(in_count){
833
                buf_set(&tmp, &s->in_buffer, s->in_buffer_index + s->in_buffer_count);
834 835 836 837
                copy(&tmp, in, in_count);
                s->in_buffer_count += in_count;
            }
        }
838 839
        if(ret2>0 && !s->drop_output)
            s->outpts += ret2 * (int64_t)s->in_sample_rate;
840 841 842 843
        return ret2;
    }
}

844 845 846 847 848 849 850 851 852 853
int swr_drop_output(struct SwrContext *s, int count){
    s->drop_output += count;

    if(s->drop_output <= 0)
        return 0;

    av_log(s, AV_LOG_VERBOSE, "discarding %d audio samples\n", count);
    return swr_convert(s, NULL, s->drop_output, NULL, 0);
}

854 855 856 857 858 859 860
int swr_inject_silence(struct SwrContext *s, int count){
    int ret, i;
    uint8_t *tmp_arg[SWR_CH_MAX];

    if(count <= 0)
        return 0;

861 862 863 864 865 866 867
#define MAX_SILENCE_STEP 16384
    while (count > MAX_SILENCE_STEP) {
        if ((ret = swr_inject_silence(s, MAX_SILENCE_STEP)) < 0)
            return ret;
        count -= MAX_SILENCE_STEP;
    }

868
    if((ret=swri_realloc_audio(&s->silence, count))<0)
869 870
        return ret;

871 872
    if(s->silence.planar) for(i=0; i<s->silence.ch_count; i++) {
        memset(s->silence.ch[i], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps);
873
    } else
874
        memset(s->silence.ch[0], s->silence.bps==1 ? 0x80 : 0, count*s->silence.bps*s->silence.ch_count);
875

876
    reversefill_audiodata(&s->silence, tmp_arg);
877 878 879 880
    av_log(s, AV_LOG_VERBOSE, "adding %d audio samples of silence\n", count);
    ret = swr_convert(s, NULL, 0, (const uint8_t**)tmp_arg, count);
    return ret;
}
881

882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909
int64_t swr_get_delay(struct SwrContext *s, int64_t base){
    if (s->resampler && s->resample){
        return s->resampler->get_delay(s, base);
    }else{
        return (s->in_buffer_count*base + (s->in_sample_rate>>1))/ s->in_sample_rate;
    }
}

int swr_set_compensation(struct SwrContext *s, int sample_delta, int compensation_distance){
    int ret;

    if (!s || compensation_distance < 0)
        return AVERROR(EINVAL);
    if (!compensation_distance && sample_delta)
        return AVERROR(EINVAL);
    if (!s->resample) {
        s->flags |= SWR_FLAG_RESAMPLE;
        ret = swr_init(s);
        if (ret < 0)
            return ret;
    }
    if (!s->resampler->set_compensation){
        return AVERROR(EINVAL);
    }else{
        return s->resampler->set_compensation(s->resample, sample_delta, compensation_distance);
    }
}

910 911 912
int64_t swr_next_pts(struct SwrContext *s, int64_t pts){
    if(pts == INT64_MIN)
        return s->outpts;
913 914 915 916

    if (s->firstpts == AV_NOPTS_VALUE)
        s->outpts = s->firstpts = pts;

917 918 919
    if(s->min_compensation >= FLT_MAX) {
        return (s->outpts = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate));
    } else {
920
        int64_t delta = pts - swr_get_delay(s, s->in_sample_rate * (int64_t)s->out_sample_rate) - s->outpts + s->drop_output*(int64_t)s->in_sample_rate;
921 922 923
        double fdelta = delta /(double)(s->in_sample_rate * (int64_t)s->out_sample_rate);

        if(fabs(fdelta) > s->min_compensation) {
924
            if(s->outpts == s->firstpts || fabs(fdelta) > s->min_hard_compensation){
925 926 927 928 929 930
                int ret;
                if(delta > 0) ret = swr_inject_silence(s,  delta / s->out_sample_rate);
                else          ret = swr_drop_output   (s, -delta / s-> in_sample_rate);
                if(ret<0){
                    av_log(s, AV_LOG_ERROR, "Failed to compensate for timestamp delta of %f\n", fdelta);
                }
931
            } else if(s->soft_compensation_duration && s->max_soft_compensation) {
932
                int duration = s->out_sample_rate * s->soft_compensation_duration;
933 934
                double max_soft_compensation = s->max_soft_compensation / (s->max_soft_compensation < 0 ? -s->in_sample_rate : 1);
                int comp = av_clipf(fdelta, -max_soft_compensation, max_soft_compensation) * duration ;
935 936 937 938 939 940 941 942
                av_log(s, AV_LOG_VERBOSE, "compensating audio timestamp drift:%f compensation:%d in:%d\n", fdelta, comp, duration);
                swr_set_compensation(s, comp, duration);
            }
        }

        return s->outpts;
    }
}