]> git.sesse.net Git - ffmpeg/blobdiff - libswresample/swresample.c
swr-test: support fabrices audiogen like avr
[ffmpeg] / libswresample / swresample.c
index 205e4360b0ff6ebd3fa3edc87488893874153e5f..414d84cafb2cdd8099d504e8d946b606fba7e105 100644 (file)
 
 //TODO split options array out?
 #define OFFSET(x) offsetof(SwrContext,x)
-static const AVOption options[]={
-{"ich",  "input channel count", OFFSET( in.ch_count   ), AV_OPT_TYPE_INT, {.dbl=2}, 0, SWR_CH_MAX, 0},
-{"och", "output channel count", OFFSET(out.ch_count   ), AV_OPT_TYPE_INT, {.dbl=2}, 0, SWR_CH_MAX, 0},
-{"uch",   "used channel count", OFFSET(used_ch_count  ), AV_OPT_TYPE_INT, {.dbl=0}, 0, SWR_CH_MAX, 0},
-{"isr",  "input sample rate"  , OFFSET( in_sample_rate), AV_OPT_TYPE_INT, {.dbl=48000}, 1, INT_MAX, 0},
-{"osr", "output sample rate"  , OFFSET(out_sample_rate), AV_OPT_TYPE_INT, {.dbl=48000}, 1, INT_MAX, 0},
-//{"ip" ,  "input planar"       , OFFSET( in.planar     ), AV_OPT_TYPE_INT, {.dbl=0},    0,       1, 0},
-//{"op" , "output planar"       , OFFSET(out.planar     ), AV_OPT_TYPE_INT, {.dbl=0},    0,       1, 0},
-{"isf",  "input sample format", OFFSET( in_sample_fmt ), AV_OPT_TYPE_INT, {.dbl=AV_SAMPLE_FMT_S16}, 0, AV_SAMPLE_FMT_NB-1+256, 0},
-{"osf", "output sample format", OFFSET(out_sample_fmt ), AV_OPT_TYPE_INT, {.dbl=AV_SAMPLE_FMT_S16}, 0, AV_SAMPLE_FMT_NB-1+256, 0},
-{"tsf", "internal sample format", OFFSET(int_sample_fmt ), AV_OPT_TYPE_INT, {.dbl=AV_SAMPLE_FMT_NONE}, -1, AV_SAMPLE_FMT_FLT, 0},
-{"icl",  "input channel layout" , OFFSET( in_ch_layout), AV_OPT_TYPE_INT64, {.dbl=0}, 0, INT64_MAX, 0, "channel_layout"},
-{"ocl",  "output channel layout", OFFSET(out_ch_layout), AV_OPT_TYPE_INT64, {.dbl=0}, 0, INT64_MAX, 0, "channel_layout"},
-{"clev", "center mix level"     , OFFSET(clev)         , AV_OPT_TYPE_FLOAT, {.dbl=C_30DB}, 0, 4, 0},
-{"slev", "sourround mix level"  , OFFSET(slev)         , AV_OPT_TYPE_FLOAT, {.dbl=C_30DB}, 0, 4, 0},
-{"rmvol", "rematrix volume"     , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0}, -1000, 1000, 0},
-{"flags", NULL                  , OFFSET(flags)        , AV_OPT_TYPE_FLAGS, {.dbl=0}, 0,  UINT_MAX, 0, "flags"},
-{"res", "force resampling", 0, AV_OPT_TYPE_CONST, {.dbl=SWR_FLAG_RESAMPLE}, INT_MIN, INT_MAX, 0, "flags"},
+#define PARAM AV_OPT_FLAG_AUDIO_PARAM
 
+static const AVOption options[]={
+{"ich"                  ,  "Input Channel Count"        , OFFSET( in.ch_count   ), AV_OPT_TYPE_INT  , {.dbl=2                     }, 0      , SWR_CH_MAX, PARAM},
+{"in_channel_count"     ,  "Input Channel Count"        , OFFSET( in.ch_count   ), AV_OPT_TYPE_INT  , {.dbl=2                     }, 0      , SWR_CH_MAX, PARAM},
+{"och"                  , "Output Channel Count"        , OFFSET(out.ch_count   ), AV_OPT_TYPE_INT  , {.dbl=2                     }, 0      , SWR_CH_MAX, PARAM},
+{"out_channel_count"    , "Output Channel Count"        , OFFSET(out.ch_count   ), AV_OPT_TYPE_INT  , {.dbl=2                     }, 0      , SWR_CH_MAX, PARAM},
+{"uch"                  ,   "Used Channel Count"        , OFFSET(used_ch_count  ), AV_OPT_TYPE_INT  , {.dbl=0                     }, 0      , SWR_CH_MAX, PARAM},
+{"used_channel_count"   ,   "Used Channel Count"        , OFFSET(used_ch_count  ), AV_OPT_TYPE_INT  , {.dbl=0                     }, 0      , SWR_CH_MAX, PARAM},
+{"isr"                  ,  "Input Sample Rate"          , OFFSET( in_sample_rate), AV_OPT_TYPE_INT  , {.dbl=48000                 }, 1      , INT_MAX   , PARAM},
+{"in_sample_rate"       ,  "Input Sample Rate"          , OFFSET( in_sample_rate), AV_OPT_TYPE_INT  , {.dbl=48000                 }, 1      , INT_MAX   , PARAM},
+{"osr"                  , "Output Sample Rate"          , OFFSET(out_sample_rate), AV_OPT_TYPE_INT  , {.dbl=48000                 }, 1      , INT_MAX   , PARAM},
+{"out_sample_rate"      , "Output Sample Rate"          , OFFSET(out_sample_rate), AV_OPT_TYPE_INT  , {.dbl=48000                 }, 1      , INT_MAX   , PARAM},
+{"isf"                  ,    "Input Sample Format"      , OFFSET( in_sample_fmt ), AV_OPT_TYPE_INT  , {.dbl=AV_SAMPLE_FMT_S16     }, 0      , AV_SAMPLE_FMT_NB-1+256, PARAM},
+{"in_sample_fmt"        ,    "Input Sample Format"      , OFFSET( in_sample_fmt ), AV_OPT_TYPE_INT  , {.dbl=AV_SAMPLE_FMT_S16     }, 0      , AV_SAMPLE_FMT_NB-1+256, PARAM},
+{"osf"                  ,   "Output Sample Format"      , OFFSET(out_sample_fmt ), AV_OPT_TYPE_INT  , {.dbl=AV_SAMPLE_FMT_S16     }, 0      , AV_SAMPLE_FMT_NB-1+256, PARAM},
+{"out_sample_fmt"       ,   "Output Sample Format"      , OFFSET(out_sample_fmt ), AV_OPT_TYPE_INT  , {.dbl=AV_SAMPLE_FMT_S16     }, 0      , AV_SAMPLE_FMT_NB-1+256, PARAM},
+{"tsf"                  , "Internal Sample Format"      , OFFSET(int_sample_fmt ), AV_OPT_TYPE_INT  , {.dbl=AV_SAMPLE_FMT_NONE    }, -1     , AV_SAMPLE_FMT_FLT, PARAM},
+{"internal_sample_fmt"  , "Internal Sample Format"      , OFFSET(int_sample_fmt ), AV_OPT_TYPE_INT  , {.dbl=AV_SAMPLE_FMT_NONE    }, -1     , AV_SAMPLE_FMT_FLT, PARAM},
+{"icl"                  ,   "Input Channel Layout"      , OFFSET( in_ch_layout  ), AV_OPT_TYPE_INT64, {.dbl=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
+{"in_channel_layout"    ,   "Input Channel Layout"      , OFFSET( in_ch_layout  ), AV_OPT_TYPE_INT64, {.dbl=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
+{"ocl"                  ,  "Output Channel Layout"      , OFFSET(out_ch_layout  ), AV_OPT_TYPE_INT64, {.dbl=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
+{"out_channel_layout"   ,  "Output Channel Layout"      , OFFSET(out_ch_layout  ), AV_OPT_TYPE_INT64, {.dbl=0                     }, 0      , INT64_MAX , PARAM, "channel_layout"},
+{"clev"                 ,    "Center Mix Level"         , OFFSET(clev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
+{"center_mix_level"     ,    "Center Mix Level"         , OFFSET(clev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
+{"slev"                 , "Sourround Mix Level"         , OFFSET(slev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
+{"surround_mix_level"   , "Sourround Mix Level"         , OFFSET(slev           ), AV_OPT_TYPE_FLOAT, {.dbl=C_30DB                }, -32    , 32        , PARAM},
+{"rmvol"                , "Rematrix Volume"             , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0                   }, -1000  , 1000      , PARAM},
+{"rematrix_volume"      , "Rematrix Volume"             , OFFSET(rematrix_volume), AV_OPT_TYPE_FLOAT, {.dbl=1.0                   }, -1000  , 1000      , PARAM},
+{"flags"                , NULL                          , OFFSET(flags          ), AV_OPT_TYPE_FLAGS, {.dbl=0                     }, 0      , UINT_MAX  , PARAM, "flags"},
+{"swr_flags"            , NULL                          , OFFSET(flags          ), AV_OPT_TYPE_FLAGS, {.dbl=0                     }, 0      , UINT_MAX  , PARAM, "flags"},
+{"res"                  , "Force Resampling"            , 0                      , AV_OPT_TYPE_CONST, {.dbl=SWR_FLAG_RESAMPLE     }, INT_MIN, INT_MAX   , PARAM, "flags"},
+{"dither_scale"         , "Dither Scale"                , OFFSET(dither_scale   ), AV_OPT_TYPE_FLOAT, {.dbl=1                     }, 0      , INT_MAX   , PARAM},
+{"dither_method"        , "Dither Method"               , OFFSET(dither_method  ), AV_OPT_TYPE_INT  , {.dbl=0                     }, 0      , SWR_DITHER_NB-1, PARAM, "dither_method"},
+{"rectangular"          , "Rectangular Dither"          , 0                      , AV_OPT_TYPE_CONST, {.dbl=SWR_DITHER_RECTANGULAR}, INT_MIN, INT_MAX   , PARAM, "dither_method"},
+{"triangular"           ,  "Triangular Dither"          , 0                      , AV_OPT_TYPE_CONST, {.dbl=SWR_DITHER_TRIANGULAR }, INT_MIN, INT_MAX   , PARAM, "dither_method"},
+{"triangular_hp"        , "Triangular Dither With High Pass" , 0                 , AV_OPT_TYPE_CONST, {.dbl=SWR_DITHER_TRIANGULAR_HIGHPASS }, INT_MIN, INT_MAX, PARAM, "dither_method"},
+{"filter_size"          , "Resampling Filter Size"      , OFFSET(filter_size)    , AV_OPT_TYPE_INT  , {.dbl=16                    }, 0      , INT_MAX   , PARAM },
+{"phase_shift"          , "Resampling Phase Shift"      , OFFSET(phase_shift)    , AV_OPT_TYPE_INT  , {.dbl=10                    }, 0      , 30        , PARAM },
+{"linear_interp"        , "Use Linear Interpolation"    , OFFSET(linear_interp)  , AV_OPT_TYPE_INT  , {.dbl=0                     }, 0      , 1         , PARAM },
+{"cutoff"               , "Cutoff Frequency Ratio"      , OFFSET(cutoff)         , AV_OPT_TYPE_DOUBLE,{.dbl=0.8                   }, 0      , 1         , PARAM },
 {0}
 };
 
@@ -94,6 +116,11 @@ int swr_set_channel_mapping(struct SwrContext *s, const int *channel_map){
     return 0;
 }
 
+const AVClass *swr_get_class(void)
+{
+    return &av_class;
+}
+
 struct SwrContext *swr_alloc(void){
     SwrContext *s= av_mallocz(sizeof(SwrContext));
     if(s){
@@ -119,7 +146,7 @@ struct SwrContext *swr_alloc_set_opts(struct SwrContext *s,
     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);
-    av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_S16, 0);
+    av_opt_set_int(s, "tsf", AV_SAMPLE_FMT_NONE,   0);
     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);
     av_opt_set_int(s, "uch", 0, 0);
@@ -139,6 +166,7 @@ void swr_free(SwrContext **ss){
         free_temp(&s->midbuf);
         free_temp(&s->preout);
         free_temp(&s->in_buffer);
+        free_temp(&s->dither);
         swri_audio_convert_free(&s-> in_convert);
         swri_audio_convert_free(&s->out_convert);
         swri_audio_convert_free(&s->full_convert);
@@ -156,6 +184,7 @@ int swr_init(struct SwrContext *s){
     free_temp(&s->midbuf);
     free_temp(&s->preout);
     free_temp(&s->in_buffer);
+    free_temp(&s->dither);
     swri_audio_convert_free(&s-> in_convert);
     swri_audio_convert_free(&s->out_convert);
     swri_audio_convert_free(&s->full_convert);
@@ -176,25 +205,28 @@ int swr_init(struct SwrContext *s){
         return AVERROR(EINVAL);
     }
 
-    if(   s->int_sample_fmt != AV_SAMPLE_FMT_S16
-        &&s->int_sample_fmt != AV_SAMPLE_FMT_FLT){
-        av_log(s, AV_LOG_ERROR, "Requested sample format %s is not supported internally, only float & S16 is supported\n", av_get_sample_fmt_name(s->int_sample_fmt));
-        return AVERROR(EINVAL);
-    }
-
     //FIXME should we allow/support using FLT on material that doesnt need it ?
     if(s->in_sample_fmt <= AV_SAMPLE_FMT_S16 || s->int_sample_fmt==AV_SAMPLE_FMT_S16){
         s->int_sample_fmt= AV_SAMPLE_FMT_S16;
     }else
         s->int_sample_fmt= AV_SAMPLE_FMT_FLT;
 
+    if(   s->int_sample_fmt != AV_SAMPLE_FMT_S16
+        &&s->int_sample_fmt != AV_SAMPLE_FMT_S32
+        &&s->int_sample_fmt != AV_SAMPLE_FMT_FLT){
+        av_log(s, AV_LOG_ERROR, "Requested sample format %s is not supported internally, S16/S32/FLT is supported\n", av_get_sample_fmt_name(s->int_sample_fmt));
+        return AVERROR(EINVAL);
+    }
 
     if (s->out_sample_rate!=s->in_sample_rate || (s->flags & SWR_FLAG_RESAMPLE)){
-        s->resample = swri_resample_init(s->resample, s->out_sample_rate, s->in_sample_rate, 16, 10, 0, 0.8);
+        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);
     }else
         swri_resample_free(&s->resample);
-    if(s->int_sample_fmt != AV_SAMPLE_FMT_S16 && s->resample){
-        av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16 currently\n"); //FIXME
+    if(    s->int_sample_fmt != AV_SAMPLE_FMT_S16
+        && s->int_sample_fmt != AV_SAMPLE_FMT_S32
+        && s->int_sample_fmt != AV_SAMPLE_FMT_FLT
+        && s->resample){
+        av_log(s, AV_LOG_ERROR, "Resampling only supported with internal s16/s32/flt\n");
         return -1;
     }
 
@@ -278,6 +310,8 @@ av_assert0(s->out.ch_count);
         s->in_buffer.planar = 1;
     }
 
+    s->dither = s->preout;
+
     if(s->rematrix)
         return swri_rematrix_init(s);
 
@@ -342,12 +376,14 @@ static void fill_audiodata(AudioData *out, uint8_t *in_arg [SWR_CH_MAX]){
  * out may be equal in.
  */
 static void buf_set(AudioData *out, AudioData *in, int count){
+    int ch;
     if(in->planar){
-        int ch;
         for(ch=0; ch<out->ch_count; ch++)
             out->ch[ch]= in->ch[ch] + count*out->bps;
-    }else
-        out->ch[0]= in->ch[0] + count*out->ch_count*out->bps;
+    }else{
+        for(ch=0; ch<out->ch_count; ch++)
+            out->ch[ch]= in->ch[0] + (ch + count*out->ch_count) * out->bps;
+    }
 }
 
 /**
@@ -502,6 +538,25 @@ static int swr_convert_internal(struct SwrContext *s, AudioData *out, int out_co
     }
 
     if(preout != out && out_count){
+        if(s->dither_method){
+            int ch;
+            int dither_count= FFMAX(out_count, 1<<16);
+            av_assert0(preout != in);
+
+            if((ret=realloc_audio(&s->dither, dither_count))<0)
+                return ret;
+            if(ret)
+                for(ch=0; ch<s->dither.ch_count; ch++)
+                    swri_get_dither(s, s->dither.ch[ch], s->dither.count, 12345678913579<<ch, s->out_sample_fmt, s->int_sample_fmt);
+            av_assert0(s->dither.ch_count == preout->ch_count);
+
+            if(s->dither_pos + out_count > s->dither.count)
+                s->dither_pos = 0;
+            for(ch=0; ch<preout->ch_count; ch++)
+                swri_sum2(s->int_sample_fmt, preout->ch[ch], preout->ch[ch], s->dither.ch[ch] + s->dither.bps * s->dither_pos, 1, 1, out_count);
+
+            s->dither_pos += out_count;
+        }
 //FIXME packed doesnt need more than 1 chan here!
         swri_audio_convert(s->out_convert, out, preout, out_count);
     }