swr: move format convert code to dedicated functions.

This should easier common case optimizations.

libswresample/audioconvert.c

@@ -32,23 +32,97 @@
 #include "audioconvert.h"
 
 
+typedef void (conv_func_type)(uint8_t *po, const uint8_t *pi, int is, int os, uint8_t *end);
+
 struct AVAudioConvert {
     int channels;
-    int fmt_pair;
+    conv_func_type *conv_f;
     const int *ch_map;
 };
 
+#define CONV_FUNC_NAME(dst_fmt, src_fmt) conv_ ## src_fmt ## _to_ ## dst_fmt
+
+//FIXME rounding ?
+#define CONV_FUNC(ofmt, otype, ifmt, expr)\
+static void CONV_FUNC_NAME(ofmt, ifmt)(uint8_t *po, const uint8_t *pi, int is, int os, uint8_t *end)\
+{\
+    do{\
+        *(otype*)po = expr; pi += is; po += os;\
+    }while(po < end);\
+}
+
+//FIXME put things below under ifdefs so we do not waste space for cases no codec will need
+CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_U8 ,  *(const uint8_t*)pi)
+CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)<<8)
+CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)<<24)
+CONV_FUNC(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
+CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
+CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S16, (*(const int16_t*)pi>>8) + 0x80)
+CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S16,  *(const int16_t*)pi)
+CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S16,  *(const int16_t*)pi<<16)
+CONV_FUNC(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_S16,  *(const int16_t*)pi*(1.0 / (1<<15)))
+CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S16,  *(const int16_t*)pi*(1.0 / (1<<15)))
+CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S32, (*(const int32_t*)pi>>24) + 0x80)
+CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S32,  *(const int32_t*)pi>>16)
+CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S32,  *(const int32_t*)pi)
+CONV_FUNC(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_S32,  *(const int32_t*)pi*(1.0 / (1U<<31)))
+CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S32,  *(const int32_t*)pi*(1.0 / (1U<<31)))
+CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_FLT, av_clip_uint8(  lrintf(*(const float*)pi * (1<<7)) + 0x80))
+CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, av_clip_int16(  lrintf(*(const float*)pi * (1<<15))))
+CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, av_clipl_int32(llrintf(*(const float*)pi * (1U<<31))))
+CONV_FUNC(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_FLT, *(const float*)pi)
+CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_FLT, *(const float*)pi)
+CONV_FUNC(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_DBL, av_clip_uint8(  lrint(*(const double*)pi * (1<<7)) + 0x80))
+CONV_FUNC(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, av_clip_int16(  lrint(*(const double*)pi * (1<<15))))
+CONV_FUNC(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, av_clipl_int32(llrint(*(const double*)pi * (1U<<31))))
+CONV_FUNC(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_DBL, *(const double*)pi)
+CONV_FUNC(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_DBL, *(const double*)pi)
+
+#define FMT_PAIR_FUNC(out, in) [out + AV_SAMPLE_FMT_NB*in] = CONV_FUNC_NAME(out, in)
+
+conv_func_type *fmt_pair_to_conv_functions[AV_SAMPLE_FMT_NB*AV_SAMPLE_FMT_NB] = {
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_U8 , AV_SAMPLE_FMT_U8 ),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_U8 ),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_U8 ),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_U8 ),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_U8 ),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_U8 , AV_SAMPLE_FMT_S16),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S16),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_S16),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_S16),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_U8 , AV_SAMPLE_FMT_S32),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_S32),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_S32),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S32),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_S32),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_U8 , AV_SAMPLE_FMT_FLT),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_FLT),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_FLT),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_FLT),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_FLT),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_U8 , AV_SAMPLE_FMT_DBL),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_DBL),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_S32, AV_SAMPLE_FMT_DBL),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_DBL),
+    FMT_PAIR_FUNC(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBL),
+};
+
 AVAudioConvert *swr_audio_convert_alloc(enum AVSampleFormat out_fmt,
                                         enum AVSampleFormat in_fmt,
                                         int channels, const int *ch_map,
                                         int flags)
 {
     AVAudioConvert *ctx;
+    conv_func_type *f = fmt_pair_to_conv_functions[out_fmt + AV_SAMPLE_FMT_NB*in_fmt];
+
+    if (!f)
+        return NULL;
     ctx = av_malloc(sizeof(AVAudioConvert));
     if (!ctx)
         return NULL;
     ctx->channels = channels;
-    ctx->fmt_pair = out_fmt + AV_SAMPLE_FMT_NB*in_fmt;
+    ctx->conv_f   = f;
     ctx->ch_map   = ch_map;
     return ctx;
 }
@@ -76,43 +150,7 @@ int swr_audio_convert(AVAudioConvert *ctx, AudioData *out, AudioData*in, int len
         uint8_t *end= po + os*len;
         if(!po)
             continue;
-
-#define CONV(ofmt, otype, ifmt, expr)\
-if(ctx->fmt_pair == ofmt + AV_SAMPLE_FMT_NB*ifmt){\
-    do{\
-        *(otype*)po = expr; pi += is; po += os;\
-    }while(po < end);\
-}
-
-//FIXME put things below under ifdefs so we do not waste space for cases no codec will need
-//FIXME rounding ?
-
-             CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_U8 ,  *(const uint8_t*)pi)
-        else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)<<8)
-        else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)<<24)
-        else CONV(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
-        else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_U8 , (*(const uint8_t*)pi - 0x80)*(1.0 / (1<<7)))
-        else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S16, (*(const int16_t*)pi>>8) + 0x80)
-        else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S16,  *(const int16_t*)pi)
-        else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S16,  *(const int16_t*)pi<<16)
-        else CONV(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_S16,  *(const int16_t*)pi*(1.0 / (1<<15)))
-        else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S16,  *(const int16_t*)pi*(1.0 / (1<<15)))
-        else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_S32, (*(const int32_t*)pi>>24) + 0x80)
-        else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_S32,  *(const int32_t*)pi>>16)
-        else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_S32,  *(const int32_t*)pi)
-        else CONV(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_S32,  *(const int32_t*)pi*(1.0 / (1U<<31)))
-        else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_S32,  *(const int32_t*)pi*(1.0 / (1U<<31)))
-        else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_FLT, av_clip_uint8(  lrintf(*(const float*)pi * (1<<7)) + 0x80))
-        else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_FLT, av_clip_int16(  lrintf(*(const float*)pi * (1<<15))))
-        else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_FLT, av_clipl_int32(llrintf(*(const float*)pi * (1U<<31))))
-        else CONV(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_FLT, *(const float*)pi)
-        else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_FLT, *(const float*)pi)
-        else CONV(AV_SAMPLE_FMT_U8 , uint8_t, AV_SAMPLE_FMT_DBL, av_clip_uint8(  lrint(*(const double*)pi * (1<<7)) + 0x80))
-        else CONV(AV_SAMPLE_FMT_S16, int16_t, AV_SAMPLE_FMT_DBL, av_clip_int16(  lrint(*(const double*)pi * (1<<15))))
-        else CONV(AV_SAMPLE_FMT_S32, int32_t, AV_SAMPLE_FMT_DBL, av_clipl_int32(llrint(*(const double*)pi * (1U<<31))))
-        else CONV(AV_SAMPLE_FMT_FLT, float  , AV_SAMPLE_FMT_DBL, *(const double*)pi)
-        else CONV(AV_SAMPLE_FMT_DBL, double , AV_SAMPLE_FMT_DBL, *(const double*)pi)
-        else return -1;
+        ctx->conv_f(po, pi, is, os, end);
     }
     return 0;
 }