swscale: re-add support for non-native endianness.
[ffmpeg.git] / libswscale / swscale.c
index f6ed453..e0dce96 100644 (file)
@@ -285,42 +285,13 @@ yuv2NBPS(10, LE, 0);
 yuv2NBPS(16, BE, 1);
 yuv2NBPS(16, LE, 0);
 
-static inline void yuv2yuvX16_c(SwsContext *c, const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
-                                const int16_t *chrFilter, const int16_t **chrUSrc, const int16_t **chrVSrc, int chrFilterSize,
-                                const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest, uint8_t *vDest, uint8_t *aDest, int dstW, int chrDstW,
-                                enum PixelFormat dstFormat)
-{
-#define conv16(bits) \
-    if (isBE(dstFormat)) { \
-        yuv2yuvX ## bits ## BE_c(c, lumFilter, lumSrc, lumFilterSize, \
-                                 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
-                                 alpSrc, \
-                                 dest, uDest, vDest, aDest, \
-                                 dstW, chrDstW); \
-    } else { \
-        yuv2yuvX ## bits ## LE_c(c, lumFilter, lumSrc, lumFilterSize, \
-                                 chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
-                                 alpSrc, \
-                                 dest, uDest, vDest, aDest, \
-                                 dstW, chrDstW); \
-    }
-    if (is16BPS(dstFormat)) {
-        conv16(16);
-    } else if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
-        conv16(9);
-    } else {
-        conv16(10);
-    }
-#undef conv16
-}
-
-static inline void yuv2yuvX_c(SwsContext *c, const int16_t *lumFilter,
-                              const int16_t **lumSrc, int lumFilterSize,
-                              const int16_t *chrFilter, const int16_t **chrUSrc,
-                              const int16_t **chrVSrc,
-                              int chrFilterSize, const int16_t **alpSrc,
-                              uint8_t *dest, uint8_t *uDest, uint8_t *vDest,
-                              uint8_t *aDest, int dstW, int chrDstW)
+static void yuv2yuvX_c(SwsContext *c, const int16_t *lumFilter,
+                       const int16_t **lumSrc, int lumFilterSize,
+                       const int16_t *chrFilter, const int16_t **chrUSrc,
+                       const int16_t **chrVSrc,
+                       int chrFilterSize, const int16_t **alpSrc,
+                       uint8_t *dest, uint8_t *uDest, uint8_t *vDest,
+                       uint8_t *aDest, int dstW, int chrDstW)
 {
     //FIXME Optimize (just quickly written not optimized..)
     int i;
@@ -356,16 +327,45 @@ static inline void yuv2yuvX_c(SwsContext *c, const int16_t *lumFilter,
 
             aDest[i]= av_clip_uint8(val>>19);
         }
+}
+
+static void yuv2yuv1_c(SwsContext *c, const int16_t *lumSrc,
+                       const int16_t *chrUSrc, const int16_t *chrVSrc,
+                       const int16_t *alpSrc,
+                       uint8_t *dest, uint8_t *uDest, uint8_t *vDest,
+                       uint8_t *aDest, int dstW, int chrDstW)
+{
+    int i;
+    for (i=0; i<dstW; i++) {
+        int val= (lumSrc[i]+64)>>7;
+        dest[i]= av_clip_uint8(val);
+    }
+
+    if (uDest)
+        for (i=0; i<chrDstW; i++) {
+            int u=(chrUSrc[i]+64)>>7;
+            int v=(chrVSrc[i]+64)>>7;
+            uDest[i]= av_clip_uint8(u);
+            vDest[i]= av_clip_uint8(v);
+        }
 
+    if (CONFIG_SWSCALE_ALPHA && aDest)
+        for (i=0; i<dstW; i++) {
+            int val= (alpSrc[i]+64)>>7;
+            aDest[i]= av_clip_uint8(val);
+        }
 }
 
-static inline void yuv2nv12X_c(SwsContext *c, const int16_t *lumFilter,
-                               const int16_t **lumSrc, int lumFilterSize,
-                               const int16_t *chrFilter, const int16_t **chrUSrc,
-                               const int16_t **chrVSrc,
-                               int chrFilterSize, uint8_t *dest, uint8_t *uDest,
-                               int dstW, int chrDstW, enum PixelFormat dstFormat)
+static void yuv2nv12X_c(SwsContext *c, const int16_t *lumFilter,
+                        const int16_t **lumSrc, int lumFilterSize,
+                        const int16_t *chrFilter, const int16_t **chrUSrc,
+                        const int16_t **chrVSrc, int chrFilterSize,
+                        const int16_t **alpSrc, uint8_t *dest, uint8_t *uDest,
+                        uint8_t *vDest, uint8_t *aDest,
+                        int dstW, int chrDstW)
 {
+    enum PixelFormat dstFormat = c->dstFormat;
+
     //FIXME Optimize (just quickly written not optimized..)
     int i;
     for (i=0; i<dstW; i++) {
@@ -408,568 +408,914 @@ static inline void yuv2nv12X_c(SwsContext *c, const int16_t *lumFilter,
         }
 }
 
-#define YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha) \
-    for (i=0; i<(dstW>>1); i++) {\
-        int j;\
-        int Y1 = 1<<18;\
-        int Y2 = 1<<18;\
-        int U  = 1<<18;\
-        int V  = 1<<18;\
-        int av_unused A1, A2;\
-        type av_unused *r, *b, *g;\
-        const int i2= 2*i;\
-        \
-        for (j=0; j<lumFilterSize; j++) {\
-            Y1 += lumSrc[j][i2] * lumFilter[j];\
-            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
-        }\
-        for (j=0; j<chrFilterSize; j++) {\
-            U += chrUSrc[j][i] * chrFilter[j];\
-            V += chrVSrc[j][i] * chrFilter[j];\
-        }\
-        Y1>>=19;\
-        Y2>>=19;\
-        U >>=19;\
-        V >>=19;\
-        if (alpha) {\
-            A1 = 1<<18;\
-            A2 = 1<<18;\
-            for (j=0; j<lumFilterSize; j++) {\
-                A1 += alpSrc[j][i2  ] * lumFilter[j];\
-                A2 += alpSrc[j][i2+1] * lumFilter[j];\
-            }\
-            A1>>=19;\
-            A2>>=19;\
+#define output_pixel(pos, val) \
+        if (target == PIX_FMT_GRAY16BE) { \
+            AV_WB16(pos, val); \
+        } else { \
+            AV_WL16(pos, val); \
         }
 
-#define YSCALE_YUV_2_PACKEDX_C(type,alpha) \
-        YSCALE_YUV_2_PACKEDX_NOCLIP_C(type,alpha)\
-        if ((Y1|Y2|U|V)&256) {\
-            if (Y1>255)   Y1=255; \
-            else if (Y1<0)Y1=0;   \
-            if (Y2>255)   Y2=255; \
-            else if (Y2<0)Y2=0;   \
-            if (U>255)    U=255;  \
-            else if (U<0) U=0;    \
-            if (V>255)    V=255;  \
-            else if (V<0) V=0;    \
-        }\
-        if (alpha && ((A1|A2)&256)) {\
-            A1=av_clip_uint8(A1);\
-            A2=av_clip_uint8(A2);\
-        }
+static av_always_inline void
+yuv2gray16_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                        const int16_t **lumSrc, int lumFilterSize,
+                        const int16_t *chrFilter, const int16_t **chrUSrc,
+                        const int16_t **chrVSrc, int chrFilterSize,
+                        const int16_t **alpSrc, uint8_t *dest, int dstW,
+                        int y, enum PixelFormat target)
+{
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int j;
+        int Y1 = 1 << 18;
+        int Y2 = 1 << 18;
+        const int i2 = 2 * i;
 
-#define YSCALE_YUV_2_PACKEDX_FULL_C(rnd,alpha) \
-    for (i=0; i<dstW; i++) {\
-        int j;\
-        int Y = 0;\
-        int U = -128<<19;\
-        int V = -128<<19;\
-        int av_unused A;\
-        int R,G,B;\
-        \
-        for (j=0; j<lumFilterSize; j++) {\
-            Y += lumSrc[j][i     ] * lumFilter[j];\
-        }\
-        for (j=0; j<chrFilterSize; j++) {\
-            U += chrUSrc[j][i] * chrFilter[j];\
-            V += chrVSrc[j][i] * chrFilter[j];\
-        }\
-        Y >>=10;\
-        U >>=10;\
-        V >>=10;\
-        if (alpha) {\
-            A = rnd;\
-            for (j=0; j<lumFilterSize; j++)\
-                A += alpSrc[j][i     ] * lumFilter[j];\
-            A >>=19;\
-            if (A&256)\
-                A = av_clip_uint8(A);\
+        for (j = 0; j < lumFilterSize; j++) {
+            Y1 += lumSrc[j][i2]   * lumFilter[j];
+            Y2 += lumSrc[j][i2+1] * lumFilter[j];
         }
+        Y1 >>= 11;
+        Y2 >>= 11;
+        if ((Y1 | Y2) & 0x10000) {
+            Y1 = av_clip_uint16(Y1);
+            Y2 = av_clip_uint16(Y2);
+        }
+        output_pixel(&dest[2 * i2 + 0], Y1);
+        output_pixel(&dest[2 * i2 + 2], Y2);
+    }
+}
+
+static av_always_inline void
+yuv2gray16_2_c_template(SwsContext *c, const uint16_t *buf0,
+                        const uint16_t *buf1, const uint16_t *ubuf0,
+                        const uint16_t *ubuf1, const uint16_t *vbuf0,
+                        const uint16_t *vbuf1, const uint16_t *abuf0,
+                        const uint16_t *abuf1, uint8_t *dest, int dstW,
+                        int yalpha, int uvalpha, int y,
+                        enum PixelFormat target)
+{
+    int  yalpha1 = 4095 - yalpha; \
+    int i;
 
-#define YSCALE_YUV_2_RGBX_FULL_C(rnd,alpha) \
-    YSCALE_YUV_2_PACKEDX_FULL_C(rnd>>3,alpha)\
-        Y-= c->yuv2rgb_y_offset;\
-        Y*= c->yuv2rgb_y_coeff;\
-        Y+= rnd;\
-        R= Y + V*c->yuv2rgb_v2r_coeff;\
-        G= Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;\
-        B= Y +                          U*c->yuv2rgb_u2b_coeff;\
-        if ((R|G|B)&(0xC0000000)) {\
-            if (R>=(256<<22))   R=(256<<22)-1; \
-            else if (R<0)R=0;   \
-            if (G>=(256<<22))   G=(256<<22)-1; \
-            else if (G<0)G=0;   \
-            if (B>=(256<<22))   B=(256<<22)-1; \
-            else if (B<0)B=0;   \
+    for (i = 0; i < (dstW >> 1); i++) {
+        const int i2 = 2 * i;
+        int Y1 = (buf0[i2  ] * yalpha1 + buf1[i2  ] * yalpha) >> 11;
+        int Y2 = (buf0[i2+1] * yalpha1 + buf1[i2+1] * yalpha) >> 11;
+
+        output_pixel(&dest[2 * i2 + 0], Y1);
+        output_pixel(&dest[2 * i2 + 2], Y2);
+    }
+}
+
+static av_always_inline void
+yuv2gray16_1_c_template(SwsContext *c, const uint16_t *buf0,
+                        const uint16_t *ubuf0, const uint16_t *ubuf1,
+                        const uint16_t *vbuf0, const uint16_t *vbuf1,
+                        const uint16_t *abuf0, uint8_t *dest, int dstW,
+                        int uvalpha, enum PixelFormat dstFormat,
+                        int flags, int y, enum PixelFormat target)
+{
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        const int i2 = 2 * i;
+        int Y1 = buf0[i2  ] << 1;
+        int Y2 = buf0[i2+1] << 1;
+
+        output_pixel(&dest[2 * i2 + 0], Y1);
+        output_pixel(&dest[2 * i2 + 2], Y2);
+    }
+}
+
+#undef output_pixel
+
+#define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
+static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
+                        const int16_t **lumSrc, int lumFilterSize, \
+                        const int16_t *chrFilter, const int16_t **chrUSrc, \
+                        const int16_t **chrVSrc, int chrFilterSize, \
+                        const int16_t **alpSrc, uint8_t *dest, int dstW, \
+                        int y) \
+{ \
+    name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
+                          chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
+                          alpSrc, dest, dstW, y, fmt); \
+} \
+ \
+static void name ## ext ## _2_c(SwsContext *c, const uint16_t *buf0, \
+                        const uint16_t *buf1, const uint16_t *ubuf0, \
+                        const uint16_t *ubuf1, const uint16_t *vbuf0, \
+                        const uint16_t *vbuf1, const uint16_t *abuf0, \
+                        const uint16_t *abuf1, uint8_t *dest, int dstW, \
+                        int yalpha, int uvalpha, int y) \
+{ \
+    name ## base ## _2_c_template(c, buf0, buf1, ubuf0, ubuf1, \
+                          vbuf0, vbuf1, abuf0, abuf1, \
+                          dest, dstW, yalpha, uvalpha, y, fmt); \
+} \
+ \
+static void name ## ext ## _1_c(SwsContext *c, const uint16_t *buf0, \
+                        const uint16_t *ubuf0, const uint16_t *ubuf1, \
+                        const uint16_t *vbuf0, const uint16_t *vbuf1, \
+                        const uint16_t *abuf0, uint8_t *dest, int dstW, \
+                        int uvalpha, enum PixelFormat dstFormat, \
+                        int flags, int y) \
+{ \
+    name ## base ## _1_c_template(c, buf0, ubuf0, ubuf1, vbuf0, \
+                          vbuf1, abuf0, dest, dstW, uvalpha, \
+                          dstFormat, flags, y, fmt); \
+}
+
+YUV2PACKEDWRAPPER(yuv2gray16,, LE, PIX_FMT_GRAY16LE);
+YUV2PACKEDWRAPPER(yuv2gray16,, BE, PIX_FMT_GRAY16BE);
+
+#define output_pixel(pos, acc) \
+    if (target == PIX_FMT_MONOBLACK) { \
+        pos = acc; \
+    } else { \
+        pos = ~acc; \
+    }
+
+static av_always_inline void
+yuv2mono_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                      const int16_t **lumSrc, int lumFilterSize,
+                      const int16_t *chrFilter, const int16_t **chrUSrc,
+                      const int16_t **chrVSrc, int chrFilterSize,
+                      const int16_t **alpSrc, uint8_t *dest, int dstW,
+                      int y, enum PixelFormat target)
+{
+    const uint8_t * const d128=dither_8x8_220[y&7];
+    uint8_t *g = c->table_gU[128] + c->table_gV[128];
+    int i;
+    int acc = 0;
+
+    for (i = 0; i < dstW - 1; i += 2) {
+        int j;
+        int Y1 = 1 << 18;
+        int Y2 = 1 << 18;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y1 += lumSrc[j][i]   * lumFilter[j];
+            Y2 += lumSrc[j][i+1] * lumFilter[j];
+        }
+        Y1 >>= 19;
+        Y2 >>= 19;
+        if ((Y1 | Y2) & 0x100) {
+            Y1 = av_clip_uint8(Y1);
+            Y2 = av_clip_uint8(Y2);
         }
+        acc += acc + g[Y1 + d128[(i + 0) & 7]];
+        acc += acc + g[Y2 + d128[(i + 1) & 7]];
+        if ((i & 7) == 6) {
+            output_pixel(*dest++, acc);
+        }
+    }
+}
+
+static av_always_inline void
+yuv2mono_2_c_template(SwsContext *c, const uint16_t *buf0,
+                      const uint16_t *buf1, const uint16_t *ubuf0,
+                      const uint16_t *ubuf1, const uint16_t *vbuf0,
+                      const uint16_t *vbuf1, const uint16_t *abuf0,
+                      const uint16_t *abuf1, uint8_t *dest, int dstW,
+                      int yalpha, int uvalpha, int y,
+                      enum PixelFormat target)
+{
+    const uint8_t * const d128 = dither_8x8_220[y & 7];
+    uint8_t *g = c->table_gU[128] + c->table_gV[128];
+    int  yalpha1 = 4095 - yalpha;
+    int i;
 
-#define YSCALE_YUV_2_GRAY16_C \
-    for (i=0; i<(dstW>>1); i++) {\
-        int j;\
-        int Y1 = 1<<18;\
-        int Y2 = 1<<18;\
-        int U  = 1<<18;\
-        int V  = 1<<18;\
-        \
-        const int i2= 2*i;\
-        \
-        for (j=0; j<lumFilterSize; j++) {\
-            Y1 += lumSrc[j][i2] * lumFilter[j];\
-            Y2 += lumSrc[j][i2+1] * lumFilter[j];\
-        }\
-        Y1>>=11;\
-        Y2>>=11;\
-        if ((Y1|Y2|U|V)&65536) {\
-            if (Y1>65535)   Y1=65535; \
-            else if (Y1<0)Y1=0;   \
-            if (Y2>65535)   Y2=65535; \
-            else if (Y2<0)Y2=0;   \
+    for (i = 0; i < dstW - 7; i += 8) {
+        int acc =    g[((buf0[i    ] * yalpha1 + buf1[i    ] * yalpha) >> 19) + d128[0]];
+        acc += acc + g[((buf0[i + 1] * yalpha1 + buf1[i + 1] * yalpha) >> 19) + d128[1]];
+        acc += acc + g[((buf0[i + 2] * yalpha1 + buf1[i + 2] * yalpha) >> 19) + d128[2]];
+        acc += acc + g[((buf0[i + 3] * yalpha1 + buf1[i + 3] * yalpha) >> 19) + d128[3]];
+        acc += acc + g[((buf0[i + 4] * yalpha1 + buf1[i + 4] * yalpha) >> 19) + d128[4]];
+        acc += acc + g[((buf0[i + 5] * yalpha1 + buf1[i + 5] * yalpha) >> 19) + d128[5]];
+        acc += acc + g[((buf0[i + 6] * yalpha1 + buf1[i + 6] * yalpha) >> 19) + d128[6]];
+        acc += acc + g[((buf0[i + 7] * yalpha1 + buf1[i + 7] * yalpha) >> 19) + d128[7]];
+        output_pixel(*dest++, acc);
+    }
+}
+
+static av_always_inline void
+yuv2mono_1_c_template(SwsContext *c, const uint16_t *buf0,
+                      const uint16_t *ubuf0, const uint16_t *ubuf1,
+                      const uint16_t *vbuf0, const uint16_t *vbuf1,
+                      const uint16_t *abuf0, uint8_t *dest, int dstW,
+                      int uvalpha, enum PixelFormat dstFormat,
+                      int flags, int y, enum PixelFormat target)
+{
+    const uint8_t * const d128 = dither_8x8_220[y & 7];
+    uint8_t *g = c->table_gU[128] + c->table_gV[128];
+    int i;
+
+    for (i = 0; i < dstW - 7; i += 8) {
+        int acc =    g[(buf0[i    ] >> 7) + d128[0]];
+        acc += acc + g[(buf0[i + 1] >> 7) + d128[1]];
+        acc += acc + g[(buf0[i + 2] >> 7) + d128[2]];
+        acc += acc + g[(buf0[i + 3] >> 7) + d128[3]];
+        acc += acc + g[(buf0[i + 4] >> 7) + d128[4]];
+        acc += acc + g[(buf0[i + 5] >> 7) + d128[5]];
+        acc += acc + g[(buf0[i + 6] >> 7) + d128[6]];
+        acc += acc + g[(buf0[i + 7] >> 7) + d128[7]];
+        output_pixel(*dest++, acc);
+    }
+}
+
+#undef output_pixel
+
+YUV2PACKEDWRAPPER(yuv2mono,, white, PIX_FMT_MONOWHITE);
+YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK);
+
+#define output_pixels(pos, Y1, U, Y2, V) \
+    if (target == PIX_FMT_YUYV422) { \
+        dest[pos + 0] = Y1; \
+        dest[pos + 1] = U;  \
+        dest[pos + 2] = Y2; \
+        dest[pos + 3] = V;  \
+    } else { \
+        dest[pos + 0] = U;  \
+        dest[pos + 1] = Y1; \
+        dest[pos + 2] = V;  \
+        dest[pos + 3] = Y2; \
+    }
+
+static av_always_inline void
+yuv2422_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                     const int16_t **lumSrc, int lumFilterSize,
+                     const int16_t *chrFilter, const int16_t **chrUSrc,
+                     const int16_t **chrVSrc, int chrFilterSize,
+                     const int16_t **alpSrc, uint8_t *dest, int dstW,
+                     int y, enum PixelFormat target)
+{
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int j;
+        int Y1 = 1 << 18;
+        int Y2 = 1 << 18;
+        int U  = 1 << 18;
+        int V  = 1 << 18;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y1 += lumSrc[j][i * 2]     * lumFilter[j];
+            Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
+        }
+        for (j = 0; j < chrFilterSize; j++) {
+            U += chrUSrc[j][i] * chrFilter[j];
+            V += chrVSrc[j][i] * chrFilter[j];
         }
+        Y1 >>= 19;
+        Y2 >>= 19;
+        U  >>= 19;
+        V  >>= 19;
+        if ((Y1 | Y2 | U | V) & 0x100) {
+            Y1 = av_clip_uint8(Y1);
+            Y2 = av_clip_uint8(Y2);
+            U  = av_clip_uint8(U);
+            V  = av_clip_uint8(V);
+        }
+        output_pixels(4*i, Y1, U, Y2, V);
+    }
+}
 
-#define YSCALE_YUV_2_RGBX_C(type,alpha) \
-    YSCALE_YUV_2_PACKEDX_C(type,alpha)  /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/\
-    r = (type *)c->table_rV[V];   \
-    g = (type *)(c->table_gU[U] + c->table_gV[V]); \
-    b = (type *)c->table_bU[U];
-
-#define YSCALE_YUV_2_PACKED2_C(type,alpha)   \
-    for (i=0; i<(dstW>>1); i++) { \
-        const int i2= 2*i;       \
-        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>19;           \
-        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>19;           \
-        int U= (ubuf0[i]*uvalpha1+ubuf1[i]*uvalpha)>>19;              \
-        int V= (vbuf0[i]*uvalpha1+vbuf1[i]*uvalpha)>>19;              \
-        type av_unused *r, *b, *g;                                    \
-        int av_unused A1, A2;                                         \
-        if (alpha) {\
-            A1= (abuf0[i2  ]*yalpha1+abuf1[i2  ]*yalpha)>>19;         \
-            A2= (abuf0[i2+1]*yalpha1+abuf1[i2+1]*yalpha)>>19;         \
+static av_always_inline void
+yuv2422_2_c_template(SwsContext *c, const uint16_t *buf0,
+                     const uint16_t *buf1, const uint16_t *ubuf0,
+                     const uint16_t *ubuf1, const uint16_t *vbuf0,
+                     const uint16_t *vbuf1, const uint16_t *abuf0,
+                     const uint16_t *abuf1, uint8_t *dest, int dstW,
+                     int yalpha, int uvalpha, int y,
+                     enum PixelFormat target)
+{
+    int  yalpha1 = 4095 - yalpha;
+    int uvalpha1 = 4095 - uvalpha;
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int Y1 = (buf0[i * 2]     * yalpha1  + buf1[i * 2]     * yalpha)  >> 19;
+        int Y2 = (buf0[i * 2 + 1] * yalpha1  + buf1[i * 2 + 1] * yalpha)  >> 19;
+        int U  = (ubuf0[i]        * uvalpha1 + ubuf1[i]        * uvalpha) >> 19;
+        int V  = (vbuf0[i]        * uvalpha1 + vbuf1[i]        * uvalpha) >> 19;
+
+        output_pixels(i * 4, Y1, U, Y2, V);
+    }
+}
+
+static av_always_inline void
+yuv2422_1_c_template(SwsContext *c, const uint16_t *buf0,
+                     const uint16_t *ubuf0, const uint16_t *ubuf1,
+                     const uint16_t *vbuf0, const uint16_t *vbuf1,
+                     const uint16_t *abuf0, uint8_t *dest, int dstW,
+                     int uvalpha, enum PixelFormat dstFormat,
+                     int flags, int y, enum PixelFormat target)
+{
+    int i;
+
+    if (uvalpha < 2048) {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 = buf0[i * 2]     >> 7;
+            int Y2 = buf0[i * 2 + 1] >> 7;
+            int U  = ubuf1[i]        >> 7;
+            int V  = vbuf1[i]        >> 7;
+
+            output_pixels(i * 4, Y1, U, Y2, V);
         }
+    } else {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 =  buf0[i * 2]          >> 7;
+            int Y2 =  buf0[i * 2 + 1]      >> 7;
+            int U  = (ubuf0[i] + ubuf1[i]) >> 8;
+            int V  = (vbuf0[i] + vbuf1[i]) >> 8;
 
-#define YSCALE_YUV_2_GRAY16_2_C   \
-    for (i=0; i<(dstW>>1); i++) { \
-        const int i2= 2*i;       \
-        int Y1= (buf0[i2  ]*yalpha1+buf1[i2  ]*yalpha)>>11;           \
-        int Y2= (buf0[i2+1]*yalpha1+buf1[i2+1]*yalpha)>>11;
-
-#define YSCALE_YUV_2_RGB2_C(type,alpha) \
-    YSCALE_YUV_2_PACKED2_C(type,alpha)\
-    r = (type *)c->table_rV[V];\
-    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
-    b = (type *)c->table_bU[U];
-
-#define YSCALE_YUV_2_PACKED1_C(type,alpha) \
-    for (i=0; i<(dstW>>1); i++) {\
-        const int i2= 2*i;\
-        int Y1= buf0[i2  ]>>7;\
-        int Y2= buf0[i2+1]>>7;\
-        int U= (ubuf1[i])>>7;\
-        int V= (vbuf1[i])>>7;\
-        type av_unused *r, *b, *g;\
-        int av_unused A1, A2;\
-        if (alpha) {\
-            A1= abuf0[i2  ]>>7;\
-            A2= abuf0[i2+1]>>7;\
+            output_pixels(i * 4, Y1, U, Y2, V);
         }
+    }
+}
+
+#undef output_pixels
+
+YUV2PACKEDWRAPPER(yuv2, 422, yuyv422, PIX_FMT_YUYV422);
+YUV2PACKEDWRAPPER(yuv2, 422, uyvy422, PIX_FMT_UYVY422);
+
+#define r_b ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? r : b)
+#define b_r ((target == PIX_FMT_RGB48LE || target == PIX_FMT_RGB48BE) ? b : r)
 
-#define YSCALE_YUV_2_GRAY16_1_C \
-    for (i=0; i<(dstW>>1); i++) {\
-        const int i2= 2*i;\
-        int Y1= buf0[i2  ]<<1;\
-        int Y2= buf0[i2+1]<<1;
-
-#define YSCALE_YUV_2_RGB1_C(type,alpha) \
-    YSCALE_YUV_2_PACKED1_C(type,alpha)\
-    r = (type *)c->table_rV[V];\
-    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
-    b = (type *)c->table_bU[U];
-
-#define YSCALE_YUV_2_PACKED1B_C(type,alpha) \
-    for (i=0; i<(dstW>>1); i++) {\
-        const int i2= 2*i;\
-        int Y1= buf0[i2  ]>>7;\
-        int Y2= buf0[i2+1]>>7;\
-        int U= (ubuf0[i] + ubuf1[i])>>8;\
-        int V= (vbuf0[i] + vbuf1[i])>>8;\
-        type av_unused *r, *b, *g;\
-        int av_unused A1, A2;\
-        if (alpha) {\
-            A1= abuf0[i2  ]>>7;\
-            A2= abuf0[i2+1]>>7;\
+static av_always_inline void
+yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                       const int16_t **lumSrc, int lumFilterSize,
+                       const int16_t *chrFilter, const int16_t **chrUSrc,
+                       const int16_t **chrVSrc, int chrFilterSize,
+                       const int16_t **alpSrc, uint8_t *dest, int dstW,
+                       int y, enum PixelFormat target)
+{
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int j;
+        int Y1 = 1 << 18;
+        int Y2 = 1 << 18;
+        int U  = 1 << 18;
+        int V  = 1 << 18;
+        const uint8_t *r, *g, *b;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y1 += lumSrc[j][i * 2]     * lumFilter[j];
+            Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
+        }
+        for (j = 0; j < chrFilterSize; j++) {
+            U += chrUSrc[j][i] * chrFilter[j];
+            V += chrVSrc[j][i] * chrFilter[j];
+        }
+        Y1 >>= 19;
+        Y2 >>= 19;
+        U  >>= 19;
+        V  >>= 19;
+        if ((Y1 | Y2 | U | V) & 0x100) {
+            Y1 = av_clip_uint8(Y1);
+            Y2 = av_clip_uint8(Y2);
+            U  = av_clip_uint8(U);
+            V  = av_clip_uint8(V);
         }
 
-#define YSCALE_YUV_2_RGB1B_C(type,alpha) \
-    YSCALE_YUV_2_PACKED1B_C(type,alpha)\
-    r = (type *)c->table_rV[V];\
-    g = (type *)(c->table_gU[U] + c->table_gV[V]);\
-    b = (type *)c->table_bU[U];
-
-#define YSCALE_YUV_2_MONO2_C \
-    const uint8_t * const d128=dither_8x8_220[y&7];\
-    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
-    for (i=0; i<dstW-7; i+=8) {\
-        int acc;\
-        acc =       g[((buf0[i  ]*yalpha1+buf1[i  ]*yalpha)>>19) + d128[0]];\
-        acc+= acc + g[((buf0[i+1]*yalpha1+buf1[i+1]*yalpha)>>19) + d128[1]];\
-        acc+= acc + g[((buf0[i+2]*yalpha1+buf1[i+2]*yalpha)>>19) + d128[2]];\
-        acc+= acc + g[((buf0[i+3]*yalpha1+buf1[i+3]*yalpha)>>19) + d128[3]];\
-        acc+= acc + g[((buf0[i+4]*yalpha1+buf1[i+4]*yalpha)>>19) + d128[4]];\
-        acc+= acc + g[((buf0[i+5]*yalpha1+buf1[i+5]*yalpha)>>19) + d128[5]];\
-        acc+= acc + g[((buf0[i+6]*yalpha1+buf1[i+6]*yalpha)>>19) + d128[6]];\
-        acc+= acc + g[((buf0[i+7]*yalpha1+buf1[i+7]*yalpha)>>19) + d128[7]];\
-        ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
-        dest++;\
-    }
-
-#define YSCALE_YUV_2_MONOX_C \
-    const uint8_t * const d128=dither_8x8_220[y&7];\
-    uint8_t *g= c->table_gU[128] + c->table_gV[128];\
-    int acc=0;\
-    for (i=0; i<dstW-1; i+=2) {\
-        int j;\
-        int Y1=1<<18;\
-        int Y2=1<<18;\
-\
-        for (j=0; j<lumFilterSize; j++) {\
-            Y1 += lumSrc[j][i] * lumFilter[j];\
-            Y2 += lumSrc[j][i+1] * lumFilter[j];\
-        }\
-        Y1>>=19;\
-        Y2>>=19;\
-        if ((Y1|Y2)&256) {\
-            if (Y1>255)   Y1=255;\
-            else if (Y1<0)Y1=0;\
-            if (Y2>255)   Y2=255;\
-            else if (Y2<0)Y2=0;\
-        }\
-        acc+= acc + g[Y1+d128[(i+0)&7]];\
-        acc+= acc + g[Y2+d128[(i+1)&7]];\
-        if ((i&7)==6) {\
-            ((uint8_t*)dest)[0]= c->dstFormat == PIX_FMT_MONOBLACK ? acc : ~acc;\
-            dest++;\
-        }\
-    }
-
-#define YSCALE_YUV_2_ANYRGB_C(func, func2, func_g16, func_monoblack)\
-    switch(c->dstFormat) {\
-    case PIX_FMT_RGB48BE:\
-    case PIX_FMT_RGB48LE:\
-        func(uint8_t,0)\
-            ((uint8_t*)dest)[ 0]= r[Y1];\
-            ((uint8_t*)dest)[ 1]= r[Y1];\
-            ((uint8_t*)dest)[ 2]= g[Y1];\
-            ((uint8_t*)dest)[ 3]= g[Y1];\
-            ((uint8_t*)dest)[ 4]= b[Y1];\
-            ((uint8_t*)dest)[ 5]= b[Y1];\
-            ((uint8_t*)dest)[ 6]= r[Y2];\
-            ((uint8_t*)dest)[ 7]= r[Y2];\
-            ((uint8_t*)dest)[ 8]= g[Y2];\
-            ((uint8_t*)dest)[ 9]= g[Y2];\
-            ((uint8_t*)dest)[10]= b[Y2];\
-            ((uint8_t*)dest)[11]= b[Y2];\
-            dest+=12;\
-        }\
-        break;\
-    case PIX_FMT_BGR48BE:\
-    case PIX_FMT_BGR48LE:\
-        func(uint8_t,0)\
-            ((uint8_t*)dest)[ 0] = ((uint8_t*)dest)[ 1] = b[Y1];\
-            ((uint8_t*)dest)[ 2] = ((uint8_t*)dest)[ 3] = g[Y1];\
-            ((uint8_t*)dest)[ 4] = ((uint8_t*)dest)[ 5] = r[Y1];\
-            ((uint8_t*)dest)[ 6] = ((uint8_t*)dest)[ 7] = b[Y2];\
-            ((uint8_t*)dest)[ 8] = ((uint8_t*)dest)[ 9] = g[Y2];\
-            ((uint8_t*)dest)[10] = ((uint8_t*)dest)[11] = r[Y2];\
-            dest+=12;\
-        }\
-        break;\
-    case PIX_FMT_RGBA:\
-    case PIX_FMT_BGRA:\
-        if (CONFIG_SMALL) {\
-            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
-            func(uint32_t,needAlpha)\
-                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? (A1<<24) : 0);\
-                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? (A2<<24) : 0);\
-            }\
-        } else {\
-            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
-                func(uint32_t,1)\
-                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (A1<<24);\
-                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (A2<<24);\
-                }\
-            } else {\
-                func(uint32_t,0)\
-                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
-                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
-                }\
-            }\
-        }\
-        break;\
-    case PIX_FMT_ARGB:\
-    case PIX_FMT_ABGR:\
-        if (CONFIG_SMALL) {\
-            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;\
-            func(uint32_t,needAlpha)\
-                ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + (needAlpha ? A1 : 0);\
-                ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + (needAlpha ? A2 : 0);\
-            }\
-        } else {\
-            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {\
-                func(uint32_t,1)\
-                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1] + A1;\
-                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2] + A2;\
-                }\
-            } else {\
-                func(uint32_t,0)\
-                    ((uint32_t*)dest)[i2+0]= r[Y1] + g[Y1] + b[Y1];\
-                    ((uint32_t*)dest)[i2+1]= r[Y2] + g[Y2] + b[Y2];\
-                }\
-            }\
-        }                \
-        break;\
-    case PIX_FMT_RGB24:\
-        func(uint8_t,0)\
-            ((uint8_t*)dest)[0]= r[Y1];\
-            ((uint8_t*)dest)[1]= g[Y1];\
-            ((uint8_t*)dest)[2]= b[Y1];\
-            ((uint8_t*)dest)[3]= r[Y2];\
-            ((uint8_t*)dest)[4]= g[Y2];\
-            ((uint8_t*)dest)[5]= b[Y2];\
-            dest+=6;\
-        }\
-        break;\
-    case PIX_FMT_BGR24:\
-        func(uint8_t,0)\
-            ((uint8_t*)dest)[0]= b[Y1];\
-            ((uint8_t*)dest)[1]= g[Y1];\
-            ((uint8_t*)dest)[2]= r[Y1];\
-            ((uint8_t*)dest)[3]= b[Y2];\
-            ((uint8_t*)dest)[4]= g[Y2];\
-            ((uint8_t*)dest)[5]= r[Y2];\
-            dest+=6;\
-        }\
-        break;\
-    case PIX_FMT_RGB565BE:\
-    case PIX_FMT_RGB565LE:\
-    case PIX_FMT_BGR565BE:\
-    case PIX_FMT_BGR565LE:\
-        {\
-            const int dr1= dither_2x2_8[y&1    ][0];\
-            const int dg1= dither_2x2_4[y&1    ][0];\
-            const int db1= dither_2x2_8[(y&1)^1][0];\
-            const int dr2= dither_2x2_8[y&1    ][1];\
-            const int dg2= dither_2x2_4[y&1    ][1];\
-            const int db2= dither_2x2_8[(y&1)^1][1];\
-            func(uint16_t,0)\
-                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
-                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
-            }\
-        }\
-        break;\
-    case PIX_FMT_RGB555BE:\
-    case PIX_FMT_RGB555LE:\
-    case PIX_FMT_BGR555BE:\
-    case PIX_FMT_BGR555LE:\
-        {\
-            const int dr1= dither_2x2_8[y&1    ][0];\
-            const int dg1= dither_2x2_8[y&1    ][1];\
-            const int db1= dither_2x2_8[(y&1)^1][0];\
-            const int dr2= dither_2x2_8[y&1    ][1];\
-            const int dg2= dither_2x2_8[y&1    ][0];\
-            const int db2= dither_2x2_8[(y&1)^1][1];\
-            func(uint16_t,0)\
-                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
-                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
-            }\
-        }\
-        break;\
-    case PIX_FMT_RGB444BE:\
-    case PIX_FMT_RGB444LE:\
-    case PIX_FMT_BGR444BE:\
-    case PIX_FMT_BGR444LE:\
-        {\
-            const int dr1= dither_4x4_16[y&3    ][0];\
-            const int dg1= dither_4x4_16[y&3    ][1];\
-            const int db1= dither_4x4_16[(y&3)^3][0];\
-            const int dr2= dither_4x4_16[y&3    ][1];\
-            const int dg2= dither_4x4_16[y&3    ][0];\
-            const int db2= dither_4x4_16[(y&3)^3][1];\
-            func(uint16_t,0)\
-                ((uint16_t*)dest)[i2+0]= r[Y1+dr1] + g[Y1+dg1] + b[Y1+db1];\
-                ((uint16_t*)dest)[i2+1]= r[Y2+dr2] + g[Y2+dg2] + b[Y2+db2];\
-            }\
-        }\
-        break;\
-    case PIX_FMT_RGB8:\
-    case PIX_FMT_BGR8:\
-        {\
-            const uint8_t * const d64= dither_8x8_73[y&7];\
-            const uint8_t * const d32= dither_8x8_32[y&7];\
-            func(uint8_t,0)\
-                ((uint8_t*)dest)[i2+0]= r[Y1+d32[(i2+0)&7]] + g[Y1+d32[(i2+0)&7]] + b[Y1+d64[(i2+0)&7]];\
-                ((uint8_t*)dest)[i2+1]= r[Y2+d32[(i2+1)&7]] + g[Y2+d32[(i2+1)&7]] + b[Y2+d64[(i2+1)&7]];\
-            }\
-        }\
-        break;\
-    case PIX_FMT_RGB4:\
-    case PIX_FMT_BGR4:\
-        {\
-            const uint8_t * const d64= dither_8x8_73 [y&7];\
-            const uint8_t * const d128=dither_8x8_220[y&7];\
-            func(uint8_t,0)\
-                ((uint8_t*)dest)[i]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]]\
-                                 + ((r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]])<<4);\
-            }\
-        }\
-        break;\
-    case PIX_FMT_RGB4_BYTE:\
-    case PIX_FMT_BGR4_BYTE:\
-        {\
-            const uint8_t * const d64= dither_8x8_73 [y&7];\
-            const uint8_t * const d128=dither_8x8_220[y&7];\
-            func(uint8_t,0)\
-                ((uint8_t*)dest)[i2+0]= r[Y1+d128[(i2+0)&7]] + g[Y1+d64[(i2+0)&7]] + b[Y1+d128[(i2+0)&7]];\
-                ((uint8_t*)dest)[i2+1]= r[Y2+d128[(i2+1)&7]] + g[Y2+d64[(i2+1)&7]] + b[Y2+d128[(i2+1)&7]];\
-            }\
-        }\
-        break;\
-    case PIX_FMT_MONOBLACK:\
-    case PIX_FMT_MONOWHITE:\
-        {\
-            func_monoblack\
-        }\
-        break;\
-    case PIX_FMT_YUYV422:\
-        func2\
-            ((uint8_t*)dest)[2*i2+0]= Y1;\
-            ((uint8_t*)dest)[2*i2+1]= U;\
-            ((uint8_t*)dest)[2*i2+2]= Y2;\
-            ((uint8_t*)dest)[2*i2+3]= V;\
-        }                \
-        break;\
-    case PIX_FMT_UYVY422:\
-        func2\
-            ((uint8_t*)dest)[2*i2+0]= U;\
-            ((uint8_t*)dest)[2*i2+1]= Y1;\
-            ((uint8_t*)dest)[2*i2+2]= V;\
-            ((uint8_t*)dest)[2*i2+3]= Y2;\
-        }                \
-        break;\
-    case PIX_FMT_GRAY16BE:\
-        func_g16\
-            ((uint8_t*)dest)[2*i2+0]= Y1>>8;\
-            ((uint8_t*)dest)[2*i2+1]= Y1;\
-            ((uint8_t*)dest)[2*i2+2]= Y2>>8;\
-            ((uint8_t*)dest)[2*i2+3]= Y2;\
-        }                \
-        break;\
-    case PIX_FMT_GRAY16LE:\
-        func_g16\
-            ((uint8_t*)dest)[2*i2+0]= Y1;\
-            ((uint8_t*)dest)[2*i2+1]= Y1>>8;\
-            ((uint8_t*)dest)[2*i2+2]= Y2;\
-            ((uint8_t*)dest)[2*i2+3]= Y2>>8;\
-        }                \
-        break;\
-    }
-
-static void yuv2packedX_c(SwsContext *c, const int16_t *lumFilter,
-                          const int16_t **lumSrc, int lumFilterSize,
-                          const int16_t *chrFilter, const int16_t **chrUSrc,
-                          const int16_t **chrVSrc, int chrFilterSize,
-                          const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
+        /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/
+        r = (const uint8_t *) c->table_rV[V];
+        g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]);
+        b = (const uint8_t *) c->table_bU[U];
+
+        dest[ 0] = dest[ 1] = r_b[Y1];
+        dest[ 2] = dest[ 3] =   g[Y1];
+        dest[ 4] = dest[ 5] = b_r[Y1];
+        dest[ 6] = dest[ 7] = r_b[Y2];
+        dest[ 8] = dest[ 9] =   g[Y2];
+        dest[10] = dest[11] = b_r[Y2];
+        dest += 12;
+    }
+}
+
+static av_always_inline void
+yuv2rgb48_2_c_template(SwsContext *c, const uint16_t *buf0,
+                       const uint16_t *buf1, const uint16_t *ubuf0,
+                       const uint16_t *ubuf1, const uint16_t *vbuf0,
+                       const uint16_t *vbuf1, const uint16_t *abuf0,
+                       const uint16_t *abuf1, uint8_t *dest, int dstW,
+                       int yalpha, int uvalpha, int y,
+                       enum PixelFormat target)
 {
+    int  yalpha1 = 4095 - yalpha;
+    int uvalpha1 = 4095 - uvalpha;
     int i;
-    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGBX_C, YSCALE_YUV_2_PACKEDX_C(void,0), YSCALE_YUV_2_GRAY16_C, YSCALE_YUV_2_MONOX_C)
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int Y1 = (buf0[i * 2]     * yalpha1  + buf1[i * 2]     * yalpha)  >> 19;
+        int Y2 = (buf0[i * 2 + 1] * yalpha1  + buf1[i * 2 + 1] * yalpha)  >> 19;
+        int U  = (ubuf0[i]        * uvalpha1 + ubuf1[i]        * uvalpha) >> 19;
+        int V  = (vbuf0[i]        * uvalpha1 + vbuf1[i]        * uvalpha) >> 19;
+        const uint8_t *r = (const uint8_t *) c->table_rV[V],
+                      *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]),
+                      *b = (const uint8_t *) c->table_bU[U];
+
+        dest[ 0] = dest[ 1] = r_b[Y1];
+        dest[ 2] = dest[ 3] =   g[Y1];
+        dest[ 4] = dest[ 5] = b_r[Y1];
+        dest[ 6] = dest[ 7] = r_b[Y2];
+        dest[ 8] = dest[ 9] =   g[Y2];
+        dest[10] = dest[11] = b_r[Y2];
+        dest += 12;
+    }
 }
 
-static inline void yuv2rgbX_c_full(SwsContext *c, const int16_t *lumFilter,
-                                   const int16_t **lumSrc, int lumFilterSize,
-                                   const int16_t *chrFilter, const int16_t **chrUSrc,
-                                   const int16_t **chrVSrc, int chrFilterSize,
-                                   const int16_t **alpSrc, uint8_t *dest, int dstW, int y)
+static av_always_inline void
+yuv2rgb48_1_c_template(SwsContext *c, const uint16_t *buf0,
+                       const uint16_t *ubuf0, const uint16_t *ubuf1,
+                       const uint16_t *vbuf0, const uint16_t *vbuf1,
+                       const uint16_t *abuf0, uint8_t *dest, int dstW,
+                       int uvalpha, enum PixelFormat dstFormat,
+                       int flags, int y, enum PixelFormat target)
 {
     int i;
-    int step= c->dstFormatBpp/8;
-    int aidx= 3;
-
-    switch(c->dstFormat) {
-    case PIX_FMT_ARGB:
-        dest++;
-        aidx= 0;
-    case PIX_FMT_RGB24:
-        aidx--;
-    case PIX_FMT_RGBA:
-        if (CONFIG_SMALL) {
-            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
-            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
-                dest[aidx]= needAlpha ? A : 255;
-                dest[0]= R>>22;
-                dest[1]= G>>22;
-                dest[2]= B>>22;
-                dest+= step;
-            }
+
+    if (uvalpha < 2048) {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 = buf0[i * 2]     >> 7;
+            int Y2 = buf0[i * 2 + 1] >> 7;
+            int U  = ubuf1[i]        >> 7;
+            int V  = vbuf1[i]        >> 7;
+            const uint8_t *r = (const uint8_t *) c->table_rV[V],
+                          *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]),
+                          *b = (const uint8_t *) c->table_bU[U];
+
+            dest[ 0] = dest[ 1] = r_b[Y1];
+            dest[ 2] = dest[ 3] =   g[Y1];
+            dest[ 4] = dest[ 5] = b_r[Y1];
+            dest[ 6] = dest[ 7] = r_b[Y2];
+            dest[ 8] = dest[ 9] =   g[Y2];
+            dest[10] = dest[11] = b_r[Y2];
+            dest += 12;
+        }
+    } else {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 =  buf0[i * 2]          >> 7;
+            int Y2 =  buf0[i * 2 + 1]      >> 7;
+            int U  = (ubuf0[i] + ubuf1[i]) >> 8;
+            int V  = (vbuf0[i] + vbuf1[i]) >> 8;
+            const uint8_t *r = (const uint8_t *) c->table_rV[V],
+                          *g = (const uint8_t *)(c->table_gU[U] + c->table_gV[V]),
+                          *b = (const uint8_t *) c->table_bU[U];
+
+            dest[ 0] = dest[ 1] = r_b[Y1];
+            dest[ 2] = dest[ 3] =   g[Y1];
+            dest[ 4] = dest[ 5] = b_r[Y1];
+            dest[ 6] = dest[ 7] = r_b[Y2];
+            dest[ 8] = dest[ 9] =   g[Y2];
+            dest[10] = dest[11] = b_r[Y2];
+            dest += 12;
+        }
+    }
+}
+
+#undef r_b
+#undef b_r
+
+YUV2PACKEDWRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE);
+//YUV2PACKEDWRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE);
+YUV2PACKEDWRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE);
+//YUV2PACKEDWRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE);
+
+static av_always_inline void
+yuv2rgb_write(uint8_t *_dest, int i, int Y1, int Y2,
+              int U, int V, int A1, int A2,
+              const void *_r, const void *_g, const void *_b, int y,
+              enum PixelFormat target, int hasAlpha)
+{
+    if (target == PIX_FMT_ARGB || target == PIX_FMT_RGBA ||
+        target == PIX_FMT_ABGR || target == PIX_FMT_BGRA) {
+        uint32_t *dest = (uint32_t *) _dest;
+        const uint32_t *r = (const uint32_t *) _r;
+        const uint32_t *g = (const uint32_t *) _g;
+        const uint32_t *b = (const uint32_t *) _b;
+
+#if CONFIG_SMALL
+        int sh = hasAlpha ? ((fmt == PIX_FMT_RGB32_1 || fmt == PIX_FMT_BGR32_1) ? 0 : 24) : 0;
+
+        dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (hasAlpha ? A1 << sh : 0);
+        dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (hasAlpha ? A2 << sh : 0);
+#else
+        if (hasAlpha) {
+            int sh = (target == PIX_FMT_RGB32_1 || target == PIX_FMT_BGR32_1) ? 0 : 24;
+
+            dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1] + (A1 << sh);
+            dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2] + (A2 << sh);
         } else {
-            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
-                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
-                    dest[aidx]= A;
-                    dest[0]= R>>22;
-                    dest[1]= G>>22;
-                    dest[2]= B>>22;
-                    dest+= step;
-                }
-            } else {
-                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
-                    dest[aidx]= 255;
-                    dest[0]= R>>22;
-                    dest[1]= G>>22;
-                    dest[2]= B>>22;
-                    dest+= step;
-                }
+            dest[i * 2 + 0] = r[Y1] + g[Y1] + b[Y1];
+            dest[i * 2 + 1] = r[Y2] + g[Y2] + b[Y2];
+        }
+#endif
+    } else if (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) {
+        uint8_t *dest = (uint8_t *) _dest;
+        const uint8_t *r = (const uint8_t *) _r;
+        const uint8_t *g = (const uint8_t *) _g;
+        const uint8_t *b = (const uint8_t *) _b;
+
+#define r_b ((target == PIX_FMT_RGB24) ? r : b)
+#define b_r ((target == PIX_FMT_RGB24) ? b : r)
+        dest[i * 6 + 0] = r_b[Y1];
+        dest[i * 6 + 1] =   g[Y1];
+        dest[i * 6 + 2] = b_r[Y1];
+        dest[i * 6 + 3] = r_b[Y2];
+        dest[i * 6 + 4] =   g[Y2];
+        dest[i * 6 + 5] = b_r[Y2];
+#undef r_b
+#undef b_r
+    } else if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565 ||
+               target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555 ||
+               target == PIX_FMT_RGB444 || target == PIX_FMT_BGR444) {
+        uint16_t *dest = (uint16_t *) _dest;
+        const uint16_t *r = (const uint16_t *) _r;
+        const uint16_t *g = (const uint16_t *) _g;
+        const uint16_t *b = (const uint16_t *) _b;
+        int dr1, dg1, db1, dr2, dg2, db2;
+
+        if (target == PIX_FMT_RGB565 || target == PIX_FMT_BGR565) {
+            dr1 = dither_2x2_8[ y & 1     ][0];
+            dg1 = dither_2x2_4[ y & 1     ][0];
+            db1 = dither_2x2_8[(y & 1) ^ 1][0];
+            dr2 = dither_2x2_8[ y & 1     ][1];
+            dg2 = dither_2x2_4[ y & 1     ][1];
+            db2 = dither_2x2_8[(y & 1) ^ 1][1];
+        } else if (target == PIX_FMT_RGB555 || target == PIX_FMT_BGR555) {
+            dr1 = dither_2x2_8[ y & 1     ][0];
+            dg1 = dither_2x2_8[ y & 1     ][1];
+            db1 = dither_2x2_8[(y & 1) ^ 1][0];
+            dr2 = dither_2x2_8[ y & 1     ][1];
+            dg2 = dither_2x2_8[ y & 1     ][0];
+            db2 = dither_2x2_8[(y & 1) ^ 1][1];
+        } else {
+            dr1 = dither_4x4_16[ y & 3     ][0];
+            dg1 = dither_4x4_16[ y & 3     ][1];
+            db1 = dither_4x4_16[(y & 3) ^ 3][0];
+            dr2 = dither_4x4_16[ y & 3     ][1];
+            dg2 = dither_4x4_16[ y & 3     ][0];
+            db2 = dither_4x4_16[(y & 3) ^ 3][1];
+        }
+
+        dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
+        dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
+    } else /* 8/4-bit */ {
+        uint8_t *dest = (uint8_t *) _dest;
+        const uint8_t *r = (const uint8_t *) _r;
+        const uint8_t *g = (const uint8_t *) _g;
+        const uint8_t *b = (const uint8_t *) _b;
+        int dr1, dg1, db1, dr2, dg2, db2;
+
+        if (target == PIX_FMT_RGB8 || target == PIX_FMT_BGR8) {
+            const uint8_t * const d64 = dither_8x8_73[y & 7];
+            const uint8_t * const d32 = dither_8x8_32[y & 7];
+            dr1 = dg1 = d32[(i * 2 + 0) & 7];
+            db1 =       d64[(i * 2 + 0) & 7];
+            dr2 = dg2 = d32[(i * 2 + 1) & 7];
+            db2 =       d64[(i * 2 + 1) & 7];
+        } else {
+            const uint8_t * const d64  = dither_8x8_73 [y & 7];
+            const uint8_t * const d128 = dither_8x8_220[y & 7];
+            dr1 = db1 = d128[(i * 2 + 0) & 7];
+            dg1 =        d64[(i * 2 + 0) & 7];
+            dr2 = db2 = d128[(i * 2 + 1) & 7];
+            dg2 =        d64[(i * 2 + 1) & 7];
+        }
+
+        if (target == PIX_FMT_RGB4 || target == PIX_FMT_BGR4) {
+            dest[i] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1] +
+                    ((r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2]) << 4);
+        } else {
+            dest[i * 2 + 0] = r[Y1 + dr1] + g[Y1 + dg1] + b[Y1 + db1];
+            dest[i * 2 + 1] = r[Y2 + dr2] + g[Y2 + dg2] + b[Y2 + db2];
+        }
+    }
+}
+
+static av_always_inline void
+yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                     const int16_t **lumSrc, int lumFilterSize,
+                     const int16_t *chrFilter, const int16_t **chrUSrc,
+                     const int16_t **chrVSrc, int chrFilterSize,
+                     const int16_t **alpSrc, uint8_t *dest, int dstW,
+                     int y, enum PixelFormat target, int hasAlpha)
+{
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int j;
+        int Y1 = 1 << 18;
+        int Y2 = 1 << 18;
+        int U  = 1 << 18;
+        int V  = 1 << 18;
+        int av_unused A1, A2;
+        const void *r, *g, *b;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y1 += lumSrc[j][i * 2]     * lumFilter[j];
+            Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
+        }
+        for (j = 0; j < chrFilterSize; j++) {
+            U += chrUSrc[j][i] * chrFilter[j];
+            V += chrVSrc[j][i] * chrFilter[j];
+        }
+        Y1 >>= 19;
+        Y2 >>= 19;
+        U  >>= 19;
+        V  >>= 19;
+        if ((Y1 | Y2 | U | V) & 0x100) {
+            Y1 = av_clip_uint8(Y1);
+            Y2 = av_clip_uint8(Y2);
+            U  = av_clip_uint8(U);
+            V  = av_clip_uint8(V);
+        }
+        if (hasAlpha) {\
+            A1 = 1 << 18;
+            A2 = 1 << 18;
+            for (j = 0; j < lumFilterSize; j++) {
+                A1 += alpSrc[j][i * 2    ] * lumFilter[j];
+                A2 += alpSrc[j][i * 2 + 1] * lumFilter[j];
             }
+            A1 >>= 19;
+            A2 >>= 19;
+            if ((A1 | A2) & 0x100) {
+                A1 = av_clip_uint8(A1);
+                A2 = av_clip_uint8(A2);
+            }
+        }
+
+        /* FIXME fix tables so that clipping is not needed and then use _NOCLIP*/
+        r =  c->table_rV[V];
+        g = (c->table_gU[U] + c->table_gV[V]);
+        b =  c->table_bU[U];
+
+        yuv2rgb_write(dest, i, Y1, Y2, U, V, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
+                      r, g, b, y, target, hasAlpha);
+    }
+}
+
+static av_always_inline void
+yuv2rgb_2_c_template(SwsContext *c, const uint16_t *buf0,
+                     const uint16_t *buf1, const uint16_t *ubuf0,
+                     const uint16_t *ubuf1, const uint16_t *vbuf0,
+                     const uint16_t *vbuf1, const uint16_t *abuf0,
+                     const uint16_t *abuf1, uint8_t *dest, int dstW,
+                     int yalpha, int uvalpha, int y,
+                     enum PixelFormat target, int hasAlpha)
+{
+    int  yalpha1 = 4095 - yalpha;
+    int uvalpha1 = 4095 - uvalpha;
+    int i;
+
+    for (i = 0; i < (dstW >> 1); i++) {
+        int Y1 = (buf0[i * 2]     * yalpha1  + buf1[i * 2]     * yalpha)  >> 19;
+        int Y2 = (buf0[i * 2 + 1] * yalpha1  + buf1[i * 2 + 1] * yalpha)  >> 19;
+        int U  = (ubuf0[i]        * uvalpha1 + ubuf1[i]        * uvalpha) >> 19;
+        int V  = (vbuf0[i]        * uvalpha1 + vbuf1[i]        * uvalpha) >> 19;
+        int A1, A2;
+        const void *r =  c->table_rV[V],
+                   *g = (c->table_gU[U] + c->table_gV[V]),
+                   *b =  c->table_bU[U];
+
+        if (hasAlpha) {
+            A1 = (abuf0[i * 2    ] * yalpha1 + abuf1[i * 2    ] * yalpha) >> 19;
+            A2 = (abuf0[i * 2 + 1] * yalpha1 + abuf1[i * 2 + 1] * yalpha) >> 19;
         }
-        break;
-    case PIX_FMT_ABGR:
-        dest++;
-        aidx= 0;
-    case PIX_FMT_BGR24:
-        aidx--;
-    case PIX_FMT_BGRA:
-        if (CONFIG_SMALL) {
-            int needAlpha = CONFIG_SWSCALE_ALPHA && c->alpPixBuf;
-            YSCALE_YUV_2_RGBX_FULL_C(1<<21, needAlpha)
-                dest[aidx]= needAlpha ? A : 255;
-                dest[0]= B>>22;
-                dest[1]= G>>22;
-                dest[2]= R>>22;
-                dest+= step;
+
+        yuv2rgb_write(dest, i, Y1, Y2, U, V, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
+                      r, g, b, y, target, hasAlpha);
+    }
+}
+
+static av_always_inline void
+yuv2rgb_1_c_template(SwsContext *c, const uint16_t *buf0,
+                     const uint16_t *ubuf0, const uint16_t *ubuf1,
+                     const uint16_t *vbuf0, const uint16_t *vbuf1,
+                     const uint16_t *abuf0, uint8_t *dest, int dstW,
+                     int uvalpha, enum PixelFormat dstFormat,
+                     int flags, int y, enum PixelFormat target,
+                     int hasAlpha)
+{
+    int i;
+
+    if (uvalpha < 2048) {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 = buf0[i * 2]     >> 7;
+            int Y2 = buf0[i * 2 + 1] >> 7;
+            int U  = ubuf1[i]        >> 7;
+            int V  = vbuf1[i]        >> 7;
+            int A1, A2;
+            const void *r =  c->table_rV[V],
+                       *g = (c->table_gU[U] + c->table_gV[V]),
+                       *b =  c->table_bU[U];
+
+            if (hasAlpha) {
+                A1 = abuf0[i * 2    ] >> 7;
+                A2 = abuf0[i * 2 + 1] >> 7;
             }
-        } else {
-            if (CONFIG_SWSCALE_ALPHA && c->alpPixBuf) {
-                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 1)
-                    dest[aidx]= A;
-                    dest[0]= B>>22;
-                    dest[1]= G>>22;
-                    dest[2]= R>>22;
-                    dest+= step;
-                }
-            } else {
-                YSCALE_YUV_2_RGBX_FULL_C(1<<21, 0)
-                    dest[aidx]= 255;
-                    dest[0]= B>>22;
-                    dest[1]= G>>22;
-                    dest[2]= R>>22;
-                    dest+= step;
-                }
+
+            yuv2rgb_write(dest, i, Y1, Y2, U, V, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
+                          r, g, b, y, target, hasAlpha);
+        }
+    } else {
+        for (i = 0; i < (dstW >> 1); i++) {
+            int Y1 =  buf0[i * 2]          >> 7;
+            int Y2 =  buf0[i * 2 + 1]      >> 7;
+            int U  = (ubuf0[i] + ubuf1[i]) >> 8;
+            int V  = (vbuf0[i] + vbuf1[i]) >> 8;
+            int A1, A2;
+            const void *r =  c->table_rV[V],
+                       *g = (c->table_gU[U] + c->table_gV[V]),
+                       *b =  c->table_bU[U];
+
+            if (hasAlpha) {
+                A1 = abuf0[i * 2    ] >> 7;
+                A2 = abuf0[i * 2 + 1] >> 7;
+            }
+
+            yuv2rgb_write(dest, i, Y1, Y2, U, V, hasAlpha ? A1 : 0, hasAlpha ? A2 : 0,
+                          r, g, b, y, target, hasAlpha);
+        }
+    }
+}
+
+#define YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
+static void name ## ext ## _X_c(SwsContext *c, const int16_t *lumFilter, \
+                                const int16_t **lumSrc, int lumFilterSize, \
+                                const int16_t *chrFilter, const int16_t **chrUSrc, \
+                                const int16_t **chrVSrc, int chrFilterSize, \
+                                const int16_t **alpSrc, uint8_t *dest, int dstW, \
+                                int y) \
+{ \
+    name ## base ## _X_c_template(c, lumFilter, lumSrc, lumFilterSize, \
+                                  chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
+                                  alpSrc, dest, dstW, y, fmt, hasAlpha); \
+}
+#define YUV2RGBWRAPPER(name, base, ext, fmt, hasAlpha) \
+YUV2RGBWRAPPERX(name, base, ext, fmt, hasAlpha) \
+static void name ## ext ## _2_c(SwsContext *c, const uint16_t *buf0, \
+                                const uint16_t *buf1, const uint16_t *ubuf0, \
+                                const uint16_t *ubuf1, const uint16_t *vbuf0, \
+                                const uint16_t *vbuf1, const uint16_t *abuf0, \
+                                const uint16_t *abuf1, uint8_t *dest, int dstW, \
+                                int yalpha, int uvalpha, int y) \
+{ \
+    name ## base ## _2_c_template(c, buf0, buf1, ubuf0, ubuf1, \
+                                  vbuf0, vbuf1, abuf0, abuf1, \
+                                  dest, dstW, yalpha, uvalpha, y, fmt, hasAlpha); \
+} \
+ \
+static void name ## ext ## _1_c(SwsContext *c, const uint16_t *buf0, \
+                                const uint16_t *ubuf0, const uint16_t *ubuf1, \
+                                const uint16_t *vbuf0, const uint16_t *vbuf1, \
+                                const uint16_t *abuf0, uint8_t *dest, int dstW, \
+                                int uvalpha, enum PixelFormat dstFormat, \
+                                int flags, int y) \
+{ \
+    name ## base ## _1_c_template(c, buf0, ubuf0, ubuf1, vbuf0, \
+                                  vbuf1, abuf0, dest, dstW, uvalpha, \
+                                  dstFormat, flags, y, fmt, hasAlpha); \
+}
+
+#if CONFIG_SMALL
+YUV2RGBWRAPPER(yuv2rgb,,  32_1,  PIX_FMT_RGB32_1,   CONFIG_SWSCALE_ALPHA && c->alpPixBuf);
+YUV2RGBWRAPPER(yuv2rgb,,  32,    PIX_FMT_RGB32,     CONFIG_SWSCALE_ALPHA && c->alpPixBuf);
+#else
+#if CONFIG_SWSCALE_ALPHA
+YUV2RGBWRAPPER(yuv2rgb,, a32_1,  PIX_FMT_RGB32_1,   1);
+YUV2RGBWRAPPER(yuv2rgb,, a32,    PIX_FMT_RGB32,     1);
+#endif
+YUV2RGBWRAPPER(yuv2rgb,, x32_1,  PIX_FMT_RGB32_1,   0);
+YUV2RGBWRAPPER(yuv2rgb,, x32,    PIX_FMT_RGB32,     0);
+#endif
+YUV2RGBWRAPPER(yuv2, rgb, rgb24, PIX_FMT_RGB24,   0);
+YUV2RGBWRAPPER(yuv2, rgb, bgr24, PIX_FMT_BGR24,   0);
+YUV2RGBWRAPPER(yuv2rgb,,  16,    PIX_FMT_RGB565,    0);
+YUV2RGBWRAPPER(yuv2rgb,,  15,    PIX_FMT_RGB555,    0);
+YUV2RGBWRAPPER(yuv2rgb,,  12,    PIX_FMT_RGB444,    0);
+YUV2RGBWRAPPER(yuv2rgb,,   8,    PIX_FMT_RGB8,      0);
+YUV2RGBWRAPPER(yuv2rgb,,   4,    PIX_FMT_RGB4,      0);
+YUV2RGBWRAPPER(yuv2rgb,,   4b,   PIX_FMT_RGB4_BYTE, 0);
+
+static av_always_inline void
+yuv2rgb_full_X_c_template(SwsContext *c, const int16_t *lumFilter,
+                          const int16_t **lumSrc, int lumFilterSize,
+                          const int16_t *chrFilter, const int16_t **chrUSrc,
+                          const int16_t **chrVSrc, int chrFilterSize,
+                          const int16_t **alpSrc, uint8_t *dest,
+                          int dstW, int y, enum PixelFormat target, int hasAlpha)
+{
+    int i;
+    int step = (target == PIX_FMT_RGB24 || target == PIX_FMT_BGR24) ? 3 : 4;
+
+    for (i = 0; i < dstW; i++) {
+        int j;
+        int Y = 0;
+        int U = -128 << 19;
+        int V = -128 << 19;
+        int av_unused A;
+        int R, G, B;
+
+        for (j = 0; j < lumFilterSize; j++) {
+            Y += lumSrc[j][i] * lumFilter[j];
+        }
+        for (j = 0; j < chrFilterSize; j++) {
+            U += chrUSrc[j][i] * chrFilter[j];
+            V += chrVSrc[j][i] * chrFilter[j];
+        }
+        Y >>= 10;
+        U >>= 10;
+        V >>= 10;
+        if (hasAlpha) {
+            A = 1 << 21;
+            for (j = 0; j < lumFilterSize; j++) {
+                A += alpSrc[j][i] * lumFilter[j];
             }
+            A >>= 19;
+            if (A & 0x100)
+                A = av_clip_uint8(A);
+        }
+        Y -= c->yuv2rgb_y_offset;
+        Y *= c->yuv2rgb_y_coeff;
+        Y += 1 << 21;
+        R = Y + V*c->yuv2rgb_v2r_coeff;
+        G = Y + V*c->yuv2rgb_v2g_coeff + U*c->yuv2rgb_u2g_coeff;
+        B = Y +                          U*c->yuv2rgb_u2b_coeff;
+        if ((R | G | B) & 0xC0000000) {
+            R = av_clip_uintp2(R, 30);
+            G = av_clip_uintp2(G, 30);
+            B = av_clip_uintp2(B, 30);
+        }
+
+        switch(target) {
+        case PIX_FMT_ARGB:
+            dest[0] = hasAlpha ? A : 255;
+            dest[1] = R >> 22;
+            dest[2] = G >> 22;
+            dest[3] = B >> 22;
+            break;
+        case PIX_FMT_RGB24:
+            dest[0] = R >> 22;
+            dest[1] = G >> 22;
+            dest[2] = B >> 22;
+            break;
+        case PIX_FMT_RGBA:
+            dest[0] = R >> 22;
+            dest[1] = G >> 22;
+            dest[2] = B >> 22;
+            dest[3] = hasAlpha ? A : 255;
+            break;
+        case PIX_FMT_ABGR:
+            dest[0] = hasAlpha ? A : 255;
+            dest[1] = B >> 22;
+            dest[2] = G >> 22;
+            dest[3] = R >> 22;
+            dest += 4;
+            break;
+        case PIX_FMT_BGR24:
+            dest[0] = B >> 22;
+            dest[1] = G >> 22;
+            dest[2] = R >> 22;
+            break;
+        case PIX_FMT_BGRA:
+            dest[0] = B >> 22;
+            dest[1] = G >> 22;
+            dest[2] = R >> 22;
+            dest[3] = hasAlpha ? A : 255;
+            break;
         }
-        break;
-    default:
-        assert(0);
+        dest += step;
     }
 }
 
-static void fillPlane(uint8_t* plane, int stride, int width, int height, int y, uint8_t val)
+#if CONFIG_SMALL
+YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA,  CONFIG_SWSCALE_ALPHA && c->alpPixBuf);
+YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR,  CONFIG_SWSCALE_ALPHA && c->alpPixBuf);
+YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA,  CONFIG_SWSCALE_ALPHA && c->alpPixBuf);
+YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB,  CONFIG_SWSCALE_ALPHA && c->alpPixBuf);
+#else
+#if CONFIG_SWSCALE_ALPHA
+YUV2RGBWRAPPERX(yuv2, rgb_full, bgra32_full, PIX_FMT_BGRA,  1);
+YUV2RGBWRAPPERX(yuv2, rgb_full, abgr32_full, PIX_FMT_ABGR,  1);
+YUV2RGBWRAPPERX(yuv2, rgb_full, rgba32_full, PIX_FMT_RGBA,  1);
+YUV2RGBWRAPPERX(yuv2, rgb_full, argb32_full, PIX_FMT_ARGB,  1);
+#endif
+YUV2RGBWRAPPERX(yuv2, rgb_full, bgrx32_full, PIX_FMT_BGRA,  0);
+YUV2RGBWRAPPERX(yuv2, rgb_full, xbgr32_full, PIX_FMT_ABGR,  0);
+YUV2RGBWRAPPERX(yuv2, rgb_full, rgbx32_full, PIX_FMT_RGBA,  0);
+YUV2RGBWRAPPERX(yuv2, rgb_full, xrgb32_full, PIX_FMT_ARGB,  0);
+#endif
+YUV2RGBWRAPPERX(yuv2, rgb_full, bgr24_full,  PIX_FMT_BGR24, 0);
+YUV2RGBWRAPPERX(yuv2, rgb_full, rgb24_full,  PIX_FMT_RGB24, 0);
+
+static av_always_inline void fillPlane(uint8_t* plane, int stride,
+                                       int width, int height,
+                                       int y, uint8_t val)
 {
     int i;
     uint8_t *ptr = plane + stride*y;
@@ -979,81 +1325,211 @@ static void fillPlane(uint8_t* plane, int stride, int width, int height, int y,
     }
 }
 
-#define rgb48funcs(LE_BE, rfunc, compA, compB, compC) \
-static void compA ## compB ## compC ## 48 ## LE_BE ## ToY_c( \
-                       uint8_t *dst, const uint8_t *src, int width, \
-                       uint32_t *unused) \
+#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
+
+#define r ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? b_r : r_b)
+#define b ((origin == PIX_FMT_BGR48BE || origin == PIX_FMT_BGR48LE) ? r_b : b_r)
+
+static av_always_inline void
+rgb48ToY_c_template(uint8_t *dst, const uint8_t *src, int width,
+                    enum PixelFormat origin)
+{
+    int i;
+    for (i = 0; i < width; i++) {
+        int r_b = input_pixel(&src[i*6+0]) >> 8;
+        int   g = input_pixel(&src[i*6+2]) >> 8;
+        int b_r = input_pixel(&src[i*6+4]) >> 8;
+
+        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
+    }
+}
+
+static av_always_inline void
+rgb48ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
+                    const uint8_t *src1, const uint8_t *src2,
+                    int width, enum PixelFormat origin)
+{
+    int i;
+    assert(src1==src2);
+    for (i = 0; i < width; i++) {
+        int r_b = input_pixel(&src1[i*6+0]) >> 8;
+        int   g = input_pixel(&src1[i*6+2]) >> 8;
+        int b_r = input_pixel(&src1[i*6+4]) >> 8;
+
+        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
+        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
+    }
+}
+
+static av_always_inline void
+rgb48ToUV_half_c_template(uint8_t *dstU, uint8_t *dstV,
+                          const uint8_t *src1, const uint8_t *src2,
+                          int width, enum PixelFormat origin)
+{
+    int i;
+    assert(src1==src2);
+    for (i = 0; i < width; i++) {
+        int r_b = (input_pixel(&src1[12*i + 0]) >> 8) + (input_pixel(&src1[12*i + 6]) >> 8);
+        int   g = (input_pixel(&src1[12*i + 2]) >> 8) + (input_pixel(&src1[12*i + 8]) >> 8);
+        int b_r = (input_pixel(&src1[12*i + 4]) >> 8) + (input_pixel(&src1[12*i + 10]) >> 8);
+
+        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
+        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1);
+    }
+}
+
+#undef r
+#undef b
+#undef input_pixel
+
+#define rgb48funcs(pattern, BE_LE, origin) \
+static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *dst, const uint8_t *src, \
+                                    int width, uint32_t *unused) \
 { \
-    int i; \
-    for (i = 0; i < width; i++) { \
-        int compA = rfunc(&src[i*6+0]) >> 8; \
-        int compB = rfunc(&src[i*6+2]) >> 8; \
-        int compC = rfunc(&src[i*6+4]) >> 8; \
- \
-        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; \
-    } \
+    rgb48ToY_c_template(dst, src, width, origin); \
 } \
  \
-static void compA ## compB ## compC ## 48 ## LE_BE ## ToUV_c( \
-                        uint8_t *dstU, uint8_t *dstV, \
-                        const uint8_t *src1, const uint8_t *src2, \
-                        int width, uint32_t *unused) \
+static void pattern ## 48 ## BE_LE ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
+                                    const uint8_t *src1, const uint8_t *src2, \
+                                    int width, uint32_t *unused) \
 { \
-    int i; \
-    assert(src1==src2); \
-    for (i = 0; i < width; i++) { \
-        int compA = rfunc(&src1[6*i + 0]) >> 8; \
-        int compB = rfunc(&src1[6*i + 2]) >> 8; \
-        int compC = rfunc(&src1[6*i + 4]) >> 8; \
+    rgb48ToUV_c_template(dstU, dstV, src1, src2, width, origin); \
+} \
  \
-        dstU[i] = (RU*r + GU*g + BU*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; \
-        dstV[i] = (RV*r + GV*g + BV*b + (257<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT; \
-    } \
+static void pattern ## 48 ## BE_LE ## ToUV_half_c(uint8_t *dstU, uint8_t *dstV, \
+                                    const uint8_t *src1, const uint8_t *src2, \
+                                    int width, uint32_t *unused) \
+{ \
+    rgb48ToUV_half_c_template(dstU, dstV, src1, src2, width, origin); \
+}
+
+rgb48funcs(rgb, LE, PIX_FMT_RGB48LE);
+rgb48funcs(rgb, BE, PIX_FMT_RGB48BE);
+rgb48funcs(bgr, LE, PIX_FMT_BGR48LE);
+rgb48funcs(bgr, BE, PIX_FMT_BGR48BE);
+
+#define input_pixel(i) ((origin == PIX_FMT_RGBA || origin == PIX_FMT_BGRA || \
+                         origin == PIX_FMT_ARGB || origin == PIX_FMT_ABGR) ? AV_RN32A(&src[(i)*4]) : \
+                        (isBE(origin) ? AV_RB16(&src[(i)*2]) : AV_RL16(&src[(i)*2])))
+
+static av_always_inline void
+rgb16_32ToY_c_template(uint8_t *dst, const uint8_t *src,
+                       int width, enum PixelFormat origin,
+                       int shr,   int shg,   int shb, int shp,
+                       int maskr, int maskg, int maskb,
+                       int rsh,   int gsh,   int bsh, int S)
+{
+    const int ry = RY << rsh, gy = GY << gsh, by = BY << bsh,
+              rnd = 33 << (S - 1);
+    int i;
+
+    for (i = 0; i < width; i++) {
+        int px = input_pixel(i) >> shp;
+        int b = (px & maskb) >> shb;
+        int g = (px & maskg) >> shg;
+        int r = (px & maskr) >> shr;
+
+        dst[i] = (ry * r + gy * g + by * b + rnd) >> S;
+    }
+}
+
+static av_always_inline void
+rgb16_32ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
+                        const uint8_t *src, int width,
+                        enum PixelFormat origin,
+                        int shr,   int shg,   int shb, int shp,
+                        int maskr, int maskg, int maskb,
+                        int rsh,   int gsh,   int bsh, int S)
+{
+    const int ru = RU << rsh, gu = GU << gsh, bu = BU << bsh,
+              rv = RV << rsh, gv = GV << gsh, bv = BV << bsh,
+              rnd = 257 << (S - 1);
+    int i;
+
+    for (i = 0; i < width; i++) {
+        int px = input_pixel(i) >> shp;
+        int b = (px & maskb) >> shb;
+        int g = (px & maskg) >> shg;
+        int r = (px & maskr) >> shr;
+
+        dstU[i] = (ru * r + gu * g + bu * b + rnd) >> S;
+        dstV[i] = (rv * r + gv * g + bv * b + rnd) >> S;
+    }
+}
+
+static av_always_inline void
+rgb16_32ToUV_half_c_template(uint8_t *dstU, uint8_t *dstV,
+                             const uint8_t *src, int width,
+                             enum PixelFormat origin,
+                             int shr,   int shg,   int shb, int shp,
+                             int maskr, int maskg, int maskb,
+                             int rsh,   int gsh,   int bsh, int S)
+{
+    const int ru = RU << rsh, gu = GU << gsh, bu = BU << bsh,
+              rv = RV << rsh, gv = GV << gsh, bv = BV << bsh,
+              rnd = 257 << S, maskgx = ~(maskr | maskb);
+    int i;
+
+    maskr |= maskr << 1; maskb |= maskb << 1; maskg |= maskg << 1;
+    for (i = 0; i < width; i++) {
+        int px0 = input_pixel(2 * i + 0) >> shp;
+        int px1 = input_pixel(2 * i + 1) >> shp;
+        int b, r, g = (px0 & maskgx) + (px1 & maskgx);
+        int rb = px0 + px1 - g;
+
+        b = (rb & maskb) >> shb;
+        if (shp || origin == PIX_FMT_BGR565LE || origin == PIX_FMT_BGR565BE ||
+            origin == PIX_FMT_RGB565LE || origin == PIX_FMT_RGB565BE) {
+            g >>= shg;
+        } else {
+            g = (g  & maskg) >> shg;
+        }
+        r = (rb & maskr) >> shr;
+
+        dstU[i] = (ru * r + gu * g + bu * b + rnd) >> (S + 1);
+        dstV[i] = (rv * r + gv * g + bv * b + rnd) >> (S + 1);
+    }
+}
+
+#undef input_pixel
+
+#define rgb16_32_wrapper(fmt, name, shr, shg, shb, shp, maskr, \
+                         maskg, maskb, rsh, gsh, bsh, S) \
+static void name ## ToY_c(uint8_t *dst, const uint8_t *src, \
+                          int width, uint32_t *unused) \
+{ \
+    rgb16_32ToY_c_template(dst, src, width, fmt, shr, shg, shb, shp, \
+                           maskr, maskg, maskb, rsh, gsh, bsh, S); \
 } \
  \
-static void compA ## compB ## compC ## 48 ## LE_BE ## ToUV_half_c( \
-                            uint8_t *dstU, uint8_t *dstV, \
-                            const uint8_t *src1, const uint8_t *src2, \
-                            int width, uint32_t *unused) \
+static void name ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
+                           const uint8_t *src, const uint8_t *dummy, \
+                           int width, uint32_t *unused) \
 { \
-    int i; \
-    assert(src1==src2); \
-    for (i = 0; i < width; i++) { \
-        int compA = (rfunc(&src1[12*i + 0]) >> 8) + (rfunc(&src1[12*i + 6]) >> 8); \
-        int compB = (rfunc(&src1[12*i + 2]) >> 8) + (rfunc(&src1[12*i + 8]) >> 8); \
-        int compC = (rfunc(&src1[12*i + 4]) >> 8) + (rfunc(&src1[12*i + 10]) >> 8); \
+    rgb16_32ToUV_c_template(dstU, dstV, src, width, fmt, shr, shg, shb, shp, \
+                            maskr, maskg, maskb, rsh, gsh, bsh, S); \
+} \
  \
-        dstU[i]= (RU*r + GU*g + BU*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1); \
-        dstV[i]= (RV*r + GV*g + BV*b + (257<<RGB2YUV_SHIFT)) >> (RGB2YUV_SHIFT+1); \
-    } \
-}
-rgb48funcs(LE, AV_RL16, r, g, b);
-rgb48funcs(BE, AV_RB16, r, g, b);
-rgb48funcs(LE, AV_RL16, b, g, r);
-rgb48funcs(BE, AV_RB16, b, g, r);
-
-#define BGR2Y(type, name, shr, shg, shb, maskr, maskg, maskb, RY, GY, BY, S)\
-static void name ## _c(uint8_t *dst, const uint8_t *src, \
-                       int width, uint32_t *unused)\
-{\
-    int i;\
-    for (i=0; i<width; i++) {\
-        int b= (((const type*)src)[i]>>shb)&maskb;\
-        int g= (((const type*)src)[i]>>shg)&maskg;\
-        int r= (((const type*)src)[i]>>shr)&maskr;\
-\
-        dst[i]= (((RY)*r + (GY)*g + (BY)*b + (33<<((S)-1)))>>(S));\
-    }\
-}
-
-BGR2Y(uint32_t, bgr32ToY,16, 0, 0, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
-BGR2Y(uint32_t,bgr321ToY,16,16, 0, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)
-BGR2Y(uint32_t, rgb32ToY, 0, 0,16, 0x00FF, 0xFF00, 0x00FF, RY<< 8, GY   , BY<< 8, RGB2YUV_SHIFT+8)
-BGR2Y(uint32_t,rgb321ToY, 0,16,16, 0xFF00, 0x00FF, 0xFF00, RY    , GY<<8, BY    , RGB2YUV_SHIFT+8)
-BGR2Y(uint16_t, bgr16ToY, 0, 0, 0, 0x001F, 0x07E0, 0xF800, RY<<11, GY<<5, BY    , RGB2YUV_SHIFT+8)
-BGR2Y(uint16_t, bgr15ToY, 0, 0, 0, 0x001F, 0x03E0, 0x7C00, RY<<10, GY<<5, BY    , RGB2YUV_SHIFT+7)
-BGR2Y(uint16_t, rgb16ToY, 0, 0, 0, 0xF800, 0x07E0, 0x001F, RY    , GY<<5, BY<<11, RGB2YUV_SHIFT+8)
-BGR2Y(uint16_t, rgb15ToY, 0, 0, 0, 0x7C00, 0x03E0, 0x001F, RY    , GY<<5, BY<<10, RGB2YUV_SHIFT+7)
+static void name ## ToUV_half_c(uint8_t *dstU, uint8_t *dstV, \
+                                const uint8_t *src, const uint8_t *dummy, \
+                                int width, uint32_t *unused) \
+{ \
+    rgb16_32ToUV_half_c_template(dstU, dstV, src, width, fmt, shr, shg, shb, shp, \
+                                 maskr, maskg, maskb, rsh, gsh, bsh, S); \
+}
+
+rgb16_32_wrapper(PIX_FMT_BGR32,    bgr32,  16, 0,  0, 0, 0xFF0000, 0xFF00,   0x00FF,  8, 0,  8, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_BGR32_1,  bgr321, 16, 0,  0, 8, 0xFF0000, 0xFF00,   0x00FF,  8, 0,  8, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_RGB32,    rgb32,   0, 0, 16, 0,   0x00FF, 0xFF00, 0xFF0000,  8, 0,  8, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_RGB32_1,  rgb321,  0, 0, 16, 8,   0x00FF, 0xFF00, 0xFF0000,  8, 0,  8, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_BGR565LE, bgr16le, 0, 0,  0, 0,   0x001F, 0x07E0,   0xF800, 11, 5,  0, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_BGR555LE, bgr15le, 0, 0,  0, 0,   0x001F, 0x03E0,   0x7C00, 10, 5,  0, RGB2YUV_SHIFT+7);
+rgb16_32_wrapper(PIX_FMT_RGB565LE, rgb16le, 0, 0,  0, 0,   0xF800, 0x07E0,   0x001F,  0, 5, 11, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_RGB555LE, rgb15le, 0, 0,  0, 0,   0x7C00, 0x03E0,   0x001F,  0, 5, 10, RGB2YUV_SHIFT+7);
+rgb16_32_wrapper(PIX_FMT_BGR565BE, bgr16be, 0, 0,  0, 0,   0x001F, 0x07E0,   0xF800, 11, 5,  0, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_BGR555BE, bgr15be, 0, 0,  0, 0,   0x001F, 0x03E0,   0x7C00, 10, 5,  0, RGB2YUV_SHIFT+7);
+rgb16_32_wrapper(PIX_FMT_RGB565BE, rgb16be, 0, 0,  0, 0,   0xF800, 0x07E0,   0x001F,  0, 5, 11, RGB2YUV_SHIFT+8);
+rgb16_32_wrapper(PIX_FMT_RGB555BE, rgb15be, 0, 0,  0, 0,   0x7C00, 0x03E0,   0x001F,  0, 5, 10, RGB2YUV_SHIFT+7);
 
 static void abgrToA_c(uint8_t *dst, const uint8_t *src, int width, uint32_t *unused)
 {
@@ -1071,50 +1547,6 @@ static void rgbaToA_c(uint8_t *dst, const uint8_t *src, int width, uint32_t *unu
     }
 }
 
-#define BGR2UV(type, name, shr, shg, shb, shp, maskr, maskg, maskb, RU, GU, BU, RV, GV, BV, S) \
-static void name ## _c(uint8_t *dstU, uint8_t *dstV, \
-                       const uint8_t *src, const uint8_t *dummy, \
-                       int width, uint32_t *unused)\
-{\
-    int i;\
-    for (i=0; i<width; i++) {\
-        int b= ((((const type*)src)[i]>>shp)&maskb)>>shb;\
-        int g= ((((const type*)src)[i]>>shp)&maskg)>>shg;\
-        int r= ((((const type*)src)[i]>>shp)&maskr)>>shr;\
-\
-        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<((S)-1)))>>(S);\
-        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<((S)-1)))>>(S);\
-    }\
-}\
-static void name ## _half_c(uint8_t *dstU, uint8_t *dstV, \
-                            const uint8_t *src, const uint8_t *dummy, \
-                            int width, uint32_t *unused)\
-{\
-    int i;\
-    for (i=0; i<width; i++) {\
-        int pix0= ((const type*)src)[2*i+0]>>shp;\
-        int pix1= ((const type*)src)[2*i+1]>>shp;\
-        int g= (pix0&~(maskr|maskb))+(pix1&~(maskr|maskb));\
-        int b= ((pix0+pix1-g)&(maskb|(2*maskb)))>>shb;\
-        int r= ((pix0+pix1-g)&(maskr|(2*maskr)))>>shr;\
-        g&= maskg|(2*maskg);\
-\
-        g>>=shg;\
-\
-        dstU[i]= ((RU)*r + (GU)*g + (BU)*b + (257<<(S)))>>((S)+1);\
-        dstV[i]= ((RV)*r + (GV)*g + (BV)*b + (257<<(S)))>>((S)+1);\
-    }\
-}
-
-BGR2UV(uint32_t, bgr32ToUV,16, 0, 0, 0, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
-BGR2UV(uint32_t,bgr321ToUV,16, 0, 0, 8, 0xFF0000, 0xFF00,   0x00FF, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
-BGR2UV(uint32_t, rgb32ToUV, 0, 0,16, 0,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
-BGR2UV(uint32_t,rgb321ToUV, 0, 0,16, 8,   0x00FF, 0xFF00, 0xFF0000, RU<< 8, GU   , BU<< 8, RV<< 8, GV   , BV<< 8, RGB2YUV_SHIFT+8)
-BGR2UV(uint16_t, bgr16ToUV, 0, 0, 0, 0,   0x001F, 0x07E0,   0xF800, RU<<11, GU<<5, BU    , RV<<11, GV<<5, BV    , RGB2YUV_SHIFT+8)
-BGR2UV(uint16_t, bgr15ToUV, 0, 0, 0, 0,   0x001F, 0x03E0,   0x7C00, RU<<10, GU<<5, BU    , RV<<10, GV<<5, BV    , RGB2YUV_SHIFT+7)
-BGR2UV(uint16_t, rgb16ToUV, 0, 0, 0, 0,   0xF800, 0x07E0,   0x001F, RU    , GU<<5, BU<<11, RV    , GV<<5, BV<<11, RGB2YUV_SHIFT+8)
-BGR2UV(uint16_t, rgb15ToUV, 0, 0, 0, 0,   0x7C00, 0x03E0,   0x001F, RU    , GU<<5, BU<<10, RV    , GV<<5, BV<<10, RGB2YUV_SHIFT+7)
-
 static void palToY_c(uint8_t *dst, const uint8_t *src, int width, uint32_t *pal)
 {
     int i;
@@ -1161,73 +1593,6 @@ static void monoblack2Y_c(uint8_t *dst, const uint8_t *src,
     }
 }
 
-static void yuv2yuv1_c(SwsContext *c, const int16_t *lumSrc,
-                       const int16_t *chrUSrc, const int16_t *chrVSrc,
-                       const int16_t *alpSrc,
-                       uint8_t *dest, uint8_t *uDest, uint8_t *vDest,
-                       uint8_t *aDest, int dstW, int chrDstW)
-{
-    int i;
-    for (i=0; i<dstW; i++) {
-        int val= (lumSrc[i]+64)>>7;
-        dest[i]= av_clip_uint8(val);
-    }
-
-    if (uDest)
-        for (i=0; i<chrDstW; i++) {
-            int u=(chrUSrc[i]+64)>>7;
-            int v=(chrVSrc[i]+64)>>7;
-            uDest[i]= av_clip_uint8(u);
-            vDest[i]= av_clip_uint8(v);
-        }
-
-    if (CONFIG_SWSCALE_ALPHA && aDest)
-        for (i=0; i<dstW; i++) {
-            int val= (alpSrc[i]+64)>>7;
-            aDest[i]= av_clip_uint8(val);
-        }
-}
-
-/**
- * vertical bilinear scale YV12 to RGB
- */
-static void yuv2packed2_c(SwsContext *c, const uint16_t *buf0,
-                          const uint16_t *buf1, const uint16_t *ubuf0,
-                          const uint16_t *ubuf1, const uint16_t *vbuf0,
-                          const uint16_t *vbuf1, const uint16_t *abuf0,
-                          const uint16_t *abuf1, uint8_t *dest, int dstW,
-                          int yalpha, int uvalpha, int y)
-{
-    int  yalpha1=4095- yalpha;
-    int uvalpha1=4095-uvalpha;
-    int i;
-
-    YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB2_C, YSCALE_YUV_2_PACKED2_C(void,0), YSCALE_YUV_2_GRAY16_2_C, YSCALE_YUV_2_MONO2_C)
-}
-
-/**
- * YV12 to RGB without scaling or interpolating
- */
-static void yuv2packed1_c(SwsContext *c, const uint16_t *buf0,
-                          const uint16_t *ubuf0, const uint16_t *ubuf1,
-                          const uint16_t *vbuf0, const uint16_t *vbuf1,
-                          const uint16_t *abuf0, uint8_t *dest, int dstW,
-                          int uvalpha, enum PixelFormat dstFormat,
-                          int flags, int y)
-{
-    const int yalpha1=0;
-    int i;
-
-    const uint16_t *buf1= buf0; //FIXME needed for RGB1/BGR1
-    const int yalpha= 4096; //FIXME ...
-
-    if (uvalpha < 2048) {
-        YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1_C, YSCALE_YUV_2_PACKED1_C(void,0), YSCALE_YUV_2_GRAY16_1_C, YSCALE_YUV_2_MONO2_C)
-    } else {
-        YSCALE_YUV_2_ANYRGB_C(YSCALE_YUV_2_RGB1B_C, YSCALE_YUV_2_PACKED1B_C(void,0), YSCALE_YUV_2_GRAY16_1_C, YSCALE_YUV_2_MONO2_C)
-    }
-}
-
 //FIXME yuy2* can read up to 7 samples too much
 
 static void yuy2ToY_c(uint8_t *dst, const uint8_t *src, int width,
@@ -1314,34 +1679,54 @@ static void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV,
     nvXXtoUV_c(dstV, dstU, src1, width);
 }
 
+#define input_pixel(pos) (isBE(origin) ? AV_RB16(pos) : AV_RL16(pos))
+
 // FIXME Maybe dither instead.
-#define YUV_NBPS(depth, endianness, rfunc) \
-static void endianness ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
-                                          const uint8_t *_srcU, const uint8_t *_srcV, \
-                                          int width, uint32_t *unused) \
+static av_always_inline void
+yuv9_OR_10ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
+                          const uint8_t *_srcU, const uint8_t *_srcV,
+                          int width, enum PixelFormat origin, int depth)
+{
+    int i;
+    const uint16_t *srcU = (const uint16_t *) _srcU;
+    const uint16_t *srcV = (const uint16_t *) _srcV;
+
+    for (i = 0; i < width; i++) {
+        dstU[i] = input_pixel(&srcU[i]) >> (depth - 8);
+        dstV[i] = input_pixel(&srcV[i]) >> (depth - 8);
+    }
+}
+
+static av_always_inline void
+yuv9_or_10ToY_c_template(uint8_t *dstY, const uint8_t *_srcY,
+                         int width, enum PixelFormat origin, int depth)
+{
+    int i;
+    const uint16_t *srcY = (const uint16_t*)_srcY;
+
+    for (i = 0; i < width; i++)
+        dstY[i] = input_pixel(&srcY[i]) >> (depth - 8);
+}
+
+#undef input_pixel
+
+#define YUV_NBPS(depth, BE_LE, origin) \
+static void BE_LE ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
+                                     const uint8_t *srcU, const uint8_t *srcV, \
+                                     int width, uint32_t *unused) \
 { \
-    int i; \
-    const uint16_t *srcU = (const uint16_t*)_srcU; \
-    const uint16_t *srcV = (const uint16_t*)_srcV; \
-    for (i = 0; i < width; i++) { \
-        dstU[i] = rfunc(&srcU[i])>>(depth-8); \
-        dstV[i] = rfunc(&srcV[i])>>(depth-8); \
-    } \
+    yuv9_OR_10ToUV_c_template(dstU, dstV, srcU, srcV, width, origin, depth); \
 } \
-\
-static void endianness ## depth ## ToY_c(uint8_t *dstY, const uint8_t *_srcY, \
-                                         int width, uint32_t *unused) \
+static void BE_LE ## depth ## ToY_c(uint8_t *dstY, const uint8_t *srcY, \
+                                    int width, uint32_t *unused) \
 { \
-    int i; \
-    const uint16_t *srcY = (const uint16_t*)_srcY; \
-    for (i = 0; i < width; i++) \
-        dstY[i] = rfunc(&srcY[i])>>(depth-8); \
-} \
+    yuv9_or_10ToY_c_template(dstY, srcY, width, origin, depth); \
+}
 
-YUV_NBPS( 9, LE, AV_RL16)
-YUV_NBPS( 9, BE, AV_RB16)
-YUV_NBPS(10, LE, AV_RL16)
-YUV_NBPS(10, BE, AV_RB16)
+YUV_NBPS( 9, LE, PIX_FMT_YUV420P9LE);
+YUV_NBPS( 9, BE, PIX_FMT_YUV420P9BE);
+YUV_NBPS(10, LE, PIX_FMT_YUV420P10LE);
+YUV_NBPS(10, BE, PIX_FMT_YUV420P10BE);
 
 static void bgr24ToY_c(uint8_t *dst, const uint8_t *src,
                        int width, uint32_t *unused)
@@ -1429,10 +1814,8 @@ static void rgb24ToUV_half_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
     }
 }
 
-
 // bilinear / bicubic scaling
 static void hScale_c(int16_t *dst, int dstW, const uint8_t *src,
-                     int srcW, int xInc,
                      const int16_t *filter, const int16_t *filterPos,
                      int filterSize)
 {
@@ -1452,7 +1835,7 @@ static void hScale_c(int16_t *dst, int dstW, const uint8_t *src,
 
 //FIXME all pal and rgb srcFormats could do this convertion as well
 //FIXME all scalers more complex than bilinear could do half of this transform
-static void chrRangeToJpeg_c(uint16_t *dstU, uint16_t *dstV, int width)
+static void chrRangeToJpeg_c(int16_t *dstU, int16_t *dstV, int width)
 {
     int i;
     for (i = 0; i < width; i++) {
@@ -1460,7 +1843,7 @@ static void chrRangeToJpeg_c(uint16_t *dstU, uint16_t *dstV, int width)
         dstV[i] = (FFMIN(dstV[i],30775)*4663 - 9289992)>>12; //-264
     }
 }
-static void chrRangeFromJpeg_c(uint16_t *dstU, uint16_t *dstV, int width)
+static void chrRangeFromJpeg_c(int16_t *dstU, int16_t *dstV, int width)
 {
     int i;
     for (i = 0; i < width; i++) {
@@ -1468,13 +1851,13 @@ static void chrRangeFromJpeg_c(uint16_t *dstU, uint16_t *dstV, int width)
         dstV[i] = (dstV[i]*1799 + 4081085)>>11; //1469
     }
 }
-static void lumRangeToJpeg_c(uint16_t *dst, int width)
+static void lumRangeToJpeg_c(int16_t *dst, int width)
 {
     int i;
     for (i = 0; i < width; i++)
         dst[i] = (FFMIN(dst[i],30189)*19077 - 39057361)>>14;
 }
-static void lumRangeFromJpeg_c(uint16_t *dst, int width)
+static void lumRangeFromJpeg_c(int16_t *dst, int width)
 {
     int i;
     for (i = 0; i < width; i++)
@@ -1495,15 +1878,15 @@ static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
 }
 
 // *** horizontal scale Y line to temp buffer
-static inline void hyscale(SwsContext *c, uint16_t *dst, int dstWidth,
-                           const uint8_t *src, int srcW, int xInc,
-                           const int16_t *hLumFilter,
-                           const int16_t *hLumFilterPos, int hLumFilterSize,
-                           uint8_t *formatConvBuffer,
-                           uint32_t *pal, int isAlpha)
+static av_always_inline void hyscale(SwsContext *c, uint16_t *dst, int dstWidth,
+                                     const uint8_t *src, int srcW, int xInc,
+                                     const int16_t *hLumFilter,
+                                     const int16_t *hLumFilterPos, int hLumFilterSize,
+                                     uint8_t *formatConvBuffer,
+                                     uint32_t *pal, int isAlpha)
 {
     void (*toYV12)(uint8_t *, const uint8_t *, int, uint32_t *) = isAlpha ? c->alpToYV12 : c->lumToYV12;
-    void (*convertRange)(uint16_t *, int) = isAlpha ? NULL : c->lumConvertRange;
+    void (*convertRange)(int16_t *, int) = isAlpha ? NULL : c->lumConvertRange;
 
     if (toYV12) {
         toYV12(formatConvBuffer, src, srcW, pal);
@@ -1511,7 +1894,7 @@ static inline void hyscale(SwsContext *c, uint16_t *dst, int dstWidth,
     }
 
     if (!c->hyscale_fast) {
-        c->hScale(dst, dstWidth, src, srcW, xInc, hLumFilter, hLumFilterPos, hLumFilterSize);
+        c->hScale(dst, dstWidth, src, hLumFilter, hLumFilterPos, hLumFilterSize);
     } else { // fast bilinear upscale / crap downscale
         c->hyscale_fast(c, dst, dstWidth, src, srcW, xInc);
     }
@@ -1535,11 +1918,11 @@ static void hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
     }
 }
 
-static inline void hcscale(SwsContext *c, uint16_t *dst1, uint16_t *dst2, int dstWidth,
-                           const uint8_t *src1, const uint8_t *src2,
-                           int srcW, int xInc, const int16_t *hChrFilter,
-                           const int16_t *hChrFilterPos, int hChrFilterSize,
-                           uint8_t *formatConvBuffer, uint32_t *pal)
+static av_always_inline void hcscale(SwsContext *c, uint16_t *dst1, uint16_t *dst2, int dstWidth,
+                                     const uint8_t *src1, const uint8_t *src2,
+                                     int srcW, int xInc, const int16_t *hChrFilter,
+                                     const int16_t *hChrFilterPos, int hChrFilterSize,
+                                     uint8_t *formatConvBuffer, uint32_t *pal)
 {
     if (c->chrToYV12) {
         uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW, 16);
@@ -1549,8 +1932,8 @@ static inline void hcscale(SwsContext *c, uint16_t *dst1, uint16_t *dst2, int ds
     }
 
     if (!c->hcscale_fast) {
-        c->hScale(dst1, dstWidth, src1, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize);
-        c->hScale(dst2, dstWidth, src2, srcW, xInc, hChrFilter, hChrFilterPos, hChrFilterSize);
+        c->hScale(dst1, dstWidth, src1, hChrFilter, hChrFilterPos, hChrFilterSize);
+        c->hScale(dst2, dstWidth, src2, hChrFilter, hChrFilterPos, hChrFilterSize);
     } else { // fast bilinear upscale / crap downscale
         c->hcscale_fast(c, dst1, dst2, dstWidth, src1, src2, srcW, xInc);
     }
@@ -1559,6 +1942,243 @@ static inline void hcscale(SwsContext *c, uint16_t *dst1, uint16_t *dst2, int ds
         c->chrConvertRange(dst1, dst2, dstWidth);
 }
 
+static av_always_inline void
+find_c_packed_planar_out_funcs(SwsContext *c,
+                               yuv2planar1_fn *yuv2yuv1,    yuv2planarX_fn *yuv2yuvX,
+                               yuv2packed1_fn *yuv2packed1, yuv2packed2_fn *yuv2packed2,
+                               yuv2packedX_fn *yuv2packedX)
+{
+    enum PixelFormat dstFormat = c->dstFormat;
+
+    if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21) {
+        *yuv2yuvX     = yuv2nv12X_c;
+    } else if (is16BPS(dstFormat)) {
+        *yuv2yuvX     = isBE(dstFormat) ? yuv2yuvX16BE_c  : yuv2yuvX16LE_c;
+    } else if (is9_OR_10BPS(dstFormat)) {
+        if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
+            *yuv2yuvX = isBE(dstFormat) ? yuv2yuvX9BE_c :  yuv2yuvX9LE_c;
+        } else {
+            *yuv2yuvX = isBE(dstFormat) ? yuv2yuvX10BE_c : yuv2yuvX10LE_c;
+        }
+    } else {
+        *yuv2yuv1     = yuv2yuv1_c;
+        *yuv2yuvX     = yuv2yuvX_c;
+    }
+    if(c->flags & SWS_FULL_CHR_H_INT) {
+        switch (dstFormat) {
+            case PIX_FMT_RGBA:
+#if CONFIG_SMALL
+                *yuv2packedX = yuv2rgba32_full_X_c;
+#else
+#if CONFIG_SWSCALE_ALPHA
+                if (c->alpPixBuf) {
+                    *yuv2packedX = yuv2rgba32_full_X_c;
+                } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+                {
+                    *yuv2packedX = yuv2rgbx32_full_X_c;
+                }
+#endif /* !CONFIG_SMALL */
+                break;
+            case PIX_FMT_ARGB:
+#if CONFIG_SMALL
+                *yuv2packedX = yuv2argb32_full_X_c;
+#else
+#if CONFIG_SWSCALE_ALPHA
+                if (c->alpPixBuf) {
+                    *yuv2packedX = yuv2argb32_full_X_c;
+                } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+                {
+                    *yuv2packedX = yuv2xrgb32_full_X_c;
+                }
+#endif /* !CONFIG_SMALL */
+                break;
+            case PIX_FMT_BGRA:
+#if CONFIG_SMALL
+                *yuv2packedX = yuv2bgra32_full_X_c;
+#else
+#if CONFIG_SWSCALE_ALPHA
+                if (c->alpPixBuf) {
+                    *yuv2packedX = yuv2bgra32_full_X_c;
+                } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+                {
+                    *yuv2packedX = yuv2bgrx32_full_X_c;
+                }
+#endif /* !CONFIG_SMALL */
+                break;
+            case PIX_FMT_ABGR:
+#if CONFIG_SMALL
+                *yuv2packedX = yuv2abgr32_full_X_c;
+#else
+#if CONFIG_SWSCALE_ALPHA
+                if (c->alpPixBuf) {
+                    *yuv2packedX = yuv2abgr32_full_X_c;
+                } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+                {
+                    *yuv2packedX = yuv2xbgr32_full_X_c;
+                }
+#endif /* !CONFIG_SMALL */
+                break;
+            case PIX_FMT_RGB24:
+            *yuv2packedX = yuv2rgb24_full_X_c;
+            break;
+        case PIX_FMT_BGR24:
+            *yuv2packedX = yuv2bgr24_full_X_c;
+            break;
+        }
+    } else {
+        switch (dstFormat) {
+        case PIX_FMT_GRAY16BE:
+            *yuv2packed1 = yuv2gray16BE_1_c;
+            *yuv2packed2 = yuv2gray16BE_2_c;
+            *yuv2packedX = yuv2gray16BE_X_c;
+            break;
+        case PIX_FMT_GRAY16LE:
+            *yuv2packed1 = yuv2gray16LE_1_c;
+            *yuv2packed2 = yuv2gray16LE_2_c;
+            *yuv2packedX = yuv2gray16LE_X_c;
+            break;
+        case PIX_FMT_MONOWHITE:
+            *yuv2packed1 = yuv2monowhite_1_c;
+            *yuv2packed2 = yuv2monowhite_2_c;
+            *yuv2packedX = yuv2monowhite_X_c;
+            break;
+        case PIX_FMT_MONOBLACK:
+            *yuv2packed1 = yuv2monoblack_1_c;
+            *yuv2packed2 = yuv2monoblack_2_c;
+            *yuv2packedX = yuv2monoblack_X_c;
+            break;
+        case PIX_FMT_YUYV422:
+            *yuv2packed1 = yuv2yuyv422_1_c;
+            *yuv2packed2 = yuv2yuyv422_2_c;
+            *yuv2packedX = yuv2yuyv422_X_c;
+            break;
+        case PIX_FMT_UYVY422:
+            *yuv2packed1 = yuv2uyvy422_1_c;
+            *yuv2packed2 = yuv2uyvy422_2_c;
+            *yuv2packedX = yuv2uyvy422_X_c;
+            break;
+        case PIX_FMT_RGB48LE:
+            //*yuv2packed1 = yuv2rgb48le_1_c;
+            //*yuv2packed2 = yuv2rgb48le_2_c;
+            //*yuv2packedX = yuv2rgb48le_X_c;
+            //break;
+        case PIX_FMT_RGB48BE:
+            *yuv2packed1 = yuv2rgb48be_1_c;
+            *yuv2packed2 = yuv2rgb48be_2_c;
+            *yuv2packedX = yuv2rgb48be_X_c;
+            break;
+        case PIX_FMT_BGR48LE:
+            //*yuv2packed1 = yuv2bgr48le_1_c;
+            //*yuv2packed2 = yuv2bgr48le_2_c;
+            //*yuv2packedX = yuv2bgr48le_X_c;
+            //break;
+        case PIX_FMT_BGR48BE:
+            *yuv2packed1 = yuv2bgr48be_1_c;
+            *yuv2packed2 = yuv2bgr48be_2_c;
+            *yuv2packedX = yuv2bgr48be_X_c;
+            break;
+        case PIX_FMT_RGB32:
+        case PIX_FMT_BGR32:
+#if CONFIG_SMALL
+            *yuv2packed1 = yuv2rgb32_1_c;
+            *yuv2packed2 = yuv2rgb32_2_c;
+            *yuv2packedX = yuv2rgb32_X_c;
+#else
+#if CONFIG_SWSCALE_ALPHA
+                if (c->alpPixBuf) {
+                    *yuv2packed1 = yuv2rgba32_1_c;
+                    *yuv2packed2 = yuv2rgba32_2_c;
+                    *yuv2packedX = yuv2rgba32_X_c;
+                } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+                {
+                    *yuv2packed1 = yuv2rgbx32_1_c;
+                    *yuv2packed2 = yuv2rgbx32_2_c;
+                    *yuv2packedX = yuv2rgbx32_X_c;
+                }
+#endif /* !CONFIG_SMALL */
+            break;
+        case PIX_FMT_RGB32_1:
+        case PIX_FMT_BGR32_1:
+#if CONFIG_SMALL
+                *yuv2packed1 = yuv2rgb32_1_1_c;
+                *yuv2packed2 = yuv2rgb32_1_2_c;
+                *yuv2packedX = yuv2rgb32_1_X_c;
+#else
+#if CONFIG_SWSCALE_ALPHA
+                if (c->alpPixBuf) {
+                    *yuv2packed1 = yuv2rgba32_1_1_c;
+                    *yuv2packed2 = yuv2rgba32_1_2_c;
+                    *yuv2packedX = yuv2rgba32_1_X_c;
+                } else
+#endif /* CONFIG_SWSCALE_ALPHA */
+                {
+                    *yuv2packed1 = yuv2rgbx32_1_1_c;
+                    *yuv2packed2 = yuv2rgbx32_1_2_c;
+                    *yuv2packedX = yuv2rgbx32_1_X_c;
+                }
+#endif /* !CONFIG_SMALL */
+                break;
+        case PIX_FMT_RGB24:
+            *yuv2packed1 = yuv2rgb24_1_c;
+            *yuv2packed2 = yuv2rgb24_2_c;
+            *yuv2packedX = yuv2rgb24_X_c;
+            break;
+        case PIX_FMT_BGR24:
+            *yuv2packed1 = yuv2bgr24_1_c;
+            *yuv2packed2 = yuv2bgr24_2_c;
+            *yuv2packedX = yuv2bgr24_X_c;
+            break;
+        case PIX_FMT_RGB565LE:
+        case PIX_FMT_RGB565BE:
+        case PIX_FMT_BGR565LE:
+        case PIX_FMT_BGR565BE:
+            *yuv2packed1 = yuv2rgb16_1_c;
+            *yuv2packed2 = yuv2rgb16_2_c;
+            *yuv2packedX = yuv2rgb16_X_c;
+            break;
+        case PIX_FMT_RGB555LE:
+        case PIX_FMT_RGB555BE:
+        case PIX_FMT_BGR555LE:
+        case PIX_FMT_BGR555BE:
+            *yuv2packed1 = yuv2rgb15_1_c;
+            *yuv2packed2 = yuv2rgb15_2_c;
+            *yuv2packedX = yuv2rgb15_X_c;
+            break;
+        case PIX_FMT_RGB444LE:
+        case PIX_FMT_RGB444BE:
+        case PIX_FMT_BGR444LE:
+        case PIX_FMT_BGR444BE:
+            *yuv2packed1 = yuv2rgb12_1_c;
+            *yuv2packed2 = yuv2rgb12_2_c;
+            *yuv2packedX = yuv2rgb12_X_c;
+            break;
+        case PIX_FMT_RGB8:
+        case PIX_FMT_BGR8:
+            *yuv2packed1 = yuv2rgb8_1_c;
+            *yuv2packed2 = yuv2rgb8_2_c;
+            *yuv2packedX = yuv2rgb8_X_c;
+            break;
+        case PIX_FMT_RGB4:
+        case PIX_FMT_BGR4:
+            *yuv2packed1 = yuv2rgb4_1_c;
+            *yuv2packed2 = yuv2rgb4_2_c;
+            *yuv2packedX = yuv2rgb4_X_c;
+            break;
+        case PIX_FMT_RGB4_BYTE:
+        case PIX_FMT_BGR4_BYTE:
+            *yuv2packed1 = yuv2rgb4b_1_c;
+            *yuv2packed2 = yuv2rgb4b_2_c;
+            *yuv2packedX = yuv2rgb4b_X_c;
+            break;
+        }
+    }
+}
+
 #define DEBUG_SWSCALE_BUFFERS 0
 #define DEBUG_BUFFERS(...) if (DEBUG_SWSCALE_BUFFERS) av_log(c, AV_LOG_DEBUG, __VA_ARGS__)
 
@@ -1602,6 +2222,11 @@ static int swScale(SwsContext *c, const uint8_t* src[],
     const int chrSrcSliceH= -((-srcSliceH) >> c->chrSrcVSubSample);
     int lastDstY;
     uint32_t *pal=c->pal_yuv;
+    yuv2planar1_fn yuv2yuv1 = c->yuv2yuv1;
+    yuv2planarX_fn yuv2yuvX = c->yuv2yuvX;
+    yuv2packed1_fn yuv2packed1 = c->yuv2packed1;
+    yuv2packed2_fn yuv2packed2 = c->yuv2packed2;
+    yuv2packedX_fn yuv2packedX = c->yuv2packedX;
 
     /* vars which will change and which we need to store back in the context */
     int dstY= c->dstY;
@@ -1738,19 +2363,19 @@ static int swScale(SwsContext *c, const uint8_t* src[],
 #if HAVE_MMX
         updateMMXDitherTables(c, dstY, lumBufIndex, chrBufIndex, lastInLumBuf, lastInChrBuf);
 #endif
-        if (dstY < dstH-2) {
+        if (dstY >= dstH-2) {
+            // hmm looks like we can't use MMX here without overwriting this array's tail
+            find_c_packed_planar_out_funcs(c, &yuv2yuv1, &yuv2yuvX,
+                                           &yuv2packed1, &yuv2packed2,
+                                           &yuv2packedX);
+        }
+
+        {
             const int16_t **lumSrcPtr= (const int16_t **) lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
             const int16_t **chrUSrcPtr= (const int16_t **) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
             const int16_t **chrVSrcPtr= (const int16_t **) chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
             const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
-            if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21) {
-                const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
-                if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi
-                c->yuv2nv12X(c,
-                             vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
-                             vChrFilter+chrDstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                             dest, uDest, dstW, chrDstW, dstFormat);
-            } else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12 like
+            if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12 like
                 const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
                 if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi
                 if (c->yuv2yuv1 && vLumFilterSize == 1 && vChrFilterSize == 1) { // unscaled YV12
@@ -1758,10 +2383,10 @@ static int swScale(SwsContext *c, const uint8_t* src[],
                     const int16_t *chrUBuf= chrUSrcPtr[0];
                     const int16_t *chrVBuf= chrVSrcPtr[0];
                     const int16_t *alpBuf= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? alpSrcPtr[0] : NULL;
-                    c->yuv2yuv1(c, lumBuf, chrUBuf, chrVBuf, alpBuf, dest,
+                    yuv2yuv1(c, lumBuf, chrUBuf, chrVBuf, alpBuf, dest,
                                 uDest, vDest, aDest, dstW, chrDstW);
                 } else { //General YV12
-                    c->yuv2yuvX(c,
+                    yuv2yuvX(c,
                                 vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                 vChrFilter+chrDstY*vChrFilterSize, chrUSrcPtr,
                                 chrVSrcPtr, vChrFilterSize,
@@ -1770,94 +2395,28 @@ static int swScale(SwsContext *c, const uint8_t* src[],
             } else {
                 assert(lumSrcPtr  + vLumFilterSize - 1 < lumPixBuf  + vLumBufSize*2);
                 assert(chrUSrcPtr + vChrFilterSize - 1 < chrUPixBuf + vChrBufSize*2);
-                if (vLumFilterSize == 1 && vChrFilterSize == 2) { //unscaled RGB
+                if (c->yuv2packed1 && vLumFilterSize == 1 && vChrFilterSize == 2) { //unscaled RGB
                     int chrAlpha= vChrFilter[2*dstY+1];
-                    if(flags & SWS_FULL_CHR_H_INT) {
-                        yuv2rgbX_c_full(c, //FIXME write a packed1_full function
-                                        vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
-                                        vChrFilter+dstY*vChrFilterSize, chrUSrcPtr,
-                                        chrVSrcPtr, vChrFilterSize,
-                                        alpSrcPtr, dest, dstW, dstY);
-                    } else {
-                        c->yuv2packed1(c, *lumSrcPtr, *chrUSrcPtr, *(chrUSrcPtr+1),
-                                       *chrVSrcPtr, *(chrVSrcPtr+1),
-                                       alpPixBuf ? *alpSrcPtr : NULL,
-                                       dest, dstW, chrAlpha, dstFormat, flags, dstY);
-                    }
-                } else if (vLumFilterSize == 2 && vChrFilterSize == 2) { //bilinear upscale RGB
+                    yuv2packed1(c, *lumSrcPtr, *chrUSrcPtr, *(chrUSrcPtr+1),
+                                   *chrVSrcPtr, *(chrVSrcPtr+1),
+                                   alpPixBuf ? *alpSrcPtr : NULL,
+                                   dest, dstW, chrAlpha, dstFormat, flags, dstY);
+                } else if (c->yuv2packed2 && vLumFilterSize == 2 && vChrFilterSize == 2) { //bilinear upscale RGB
                     int lumAlpha= vLumFilter[2*dstY+1];
                     int chrAlpha= vChrFilter[2*dstY+1];
                     lumMmxFilter[2]=
                     lumMmxFilter[3]= vLumFilter[2*dstY   ]*0x10001;
                     chrMmxFilter[2]=
                     chrMmxFilter[3]= vChrFilter[2*chrDstY]*0x10001;
-                    if(flags & SWS_FULL_CHR_H_INT) {
-                        yuv2rgbX_c_full(c, //FIXME write a packed2_full function
-                                        vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
-                                        vChrFilter+dstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                                        alpSrcPtr, dest, dstW, dstY);
-                    } else {
-                        c->yuv2packed2(c, *lumSrcPtr, *(lumSrcPtr+1), *chrUSrcPtr, *(chrUSrcPtr+1),
-                                       *chrVSrcPtr, *(chrVSrcPtr+1),
-                                       alpPixBuf ? *alpSrcPtr : NULL, alpPixBuf ? *(alpSrcPtr+1) : NULL,
-                                       dest, dstW, lumAlpha, chrAlpha, dstY);
-                    }
+                    yuv2packed2(c, *lumSrcPtr, *(lumSrcPtr+1), *chrUSrcPtr, *(chrUSrcPtr+1),
+                                   *chrVSrcPtr, *(chrVSrcPtr+1),
+                                   alpPixBuf ? *alpSrcPtr : NULL, alpPixBuf ? *(alpSrcPtr+1) : NULL,
+                                   dest, dstW, lumAlpha, chrAlpha, dstY);
                 } else { //general RGB
-                    if(flags & SWS_FULL_CHR_H_INT) {
-                        yuv2rgbX_c_full(c,
-                                        vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
-                                        vChrFilter+dstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                                        alpSrcPtr, dest, dstW, dstY);
-                    } else {
-                        c->yuv2packedX(c,
-                                       vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
-                                       vChrFilter+dstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                                       alpSrcPtr, dest, dstW, dstY);
-                    }
-                }
-            }
-        } else { // hmm looks like we can't use MMX here without overwriting this array's tail
-            const int16_t **lumSrcPtr= (const int16_t **)lumPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize;
-            const int16_t **chrUSrcPtr= (const int16_t **)chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
-            const int16_t **chrVSrcPtr= (const int16_t **)chrVPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
-            const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **)alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
-            if (dstFormat == PIX_FMT_NV12 || dstFormat == PIX_FMT_NV21) {
-                const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
-                if (dstY&chrSkipMask) uDest= NULL; //FIXME split functions in lumi / chromi
-                yuv2nv12X_c(c, vLumFilter+dstY*vLumFilterSize,
-                            lumSrcPtr, vLumFilterSize,
-                            vChrFilter+chrDstY*vChrFilterSize,
-                            chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                            dest, uDest, dstW, chrDstW, dstFormat);
-            } else if (isPlanarYUV(dstFormat) || dstFormat==PIX_FMT_GRAY8) { //YV12
-                const int chrSkipMask= (1<<c->chrDstVSubSample)-1;
-                if ((dstY&chrSkipMask) || isGray(dstFormat)) uDest=vDest= NULL; //FIXME split functions in lumi / chromi
-                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
-                    yuv2yuvX16_c(c, vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
-                                 vChrFilter+chrDstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                                 alpSrcPtr, dest, uDest, vDest, aDest, dstW, chrDstW,
-                                 dstFormat);
-                } else {
-                    yuv2yuvX_c(c, vLumFilter+dstY*vLumFilterSize,
-                               lumSrcPtr, vLumFilterSize,
-                               vChrFilter+chrDstY*vChrFilterSize,
-                               chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                               alpSrcPtr, dest, uDest, vDest, aDest,
-                               dstW, chrDstW);
-                }
-            } else {
-                assert(lumSrcPtr + vLumFilterSize - 1 < lumPixBuf + vLumBufSize*2);
-                assert(chrUSrcPtr + vChrFilterSize - 1 < chrUPixBuf + vChrBufSize*2);
-                if(flags & SWS_FULL_CHR_H_INT) {
-                    yuv2rgbX_c_full(c,
-                                    vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
-                                    vChrFilter+dstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                                    alpSrcPtr, dest, dstW, dstY);
-                } else {
-                    yuv2packedX_c(c,
-                                  vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
-                                  vChrFilter+dstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
-                                  alpSrcPtr, dest, dstW, dstY);
+                    yuv2packedX(c,
+                                   vLumFilter+dstY*vLumFilterSize, lumSrcPtr, vLumFilterSize,
+                                   vChrFilter+dstY*vChrFilterSize, chrUSrcPtr, chrVSrcPtr, vChrFilterSize,
+                                   alpSrcPtr, dest, dstW, dstY);
                 }
             }
         }
@@ -1882,27 +2441,13 @@ static int swScale(SwsContext *c, const uint8_t* src[],
     return dstY - lastDstY;
 }
 
-static void sws_init_swScale_c(SwsContext *c)
+static av_cold void sws_init_swScale_c(SwsContext *c)
 {
-    enum PixelFormat srcFormat = c->srcFormat,
-                     dstFormat = c->dstFormat;
+    enum PixelFormat srcFormat = c->srcFormat;
 
-    c->yuv2nv12X    = yuv2nv12X_c;
-    if (is16BPS(dstFormat)) {
-        c->yuv2yuvX     = isBE(dstFormat) ? yuv2yuvX16BE_c  : yuv2yuvX16LE_c;
-    } else if (is9_OR_10BPS(dstFormat)) {
-        if (dstFormat == PIX_FMT_YUV420P9BE || dstFormat == PIX_FMT_YUV420P9LE) {
-            c->yuv2yuvX = isBE(dstFormat) ? yuv2yuvX9BE_c :  yuv2yuvX9LE_c;
-        } else {
-            c->yuv2yuvX = isBE(dstFormat) ? yuv2yuvX10BE_c : yuv2yuvX10LE_c;
-        }
-    } else {
-        c->yuv2yuv1     = yuv2yuv1_c;
-        c->yuv2yuvX     = yuv2yuvX_c;
-    }
-    c->yuv2packed1  = yuv2packed1_c;
-    c->yuv2packed2  = yuv2packed2_c;
-    c->yuv2packedX  = yuv2packedX_c;
+    find_c_packed_planar_out_funcs(c, &c->yuv2yuv1, &c->yuv2yuvX,
+                                   &c->yuv2packed1, &c->yuv2packed2,
+                                   &c->yuv2packedX);
 
     c->hScale       = hScale_c;
 
@@ -1922,9 +2467,15 @@ static void sws_init_swScale_c(SwsContext *c)
         case PIX_FMT_PAL8     :
         case PIX_FMT_BGR4_BYTE:
         case PIX_FMT_RGB4_BYTE: c->chrToYV12 = palToUV_c; break;
+        case PIX_FMT_YUV444P9BE:
         case PIX_FMT_YUV420P9BE: c->chrToYV12 = BE9ToUV_c; break;
+        case PIX_FMT_YUV444P9LE:
         case PIX_FMT_YUV420P9LE: c->chrToYV12 = LE9ToUV_c; break;
+        case PIX_FMT_YUV444P10BE:
+        case PIX_FMT_YUV422P10BE:
         case PIX_FMT_YUV420P10BE: c->chrToYV12 = BE10ToUV_c; break;
+        case PIX_FMT_YUV422P10LE:
+        case PIX_FMT_YUV444P10LE:
         case PIX_FMT_YUV420P10LE: c->chrToYV12 = LE10ToUV_c; break;
         case PIX_FMT_YUV420P16BE:
         case PIX_FMT_YUV422P16BE:
@@ -1935,46 +2486,60 @@ static void sws_init_swScale_c(SwsContext *c)
     }
     if (c->chrSrcHSubSample) {
         switch(srcFormat) {
-        case PIX_FMT_RGB48BE: c->chrToYV12 = rgb48BEToUV_half_c; break;
-        case PIX_FMT_RGB48LE: c->chrToYV12 = rgb48LEToUV_half_c; break;
-        case PIX_FMT_BGR48BE: c->chrToYV12 = bgr48BEToUV_half_c; break;
-        case PIX_FMT_BGR48LE: c->chrToYV12 = bgr48LEToUV_half_c; break;
-        case PIX_FMT_RGB32  : c->chrToYV12 = bgr32ToUV_half_c;  break;
-        case PIX_FMT_RGB32_1: c->chrToYV12 = bgr321ToUV_half_c; break;
-        case PIX_FMT_BGR24  : c->chrToYV12 = bgr24ToUV_half_c; break;
-        case PIX_FMT_BGR565 : c->chrToYV12 = bgr16ToUV_half_c; break;
-        case PIX_FMT_BGR555 : c->chrToYV12 = bgr15ToUV_half_c; break;
-        case PIX_FMT_BGR32  : c->chrToYV12 = rgb32ToUV_half_c;  break;
-        case PIX_FMT_BGR32_1: c->chrToYV12 = rgb321ToUV_half_c; break;
-        case PIX_FMT_RGB24  : c->chrToYV12 = rgb24ToUV_half_c; break;
-        case PIX_FMT_RGB565 : c->chrToYV12 = rgb16ToUV_half_c; break;
-        case PIX_FMT_RGB555 : c->chrToYV12 = rgb15ToUV_half_c; break;
+        case PIX_FMT_RGB48BE : c->chrToYV12 = rgb48BEToUV_half_c; break;
+        case PIX_FMT_RGB48LE : c->chrToYV12 = rgb48LEToUV_half_c; break;
+        case PIX_FMT_BGR48BE : c->chrToYV12 = bgr48BEToUV_half_c; break;
+        case PIX_FMT_BGR48LE : c->chrToYV12 = bgr48LEToUV_half_c; break;
+        case PIX_FMT_RGB32   : c->chrToYV12 = bgr32ToUV_half_c;   break;
+        case PIX_FMT_RGB32_1 : c->chrToYV12 = bgr321ToUV_half_c;  break;
+        case PIX_FMT_BGR24   : c->chrToYV12 = bgr24ToUV_half_c;   break;
+        case PIX_FMT_BGR565LE: c->chrToYV12 = bgr16leToUV_half_c; break;
+        case PIX_FMT_BGR565BE: c->chrToYV12 = bgr16beToUV_half_c; break;
+        case PIX_FMT_BGR555LE: c->chrToYV12 = bgr15leToUV_half_c; break;
+        case PIX_FMT_BGR555BE: c->chrToYV12 = bgr15beToUV_half_c; break;
+        case PIX_FMT_BGR32   : c->chrToYV12 = rgb32ToUV_half_c;   break;
+        case PIX_FMT_BGR32_1 : c->chrToYV12 = rgb321ToUV_half_c;  break;
+        case PIX_FMT_RGB24   : c->chrToYV12 = rgb24ToUV_half_c;   break;
+        case PIX_FMT_RGB565LE: c->chrToYV12 = rgb16leToUV_half_c; break;
+        case PIX_FMT_RGB565BE: c->chrToYV12 = rgb16beToUV_half_c; break;
+        case PIX_FMT_RGB555LE: c->chrToYV12 = rgb15leToUV_half_c; break;
+        case PIX_FMT_RGB555BE: c->chrToYV12 = rgb15beToUV_half_c; break;
         }
     } else {
         switch(srcFormat) {
-        case PIX_FMT_RGB48BE: c->chrToYV12 = rgb48BEToUV_c; break;
-        case PIX_FMT_RGB48LE: c->chrToYV12 = rgb48LEToUV_c; break;
-        case PIX_FMT_BGR48BE: c->chrToYV12 = bgr48BEToUV_c; break;
-        case PIX_FMT_BGR48LE: c->chrToYV12 = bgr48LEToUV_c; break;
-        case PIX_FMT_RGB32  : c->chrToYV12 = bgr32ToUV_c;  break;
-        case PIX_FMT_RGB32_1: c->chrToYV12 = bgr321ToUV_c; break;
-        case PIX_FMT_BGR24  : c->chrToYV12 = bgr24ToUV_c; break;
-        case PIX_FMT_BGR565 : c->chrToYV12 = bgr16ToUV_c; break;
-        case PIX_FMT_BGR555 : c->chrToYV12 = bgr15ToUV_c; break;
-        case PIX_FMT_BGR32  : c->chrToYV12 = rgb32ToUV_c;  break;
-        case PIX_FMT_BGR32_1: c->chrToYV12 = rgb321ToUV_c; break;
-        case PIX_FMT_RGB24  : c->chrToYV12 = rgb24ToUV_c; break;
-        case PIX_FMT_RGB565 : c->chrToYV12 = rgb16ToUV_c; break;
-        case PIX_FMT_RGB555 : c->chrToYV12 = rgb15ToUV_c; break;
+        case PIX_FMT_RGB48BE : c->chrToYV12 = rgb48BEToUV_c; break;
+        case PIX_FMT_RGB48LE : c->chrToYV12 = rgb48LEToUV_c; break;
+        case PIX_FMT_BGR48BE : c->chrToYV12 = bgr48BEToUV_c; break;
+        case PIX_FMT_BGR48LE : c->chrToYV12 = bgr48LEToUV_c; break;
+        case PIX_FMT_RGB32   : c->chrToYV12 = bgr32ToUV_c;   break;
+        case PIX_FMT_RGB32_1 : c->chrToYV12 = bgr321ToUV_c;  break;
+        case PIX_FMT_BGR24   : c->chrToYV12 = bgr24ToUV_c;   break;
+        case PIX_FMT_BGR565LE: c->chrToYV12 = bgr16leToUV_c; break;
+        case PIX_FMT_BGR565BE: c->chrToYV12 = bgr16beToUV_c; break;
+        case PIX_FMT_BGR555LE: c->chrToYV12 = bgr15leToUV_c; break;
+        case PIX_FMT_BGR555BE: c->chrToYV12 = bgr15beToUV_c; break;
+        case PIX_FMT_BGR32   : c->chrToYV12 = rgb32ToUV_c;   break;
+        case PIX_FMT_BGR32_1 : c->chrToYV12 = rgb321ToUV_c;  break;
+        case PIX_FMT_RGB24   : c->chrToYV12 = rgb24ToUV_c;   break;
+        case PIX_FMT_RGB565LE: c->chrToYV12 = rgb16leToUV_c; break;
+        case PIX_FMT_RGB565BE: c->chrToYV12 = rgb16beToUV_c; break;
+        case PIX_FMT_RGB555LE: c->chrToYV12 = rgb15leToUV_c; break;
+        case PIX_FMT_RGB555BE: c->chrToYV12 = rgb15beToUV_c; break;
         }
     }
 
     c->lumToYV12 = NULL;
     c->alpToYV12 = NULL;
     switch (srcFormat) {
+    case PIX_FMT_YUV444P9BE:
     case PIX_FMT_YUV420P9BE: c->lumToYV12 = BE9ToY_c; break;
+    case PIX_FMT_YUV444P9LE:
     case PIX_FMT_YUV420P9LE: c->lumToYV12 = LE9ToY_c; break;
+    case PIX_FMT_YUV444P10BE:
+    case PIX_FMT_YUV422P10BE:
     case PIX_FMT_YUV420P10BE: c->lumToYV12 = BE10ToY_c; break;
+    case PIX_FMT_YUV444P10LE:
+    case PIX_FMT_YUV422P10LE:
     case PIX_FMT_YUV420P10LE: c->lumToYV12 = LE10ToY_c; break;
     case PIX_FMT_YUYV422  :
     case PIX_FMT_YUV420P16BE:
@@ -1986,13 +2551,17 @@ static void sws_init_swScale_c(SwsContext *c)
     case PIX_FMT_YUV420P16LE:
     case PIX_FMT_YUV422P16LE:
     case PIX_FMT_YUV444P16LE:
-    case PIX_FMT_GRAY16LE : c->lumToYV12 = uyvyToY_c; break;
-    case PIX_FMT_BGR24    : c->lumToYV12 = bgr24ToY_c; break;
-    case PIX_FMT_BGR565   : c->lumToYV12 = bgr16ToY_c; break;
-    case PIX_FMT_BGR555   : c->lumToYV12 = bgr15ToY_c; break;
-    case PIX_FMT_RGB24    : c->lumToYV12 = rgb24ToY_c; break;
-    case PIX_FMT_RGB565   : c->lumToYV12 = rgb16ToY_c; break;
-    case PIX_FMT_RGB555   : c->lumToYV12 = rgb15ToY_c; break;
+    case PIX_FMT_GRAY16LE : c->lumToYV12 = uyvyToY_c;    break;
+    case PIX_FMT_BGR24    : c->lumToYV12 = bgr24ToY_c;   break;
+    case PIX_FMT_BGR565LE : c->lumToYV12 = bgr16leToY_c; break;
+    case PIX_FMT_BGR565BE : c->lumToYV12 = bgr16beToY_c; break;
+    case PIX_FMT_BGR555LE : c->lumToYV12 = bgr15leToY_c; break;
+    case PIX_FMT_BGR555BE : c->lumToYV12 = bgr15beToY_c; break;
+    case PIX_FMT_RGB24    : c->lumToYV12 = rgb24ToY_c;   break;
+    case PIX_FMT_RGB565LE : c->lumToYV12 = rgb16leToY_c; break;
+    case PIX_FMT_RGB565BE : c->lumToYV12 = rgb16beToY_c; break;
+    case PIX_FMT_RGB555LE : c->lumToYV12 = rgb15leToY_c; break;
+    case PIX_FMT_RGB555BE : c->lumToYV12 = rgb15beToY_c; break;
     case PIX_FMT_RGB8     :
     case PIX_FMT_BGR8     :
     case PIX_FMT_PAL8     :