[FFmpeg-cvslog] swscale: implement >8bit scaling support.

Ronald S. Bultje git at videolan.org
Thu Jun 30 04:44:01 CEST 2011


ffmpeg | branch: master | Ronald S. Bultje <rsbultje at gmail.com> | Wed Jun 29 09:39:43 2011 -0700| [ef1ee362b36893cd1cc1e9c7de378e9cb239b0e8] | committer: Ronald S. Bultje

swscale: implement >8bit scaling support.

This means that precision is retained when scaling between sample
formats with >8 bits per component (48bit RGB, 16bit grayscale,
9/10/16bit YUV).

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=ef1ee362b36893cd1cc1e9c7de378e9cb239b0e8
---

 libswscale/ppc/swscale_altivec.c  |    4 +-
 libswscale/swscale.c              |  694 ++++++++++++++++++++++++-------------
 libswscale/swscale_internal.h     |   10 +
 libswscale/utils.c                |    9 +-
 libswscale/x86/swscale_template.c |   73 +----
 tests/ref/lavfi/pixdesc           |   16 +-
 tests/ref/lavfi/pixfmts_copy      |   16 +-
 tests/ref/lavfi/pixfmts_crop      |   16 +-
 tests/ref/lavfi/pixfmts_hflip     |   16 +-
 tests/ref/lavfi/pixfmts_null      |   16 +-
 tests/ref/lavfi/pixfmts_scale     |   32 +-
 tests/ref/lavfi/pixfmts_vflip     |   16 +-
 12 files changed, 545 insertions(+), 373 deletions(-)

diff --git a/libswscale/ppc/swscale_altivec.c b/libswscale/ppc/swscale_altivec.c
index 396b00c..6fb3de0 100644
--- a/libswscale/ppc/swscale_altivec.c
+++ b/libswscale/ppc/swscale_altivec.c
@@ -406,7 +406,9 @@ void ff_sws_init_swScale_altivec(SwsContext *c)
     if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
         return;
 
-    c->hScale       = hScale_altivec_real;
+    if (c->scalingBpp == 8) {
+        c->hScale       = hScale_altivec_real;
+    }
     if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&
         dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21 &&
         !c->alpPixBuf) {
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index 64ba97e..987afff 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -195,18 +195,18 @@ DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 #endif
 
 static av_always_inline void
-yuv2yuvX16_c_template(const int16_t *lumFilter, const int16_t **lumSrc,
+yuv2yuvX16_c_template(const int16_t *lumFilter, const int32_t **lumSrc,
                       int lumFilterSize, const int16_t *chrFilter,
-                      const int16_t **chrUSrc, const int16_t **chrVSrc,
-                      int chrFilterSize, const int16_t **alpSrc,
+                      const int32_t **chrUSrc, const int32_t **chrVSrc,
+                      int chrFilterSize, const int32_t **alpSrc,
                       uint16_t *dest[4], int dstW, int chrDstW,
                       int big_endian, int output_bits)
 {
     //FIXME Optimize (just quickly written not optimized..)
     int i;
-    int shift = 11 + 16 - output_bits;
     uint16_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2],
              *aDest = CONFIG_SWSCALE_ALPHA ? dest[3] : NULL;
+    int shift = 15 + 16 - output_bits;
 
 #define output_pixel(pos, val) \
     if (big_endian) { \
@@ -223,7 +223,7 @@ yuv2yuvX16_c_template(const int16_t *lumFilter, const int16_t **lumSrc,
         } \
     }
     for (i = 0; i < dstW; i++) {
-        int val = 1 << (26-output_bits);
+        int val = 1 << (30-output_bits);
         int j;
 
         for (j = 0; j < lumFilterSize; j++)
@@ -234,8 +234,8 @@ yuv2yuvX16_c_template(const int16_t *lumFilter, const int16_t **lumSrc,
 
     if (uDest) {
         for (i = 0; i < chrDstW; i++) {
-            int u = 1 << (26-output_bits);
-            int v = 1 << (26-output_bits);
+            int u = 1 << (30-output_bits);
+            int v = 1 << (30-output_bits);
             int j;
 
             for (j = 0; j < chrFilterSize; j++) {
@@ -250,7 +250,7 @@ yuv2yuvX16_c_template(const int16_t *lumFilter, const int16_t **lumSrc,
 
     if (CONFIG_SWSCALE_ALPHA && aDest) {
         for (i = 0; i < dstW; i++) {
-            int val = 1 << (26-output_bits);
+            int val = 1 << (30-output_bits);
             int j;
 
             for (j = 0; j < lumFilterSize; j++)
@@ -264,12 +264,16 @@ yuv2yuvX16_c_template(const int16_t *lumFilter, const int16_t **lumSrc,
 
 #define yuv2NBPS(bits, BE_LE, is_be) \
 static void yuv2yuvX ## bits ## BE_LE ## _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, \
+                              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[4], int dstW, int chrDstW) \
 { \
+    const int32_t **lumSrc  = (const int32_t **) _lumSrc, \
+                  **chrUSrc = (const int32_t **) _chrUSrc, \
+                  **chrVSrc = (const int32_t **) _chrVSrc, \
+                  **alpSrc  = (const int32_t **) _alpSrc; \
     yuv2yuvX16_c_template(lumFilter, lumSrc, lumFilterSize, \
                           chrFilter, chrUSrc, chrVSrc, chrFilterSize, \
                           alpSrc, (uint16_t **) _dest, \
@@ -418,109 +422,121 @@ static void yuv2nv12X_c(SwsContext *c, const int16_t *lumFilter,
 
 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,
+                        const int32_t **lumSrc, int lumFilterSize,
+                        const int16_t *chrFilter, const int32_t **chrUSrc,
+                        const int32_t **chrVSrc, int chrFilterSize,
+                        const int32_t **alpSrc, uint16_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;
+        int Y1 = 1 << 14;
+        int Y2 = 1 << 14;
 
         for (j = 0; j < lumFilterSize; j++) {
-            Y1 += lumSrc[j][i2]   * lumFilter[j];
-            Y2 += lumSrc[j][i2+1] * lumFilter[j];
+            Y1 += lumSrc[j][i * 2]     * lumFilter[j];
+            Y2 += lumSrc[j][i * 2 + 1] * lumFilter[j];
         }
-        Y1 >>= 11;
-        Y2 >>= 11;
+        Y1 >>= 15;
+        Y2 >>= 15;
         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);
+        output_pixel(&dest[i * 2 + 0], Y1);
+        output_pixel(&dest[i * 2 + 1], Y2);
     }
 }
 
 static av_always_inline void
-yuv2gray16_2_c_template(SwsContext *c, const int16_t *buf[2],
-                        const int16_t *ubuf[2], const int16_t *vbuf[2],
-                        const int16_t *abuf[2], uint8_t *dest, int dstW,
+yuv2gray16_2_c_template(SwsContext *c, const int32_t *buf[2],
+                        const int32_t *ubuf[2], const int32_t *vbuf[2],
+                        const int32_t *abuf[2], uint16_t *dest, int dstW,
                         int yalpha, int uvalpha, int y,
                         enum PixelFormat target)
 {
     int  yalpha1 = 4095 - yalpha;
     int i;
-    const int16_t *buf0 = buf[0], *buf1 = buf[1];
+    const int32_t *buf0 = buf[0], *buf1 = buf[1];
 
     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;
+        int Y1 = (buf0[i * 2    ] * yalpha1 + buf1[i * 2    ] * yalpha) >> 15;
+        int Y2 = (buf0[i * 2 + 1] * yalpha1 + buf1[i * 2 + 1] * yalpha) >> 15;
 
-        output_pixel(&dest[2 * i2 + 0], Y1);
-        output_pixel(&dest[2 * i2 + 2], Y2);
+        output_pixel(&dest[i * 2 + 0], Y1);
+        output_pixel(&dest[i * 2 + 1], Y2);
     }
 }
 
 static av_always_inline void
-yuv2gray16_1_c_template(SwsContext *c, const int16_t *buf0,
-                        const int16_t *ubuf[2], const int16_t *vbuf[2],
-                        const int16_t *abuf0, uint8_t *dest, int dstW,
+yuv2gray16_1_c_template(SwsContext *c, const int32_t *buf0,
+                        const int32_t *ubuf[2], const int32_t *vbuf[2],
+                        const int32_t *abuf0, uint16_t *dest, int dstW,
                         int uvalpha, 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;
+        int Y1 = buf0[i * 2    ] << 1;
+        int Y2 = buf0[i * 2 + 1] << 1;
 
-        output_pixel(&dest[2 * i2 + 0], Y1);
-        output_pixel(&dest[2 * i2 + 2], Y2);
+        output_pixel(&dest[i * 2 + 0], Y1);
+        output_pixel(&dest[i * 2 + 1], Y2);
     }
 }
 
 #undef output_pixel
 
-#define YUV2PACKEDWRAPPER(name, base, ext, fmt) \
+#define YUV2PACKED16WRAPPER(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, \
+                        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) \
 { \
+    const int32_t **lumSrc  = (const int32_t **) _lumSrc, \
+                  **chrUSrc = (const int32_t **) _chrUSrc, \
+                  **chrVSrc = (const int32_t **) _chrVSrc, \
+                  **alpSrc  = (const int32_t **) _alpSrc; \
+    uint16_t *dest = (uint16_t *) _dest; \
     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 int16_t *buf[2], \
-                        const int16_t *ubuf[2], const int16_t *vbuf[2], \
-                        const int16_t *abuf[2], uint8_t *dest, int dstW, \
+static void name ## ext ## _2_c(SwsContext *c, const int16_t *_buf[2], \
+                        const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
+                        const int16_t *_abuf[2], uint8_t *_dest, int dstW, \
                         int yalpha, int uvalpha, int y) \
 { \
+    const int32_t **buf  = (const int32_t **) _buf, \
+                  **ubuf = (const int32_t **) _ubuf, \
+                  **vbuf = (const int32_t **) _vbuf, \
+                  **abuf = (const int32_t **) _abuf; \
+    uint16_t *dest = (uint16_t *) _dest; \
     name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
                           dest, dstW, yalpha, uvalpha, y, fmt); \
 } \
  \
-static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
-                        const int16_t *ubuf[2], const int16_t *vbuf[2], \
-                        const int16_t *abuf0, uint8_t *dest, int dstW, \
+static void name ## ext ## _1_c(SwsContext *c, const int16_t *_buf0, \
+                        const int16_t *_ubuf[2], const int16_t *_vbuf[2], \
+                        const int16_t *_abuf0, uint8_t *_dest, int dstW, \
                         int uvalpha, int y) \
 { \
+    const int32_t *buf0  = (const int32_t *)  _buf0, \
+                 **ubuf  = (const int32_t **) _ubuf, \
+                 **vbuf  = (const int32_t **) _vbuf, \
+                  *abuf0 = (const int32_t *)  _abuf0; \
+    uint16_t *dest = (uint16_t *) _dest; \
     name ## base ## _1_c_template(c, buf0, ubuf, vbuf, abuf0, dest, \
                                   dstW, uvalpha, y, fmt); \
 }
 
-YUV2PACKEDWRAPPER(yuv2gray16,, LE, PIX_FMT_GRAY16LE);
-YUV2PACKEDWRAPPER(yuv2gray16,, BE, PIX_FMT_GRAY16BE);
+YUV2PACKED16WRAPPER(yuv2gray16,, LE, PIX_FMT_GRAY16LE);
+YUV2PACKED16WRAPPER(yuv2gray16,, BE, PIX_FMT_GRAY16BE);
 
 #define output_pixel(pos, acc) \
     if (target == PIX_FMT_MONOBLACK) { \
@@ -616,6 +632,38 @@ yuv2mono_1_c_template(SwsContext *c, const int16_t *buf0,
 
 #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 int16_t *buf[2], \
+                                const int16_t *ubuf[2], const int16_t *vbuf[2], \
+                                const int16_t *abuf[2], uint8_t *dest, int dstW, \
+                                int yalpha, int uvalpha, int y) \
+{ \
+    name ## base ## _2_c_template(c, buf, ubuf, vbuf, abuf, \
+                                  dest, dstW, yalpha, uvalpha, y, fmt); \
+} \
+ \
+static void name ## ext ## _1_c(SwsContext *c, const int16_t *buf0, \
+                                const int16_t *ubuf[2], const int16_t *vbuf[2], \
+                                const int16_t *abuf0, uint8_t *dest, int dstW, \
+                                int uvalpha, int y) \
+{ \
+    name ## base ## _1_c_template(c, buf0, ubuf, vbuf, \
+                                  abuf0, dest, dstW, uvalpha, \
+                                  y, fmt); \
+}
+
 YUV2PACKEDWRAPPER(yuv2mono,, white, PIX_FMT_MONOWHITE);
 YUV2PACKEDWRAPPER(yuv2mono,, black, PIX_FMT_MONOBLACK);
 
@@ -731,26 +779,32 @@ yuv2422_1_c_template(SwsContext *c, const int16_t *buf0,
 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 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 output_pixel(pos, val) \
+    if (isBE(target)) { \
+        AV_WB16(pos, val); \
+    } else { \
+        AV_WL16(pos, val); \
+    }
 
 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,
+                       const int32_t **lumSrc, int lumFilterSize,
+                       const int16_t *chrFilter, const int32_t **chrUSrc,
+                       const int32_t **chrVSrc, int chrFilterSize,
+                       const int32_t **alpSrc, uint16_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;
+        int Y1 = 0;
+        int Y2 = 0;
+        int U  = -128 << 23; // 19
+        int V  = -128 << 23;
+        int R, G, B;
 
         for (j = 0; j < lumFilterSize; j++) {
             Y1 += lumSrc[j][i * 2]     * lumFilter[j];
@@ -760,40 +814,45 @@ yuv2rgb48_X_c_template(SwsContext *c, const int16_t *lumFilter,
             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);
-        }
 
-        /* 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];
+        // 8bit: 12+15=27; 16-bit: 12+19=31
+        Y1 >>= 14; // 10
+        Y2 >>= 14;
+        U  >>= 14;
+        V  >>= 14;
+
+        // 8bit: 27 -> 17bit, 16bit: 31 - 14 = 17bit
+        Y1 -= c->yuv2rgb_y_offset;
+        Y2 -= c->yuv2rgb_y_offset;
+        Y1 *= c->yuv2rgb_y_coeff;
+        Y2 *= c->yuv2rgb_y_coeff;
+        Y1 += 1 << 13; // 21
+        Y2 += 1 << 13;
+        // 8bit: 17 + 13bit = 30bit, 16bit: 17 + 13bit = 30bit
+
+        R = V * c->yuv2rgb_v2r_coeff;
+        G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+        B =                            U * c->yuv2rgb_u2b_coeff;
 
-        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;
+        // 8bit: 30 - 22 = 8bit, 16bit: 30bit - 14 = 16bit
+        output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
+        output_pixel(&dest[1], av_clip_uintp2(  G + Y1, 30) >> 14);
+        output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
+        output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
+        output_pixel(&dest[4], av_clip_uintp2(  G + Y2, 30) >> 14);
+        output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
+        dest += 6;
     }
 }
 
 static av_always_inline void
-yuv2rgb48_2_c_template(SwsContext *c, const int16_t *buf[2],
-                       const int16_t *ubuf[2], const int16_t *vbuf[2],
-                       const int16_t *abuf[2], uint8_t *dest, int dstW,
+yuv2rgb48_2_c_template(SwsContext *c, const int32_t *buf[2],
+                       const int32_t *ubuf[2], const int32_t *vbuf[2],
+                       const int32_t *abuf[2], uint16_t *dest, int dstW,
                        int yalpha, int uvalpha, int y,
                        enum PixelFormat target)
 {
-    const int16_t *buf0  = buf[0],  *buf1  = buf[1],
+    const int32_t *buf0  = buf[0],  *buf1  = buf[1],
                   *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
                   *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
     int  yalpha1 = 4095 - yalpha;
@@ -801,80 +860,108 @@ yuv2rgb48_2_c_template(SwsContext *c, const int16_t *buf[2],
     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;
-        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];
+        int Y1 = (buf0[i * 2]     * yalpha1  + buf1[i * 2]     * yalpha) >> 14;
+        int Y2 = (buf0[i * 2 + 1] * yalpha1  + buf1[i * 2 + 1] * yalpha) >> 14;
+        int U  = (ubuf0[i]        * uvalpha1 + ubuf1[i]        * uvalpha + (-128 << 23)) >> 14;
+        int V  = (vbuf0[i]        * uvalpha1 + vbuf1[i]        * uvalpha + (-128 << 23)) >> 14;
+        int R, G, B;
 
-        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;
+        Y1 -= c->yuv2rgb_y_offset;
+        Y2 -= c->yuv2rgb_y_offset;
+        Y1 *= c->yuv2rgb_y_coeff;
+        Y2 *= c->yuv2rgb_y_coeff;
+        Y1 += 1 << 13;
+        Y2 += 1 << 13;
+
+        R = V * c->yuv2rgb_v2r_coeff;
+        G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+        B =                            U * c->yuv2rgb_u2b_coeff;
+
+        output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
+        output_pixel(&dest[1], av_clip_uintp2(  G + Y1, 30) >> 14);
+        output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
+        output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
+        output_pixel(&dest[4], av_clip_uintp2(  G + Y2, 30) >> 14);
+        output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
+        dest += 6;
     }
 }
 
 static av_always_inline void
-yuv2rgb48_1_c_template(SwsContext *c, const int16_t *buf0,
-                       const int16_t *ubuf[2], const int16_t *vbuf[2],
-                       const int16_t *abuf0, uint8_t *dest, int dstW,
+yuv2rgb48_1_c_template(SwsContext *c, const int32_t *buf0,
+                       const int32_t *ubuf[2], const int32_t *vbuf[2],
+                       const int32_t *abuf0, uint16_t *dest, int dstW,
                        int uvalpha, int y, enum PixelFormat target)
 {
-    const int16_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
+    const int32_t *ubuf0 = ubuf[0], *ubuf1 = ubuf[1],
                   *vbuf0 = vbuf[0], *vbuf1 = vbuf[1];
     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;
-            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;
+            int Y1 = (buf0[i * 2]    ) >> 2;
+            int Y2 = (buf0[i * 2 + 1]) >> 2;
+            int U  = (ubuf0[i] + (-128 << 11)) >> 2;
+            int V  = (vbuf0[i] + (-128 << 11)) >> 2;
+            int R, G, B;
+
+            Y1 -= c->yuv2rgb_y_offset;
+            Y2 -= c->yuv2rgb_y_offset;
+            Y1 *= c->yuv2rgb_y_coeff;
+            Y2 *= c->yuv2rgb_y_coeff;
+            Y1 += 1 << 13;
+            Y2 += 1 << 13;
+
+            R = V * c->yuv2rgb_v2r_coeff;
+            G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+            B =                            U * c->yuv2rgb_u2b_coeff;
+
+            output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
+            output_pixel(&dest[1], av_clip_uintp2(  G + Y1, 30) >> 14);
+            output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
+            output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
+            output_pixel(&dest[4], av_clip_uintp2(  G + Y2, 30) >> 14);
+            output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
+            dest += 6;
         }
     } 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;
+            int Y1 = (buf0[i * 2]    ) >> 2;
+            int Y2 = (buf0[i * 2 + 1]) >> 2;
+            int U  = (ubuf0[i] + ubuf1[i] + (-128 << 11)) >> 3;
+            int V  = (vbuf0[i] + vbuf1[i] + (-128 << 11)) >> 3;
+            int R, G, B;
+
+            Y1 -= c->yuv2rgb_y_offset;
+            Y2 -= c->yuv2rgb_y_offset;
+            Y1 *= c->yuv2rgb_y_coeff;
+            Y2 *= c->yuv2rgb_y_coeff;
+            Y1 += 1 << 13;
+            Y2 += 1 << 13;
+
+            R = V * c->yuv2rgb_v2r_coeff;
+            G = V * c->yuv2rgb_v2g_coeff + U * c->yuv2rgb_u2g_coeff;
+            B =                            U * c->yuv2rgb_u2b_coeff;
+
+            output_pixel(&dest[0], av_clip_uintp2(R_B + Y1, 30) >> 14);
+            output_pixel(&dest[1], av_clip_uintp2(  G + Y1, 30) >> 14);
+            output_pixel(&dest[2], av_clip_uintp2(B_R + Y1, 30) >> 14);
+            output_pixel(&dest[3], av_clip_uintp2(R_B + Y2, 30) >> 14);
+            output_pixel(&dest[4], av_clip_uintp2(  G + Y2, 30) >> 14);
+            output_pixel(&dest[5], av_clip_uintp2(B_R + Y2, 30) >> 14);
+            dest += 6;
         }
     }
 }
 
+#undef output_pixel
 #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);
+YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48be, PIX_FMT_RGB48BE);
+YUV2PACKED16WRAPPER(yuv2, rgb48, rgb48le, PIX_FMT_RGB48LE);
+YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48be, PIX_FMT_BGR48BE);
+YUV2PACKED16WRAPPER(yuv2, rgb48, bgr48le, PIX_FMT_BGR48LE);
 
 static av_always_inline void
 yuv2rgb_write(uint8_t *_dest, int i, int Y1, int Y2,
@@ -1025,7 +1112,7 @@ yuv2rgb_X_c_template(SwsContext *c, const int16_t *lumFilter,
             U  = av_clip_uint8(U);
             V  = av_clip_uint8(V);
         }
-        if (hasAlpha) {\
+        if (hasAlpha) {
             A1 = 1 << 18;
             A2 = 1 << 18;
             for (j = 0; j < lumFilterSize; j++) {
@@ -1318,50 +1405,50 @@ static av_always_inline void fillPlane(uint8_t* plane, int stride,
 #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,
+rgb48ToY_c_template(uint16_t *dst, const uint16_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;
+        unsigned int r_b = input_pixel(&src[i*3+0]);
+        unsigned int   g = input_pixel(&src[i*3+1]);
+        unsigned int b_r = input_pixel(&src[i*3+2]);
 
-        dst[i] = (RY*r + GY*g + BY*b + (33<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
+        dst[i] = (RY*r + GY*g + BY*b + (0x2001<<(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,
+rgb48ToUV_c_template(uint16_t *dstU, uint16_t *dstV,
+                    const uint16_t *src1, const uint16_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;
+        int r_b = input_pixel(&src1[i*3+0]);
+        int   g = input_pixel(&src1[i*3+1]);
+        int b_r = input_pixel(&src1[i*3+2]);
 
-        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;
+        dstU[i] = (RU*r + GU*g + BU*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
+        dstV[i] = (RV*r + GV*g + BV*b + (0x10001<<(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,
+rgb48ToUV_half_c_template(uint16_t *dstU, uint16_t *dstV,
+                          const uint16_t *src1, const uint16_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);
+        int r_b = (input_pixel(&src1[6 * i + 0]) + input_pixel(&src1[6 * i + 3]) + 1) >> 1;
+        int   g = (input_pixel(&src1[6 * i + 1]) + input_pixel(&src1[6 * i + 4]) + 1) >> 1;
+        int b_r = (input_pixel(&src1[6 * i + 2]) + input_pixel(&src1[6 * i + 5]) + 1) >> 1;
 
-        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);
+        dstU[i]= (RU*r + GU*g + BU*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
+        dstV[i]= (RV*r + GV*g + BV*b + (0x10001<<(RGB2YUV_SHIFT-1))) >> RGB2YUV_SHIFT;
     }
 }
 
@@ -1370,23 +1457,31 @@ rgb48ToUV_half_c_template(uint8_t *dstU, uint8_t *dstV,
 #undef input_pixel
 
 #define rgb48funcs(pattern, BE_LE, origin) \
-static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *dst, const uint8_t *src, \
+static void pattern ## 48 ## BE_LE ## ToY_c(uint8_t *_dst, const uint8_t *_src, \
                                     int width, uint32_t *unused) \
 { \
+    const uint16_t *src = (const uint16_t *) _src; \
+    uint16_t *dst = (uint16_t *) _dst; \
     rgb48ToY_c_template(dst, src, width, origin); \
 } \
  \
-static void pattern ## 48 ## BE_LE ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
-                                    const uint8_t *src1, const uint8_t *src2, \
+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) \
 { \
+    const uint16_t *src1 = (const uint16_t *) _src1, \
+                   *src2 = (const uint16_t *) _src2; \
+    uint16_t *dstU = (uint16_t *) _dstU, *dstV = (uint16_t *) _dstV; \
     rgb48ToUV_c_template(dstU, dstV, src1, src2, width, origin); \
 } \
  \
-static void pattern ## 48 ## BE_LE ## ToUV_half_c(uint8_t *dstU, uint8_t *dstV, \
-                                    const uint8_t *src1, const uint8_t *src2, \
+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) \
 { \
+    const uint16_t *src1 = (const uint16_t *) _src1, \
+                   *src2 = (const uint16_t *) _src2; \
+    uint16_t *dstU = (uint16_t *) _dstU, *dstV = (uint16_t *) _dstV; \
     rgb48ToUV_half_c_template(dstU, dstV, src1, src2, width, origin); \
 }
 
@@ -1601,13 +1696,26 @@ static void yuy2ToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
     assert(src1 == src2);
 }
 
-static void LEToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
-                     const uint8_t *src2, int width, uint32_t *unused)
+static void bswap16Y_c(uint8_t *_dst, const uint8_t *_src, int width, uint32_t *unused)
 {
     int i;
+    const uint16_t *src = (const uint16_t *) _src;
+    uint16_t *dst = (uint16_t *) _dst;
     for (i=0; i<width; i++) {
-        dstU[i]= src1[2*i + 1];
-        dstV[i]= src2[2*i + 1];
+        dst[i] = av_bswap16(src[i]);
+    }
+}
+
+static void bswap16UV_c(uint8_t *_dstU, uint8_t *_dstV, const uint8_t *_src1,
+                        const uint8_t *_src2, int width, uint32_t *unused)
+{
+    int i;
+    const uint16_t *src1 = (const uint16_t *) _src1,
+                   *src2 = (const uint16_t *) _src2;
+    uint16_t *dstU = (uint16_t *) _dstU, *dstV = (uint16_t *) _dstV;
+    for (i=0; i<width; i++) {
+        dstU[i] = av_bswap16(src1[i]);
+        dstV[i] = av_bswap16(src2[i]);
     }
 }
 
@@ -1632,16 +1740,6 @@ static void uyvyToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
     assert(src1 == src2);
 }
 
-static void BEToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
-                     const uint8_t *src2, int width, uint32_t *unused)
-{
-    int i;
-    for (i=0; i<width; i++) {
-        dstU[i]= src1[2*i];
-        dstV[i]= src2[2*i];
-    }
-}
-
 static av_always_inline void nvXXtoUV_c(uint8_t *dst1, uint8_t *dst2,
                                         const uint8_t *src, int width)
 {
@@ -1670,43 +1768,49 @@ static void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV,
 
 // FIXME Maybe dither instead.
 static av_always_inline void
-yuv9_OR_10ToUV_c_template(uint8_t *dstU, uint8_t *dstV,
-                          const uint8_t *_srcU, const uint8_t *_srcV,
+yuv9_OR_10ToUV_c_template(uint16_t *dstU, uint16_t *dstV,
+                          const uint16_t *srcU, const uint16_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);
+        int upx = input_pixel(&srcU[i]);
+        int vpx = input_pixel(&srcV[i]);
+        dstU[i] =  (upx << (16 - depth)) | (upx >> (2 * depth - 16));
+        dstV[i] =  (vpx << (16 - depth)) | (vpx >> (2 * depth - 16));
     }
 }
 
 static av_always_inline void
-yuv9_or_10ToY_c_template(uint8_t *dstY, const uint8_t *_srcY,
+yuv9_or_10ToY_c_template(uint16_t *dstY, const uint16_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);
+    for (i = 0; i < width; i++) {
+        int px = input_pixel(&srcY[i]);
+        dstY[i] =  (px << (16 - depth)) | (px >> (2 * depth - 16));
+    }
 }
 
 #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, \
+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) \
 { \
+    uint16_t *dstU = (uint16_t *) _dstU, *dstV = (uint16_t *) _dstV; \
+    const uint16_t *srcU = (const uint16_t *) _srcU, \
+                   *srcV = (const uint16_t *) _srcV; \
     yuv9_OR_10ToUV_c_template(dstU, dstV, srcU, srcV, width, origin, depth); \
 } \
-static void BE_LE ## depth ## ToY_c(uint8_t *dstY, const uint8_t *srcY, \
+static void BE_LE ## depth ## ToY_c(uint8_t *_dstY, const uint8_t *_srcY, \
                                     int width, uint32_t *unused) \
 { \
+    uint16_t *dstY = (uint16_t *) _dstY; \
+    const uint16_t *srcY = (const uint16_t *) _srcY; \
     yuv9_or_10ToY_c_template(dstY, srcY, width, origin, depth); \
 }
 
@@ -1801,6 +1905,27 @@ static void rgb24ToUV_half_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
     }
 }
 
+static void hScale16_c(int16_t *_dst, int dstW, const uint8_t *_src,
+                       const int16_t *filter,
+                       const int16_t *filterPos, int filterSize)
+{
+    int i;
+    int32_t *dst = (int32_t *) _dst;
+    const uint16_t *src = (const uint16_t *) _src;
+
+    for (i = 0; i < dstW; i++) {
+        int j;
+        int srcPos = filterPos[i];
+        unsigned int val = 0;
+
+        for (j = 0; j < filterSize; j++) {
+            val += src[srcPos + j] * filter[filterSize * i + j];
+        }
+        // filter=14 bit, input=16 bit, output=30 bit, >> 11 makes 19 bit
+        dst[i] = FFMIN(val >> 11, (1 << 19) - 1);
+    }
+}
+
 // bilinear / bicubic scaling
 static void hScale_c(int16_t *dst, int dstW, const uint8_t *src,
                      const int16_t *filter, const int16_t *filterPos,
@@ -1851,6 +1976,41 @@ static void lumRangeFromJpeg_c(int16_t *dst, int width)
         dst[i] = (dst[i]*14071 + 33561947)>>14;
 }
 
+static void chrRangeToJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
+{
+    int i;
+    int32_t *dstU = (int32_t *) _dstU;
+    int32_t *dstV = (int32_t *) _dstV;
+    for (i = 0; i < width; i++) {
+        dstU[i] = (FFMIN(dstU[i],30775)*4663 - 9289992)>>12; //-264
+        dstV[i] = (FFMIN(dstV[i],30775)*4663 - 9289992)>>12; //-264
+    }
+}
+static void chrRangeFromJpeg16_c(int16_t *_dstU, int16_t *_dstV, int width)
+{
+    int i;
+    int32_t *dstU = (int32_t *) _dstU;
+    int32_t *dstV = (int32_t *) _dstV;
+    for (i = 0; i < width; i++) {
+        dstU[i] = (dstU[i]*1799 + 4081085)>>11; //1469
+        dstV[i] = (dstV[i]*1799 + 4081085)>>11; //1469
+    }
+}
+static void lumRangeToJpeg16_c(int16_t *_dst, int width)
+{
+    int i;
+    int32_t *dst = (int32_t *) _dst;
+    for (i = 0; i < width; i++)
+        dst[i] = (FFMIN(dst[i],30189)*19077 - 39057361)>>14;
+}
+static void lumRangeFromJpeg16_c(int16_t *_dst, int width)
+{
+    int i;
+    int32_t *dst = (int32_t *) _dst;
+    for (i = 0; i < width; i++)
+        dst[i] = (dst[i]*14071 + 33561947)>>14;
+}
+
 static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
                            const uint8_t *src, int srcW, int xInc)
 {
@@ -1864,8 +2024,25 @@ static void hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
     }
 }
 
+static void scale8To16Rv_c(uint16_t *_dst, const uint8_t *src, int len)
+{
+    int i;
+    uint8_t *dst = (uint8_t *) _dst;
+    for (i = len - 1; i >= 0; i--) {
+        dst[i * 2] = dst[i * 2 + 1] = src[i];
+    }
+}
+
+static void scale19To15Fw_c(int16_t *dst, const int32_t *src, int len)
+{
+    int i;
+    for (i = 0; i < len; i++) {
+        dst[i] = src[i] >> 4;
+    }
+}
+
 // *** horizontal scale Y line to temp buffer
-static av_always_inline void hyscale(SwsContext *c, uint16_t *dst, int dstWidth,
+static av_always_inline void hyscale(SwsContext *c, int16_t *dst, int dstWidth,
                                      const uint8_t *src, int srcW, int xInc,
                                      const int16_t *hLumFilter,
                                      const int16_t *hLumFilterPos, int hLumFilterSize,
@@ -1880,6 +2057,11 @@ static av_always_inline void hyscale(SwsContext *c, uint16_t *dst, int dstWidth,
         src= formatConvBuffer;
     }
 
+    if (av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1 < 8 && c->scalingBpp == 16) {
+        c->scale8To16Rv((uint16_t *) formatConvBuffer, src, srcW);
+        src = formatConvBuffer;
+    }
+
     if (!c->hyscale_fast) {
         c->hScale(dst, dstWidth, src, hLumFilter, hLumFilterPos, hLumFilterSize);
     } else { // fast bilinear upscale / crap downscale
@@ -1888,6 +2070,10 @@ static av_always_inline void hyscale(SwsContext *c, uint16_t *dst, int dstWidth,
 
     if (convertRange)
         convertRange(dst, dstWidth);
+
+    if (av_pix_fmt_descriptors[c->dstFormat].comp[0].depth_minus1 < 8 && c->scalingBpp == 16) {
+        c->scale19To15Fw(dst, (int32_t *) dst, dstWidth);
+    }
 }
 
 static void hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
@@ -1905,19 +2091,27 @@ static void hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
     }
 }
 
-static av_always_inline void hcscale(SwsContext *c, uint16_t *dst1, uint16_t *dst2, int dstWidth,
+static av_always_inline void hcscale(SwsContext *c, int16_t *dst1, int16_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);
+        uint8_t *buf2 = formatConvBuffer + FFALIGN(srcW * c->scalingBpp >> 3, 16);
         c->chrToYV12(formatConvBuffer, buf2, src1, src2, srcW, pal);
         src1= formatConvBuffer;
         src2= buf2;
     }
 
+    if (av_pix_fmt_descriptors[c->srcFormat].comp[0].depth_minus1 < 8 && c->scalingBpp == 16) {
+        uint8_t *buf2 = (formatConvBuffer + FFALIGN(srcW * 2, 16));
+        c->scale8To16Rv((uint16_t *) formatConvBuffer, src1, srcW);
+        c->scale8To16Rv((uint16_t *) buf2,             src2, srcW);
+        src1 = formatConvBuffer;
+        src2 = buf2;
+    }
+
     if (!c->hcscale_fast) {
         c->hScale(dst1, dstWidth, src1, hChrFilter, hChrFilterPos, hChrFilterSize);
         c->hScale(dst2, dstWidth, src2, hChrFilter, hChrFilterPos, hChrFilterSize);
@@ -1927,6 +2121,11 @@ static av_always_inline void hcscale(SwsContext *c, uint16_t *dst1, uint16_t *ds
 
     if (c->chrConvertRange)
         c->chrConvertRange(dst1, dst2, dstWidth);
+
+    if (av_pix_fmt_descriptors[c->dstFormat].comp[0].depth_minus1 < 8 && c->scalingBpp == 16) {
+        c->scale19To15Fw(dst1, (int32_t *) dst1, dstWidth);
+        c->scale19To15Fw(dst2, (int32_t *) dst2, dstWidth);
+    }
 }
 
 static av_always_inline void
@@ -2049,20 +2248,20 @@ find_c_packed_planar_out_funcs(SwsContext *c,
             *yuv2packedX = yuv2uyvy422_X_c;
             break;
         case PIX_FMT_RGB48LE:
-            //*yuv2packed1 = yuv2rgb48le_1_c;
-            //*yuv2packed2 = yuv2rgb48le_2_c;
-            //*yuv2packedX = yuv2rgb48le_X_c;
-            //break;
+            *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;
+            *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;
@@ -2435,13 +2634,6 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
                                    &c->yuv2packed1, &c->yuv2packed2,
                                    &c->yuv2packedX);
 
-    c->hScale       = hScale_c;
-
-    if (c->flags & SWS_FAST_BILINEAR) {
-        c->hyscale_fast = hyscale_fast_c;
-        c->hcscale_fast = hcscale_fast_c;
-    }
-
     c->chrToYV12 = NULL;
     switch(srcFormat) {
         case PIX_FMT_YUYV422  : c->chrToYV12 = yuy2ToUV_c; break;
@@ -2453,22 +2645,25 @@ static av_cold 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_YUV420P9LE:  c->chrToYV12 = LE9ToUV_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:
-        case PIX_FMT_YUV444P16BE: c->chrToYV12 = BEToUV_c; break;
+        case PIX_FMT_YUV444P9BE:
+        case PIX_FMT_YUV420P9BE:  c->chrToYV12 = BE9ToUV_c; break;
+        case PIX_FMT_YUV444P10BE:
+        case PIX_FMT_YUV422P10BE:
+        case PIX_FMT_YUV420P10BE: c->chrToYV12 = BE10ToUV_c; break;
+#if HAVE_BIGENDIAN
         case PIX_FMT_YUV420P16LE:
         case PIX_FMT_YUV422P16LE:
-        case PIX_FMT_YUV444P16LE: c->chrToYV12 = LEToUV_c; break;
+        case PIX_FMT_YUV444P16LE: c->chrToYV12 = bswap16UV_c; break;
+#else
+        case PIX_FMT_YUV420P16BE:
+        case PIX_FMT_YUV422P16BE:
+        case PIX_FMT_YUV444P16BE: c->chrToYV12 = bswap16UV_c; break;
+#endif
     }
     if (c->chrSrcHSubSample) {
         switch(srcFormat) {
@@ -2517,27 +2712,30 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
     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_YUV420P9LE:  c->lumToYV12 = LE9ToY_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:
-    case PIX_FMT_YUV422P16BE:
-    case PIX_FMT_YUV444P16BE:
-    case PIX_FMT_Y400A    :
-    case PIX_FMT_GRAY16BE : c->lumToYV12 = yuy2ToY_c; break;
-    case PIX_FMT_UYVY422  :
+    case PIX_FMT_YUV444P9BE:
+    case PIX_FMT_YUV420P9BE:  c->lumToYV12 = BE9ToY_c; break;
+    case PIX_FMT_YUV444P10BE:
+    case PIX_FMT_YUV422P10BE:
+    case PIX_FMT_YUV420P10BE: c->lumToYV12 = BE10ToY_c; break;
+#if HAVE_BIGENDIAN
     case PIX_FMT_YUV420P16LE:
     case PIX_FMT_YUV422P16LE:
     case PIX_FMT_YUV444P16LE:
-    case PIX_FMT_GRAY16LE : c->lumToYV12 = uyvyToY_c;    break;
+    case PIX_FMT_GRAY16LE: c->lumToYV12 = bswap16Y_c; break;
+#else
+    case PIX_FMT_YUV420P16BE:
+    case PIX_FMT_YUV422P16BE:
+    case PIX_FMT_YUV444P16BE:
+    case PIX_FMT_GRAY16BE: c->lumToYV12 = bswap16Y_c; break;
+#endif
+    case PIX_FMT_YUYV422  :
+    case PIX_FMT_Y400A    : c->lumToYV12 = yuy2ToY_c; break;
+    case PIX_FMT_UYVY422  : 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;
@@ -2574,6 +2772,13 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
         }
     }
 
+    if (c->scalingBpp == 8) {
+    c->hScale       = hScale_c;
+    if (c->flags & SWS_FAST_BILINEAR) {
+        c->hyscale_fast = hyscale_fast_c;
+        c->hcscale_fast = hcscale_fast_c;
+    }
+
     if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {
         if (c->srcRange) {
             c->lumConvertRange = lumRangeFromJpeg_c;
@@ -2583,6 +2788,21 @@ static av_cold void sws_init_swScale_c(SwsContext *c)
             c->chrConvertRange = chrRangeToJpeg_c;
         }
     }
+    } else {
+        c->hScale = hScale16_c;
+        c->scale19To15Fw = scale19To15Fw_c;
+        c->scale8To16Rv  = scale8To16Rv_c;
+
+        if (c->srcRange != c->dstRange && !isAnyRGB(c->dstFormat)) {
+            if (c->srcRange) {
+                c->lumConvertRange = lumRangeFromJpeg16_c;
+                c->chrConvertRange = chrRangeFromJpeg16_c;
+            } else {
+                c->lumConvertRange = lumRangeToJpeg16_c;
+                c->chrConvertRange = chrRangeToJpeg16_c;
+            }
+        }
+    }
 
     if (!(isGray(srcFormat) || isGray(c->dstFormat) ||
           srcFormat == PIX_FMT_MONOBLACK || srcFormat == PIX_FMT_MONOWHITE))
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index c1eed8a..cfb19d9 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -205,6 +205,7 @@ typedef struct SwsContext {
     enum PixelFormat srcFormat;   ///< Source      pixel format.
     int dstFormatBpp;             ///< Number of bits per pixel of the destination pixel format.
     int srcFormatBpp;             ///< Number of bits per pixel of the source      pixel format.
+    int scalingBpp;
     int chrSrcHSubSample;         ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in source      image.
     int chrSrcVSubSample;         ///< Binary logarithm of vertical   subsampling factor between luma/alpha and chroma planes in source      image.
     int chrDstHSubSample;         ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in destination image.
@@ -446,6 +447,15 @@ typedef struct SwsContext {
     void (*lumConvertRange)(int16_t *dst, int width); ///< Color range conversion function for luma plane if needed.
     void (*chrConvertRange)(int16_t *dst1, int16_t *dst2, int width); ///< Color range conversion function for chroma planes if needed.
 
+    /**
+     * dst[..] = (src[..] << 8) | src[..];
+     */
+    void (*scale8To16Rv)(uint16_t *dst, const uint8_t *src, int len);
+    /**
+     * dst[..] = src[..] >> 4;
+     */
+    void (*scale19To15Fw)(int16_t *dst, const int32_t *src, int len);
+
     int needs_hcscale; ///< Set if there are chroma planes to be converted.
 
 } SwsContext;
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 6971418..b7ccac9 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -877,8 +877,13 @@ int sws_init_context(SwsContext *c, SwsFilter *srcFilter, SwsFilter *dstFilter)
         }
     }
 
-    FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW, 16) * 2, fail);
-    if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) {
+    // FIXME it's even nicer if bpp isn't 16, but max({src,dst}formatbpp)
+    c->scalingBpp = FFMAX(av_pix_fmt_descriptors[srcFormat].comp[0].depth_minus1,
+                          av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1) >= 8 ? 16 : 8;
+    if (c->scalingBpp == 16)
+        dst_stride <<= 1;
+    FF_ALLOC_OR_GOTO(c, c->formatConvBuffer, FFALIGN(srcW, 16) * 2 * c->scalingBpp >> 3, fail);
+    if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2 && c->scalingBpp == 8) {
         c->canMMX2BeUsed= (dstW >=srcW && (dstW&31)==0 && (srcW&15)==0) ? 1 : 0;
         if (!c->canMMX2BeUsed && dstW >=srcW && (srcW&15)==0 && (flags&SWS_FAST_BILINEAR)) {
             if (flags&SWS_PRINT_INFO)
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index de0e4ab..54c7edc 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -1625,32 +1625,6 @@ static void RENAME(yuy2ToUV)(uint8_t *dstU, uint8_t *dstV,
     assert(src1 == src2);
 }
 
-static void RENAME(LEToUV)(uint8_t *dstU, uint8_t *dstV,
-                           const uint8_t *src1, const uint8_t *src2,
-                           int width, uint32_t *unused)
-{
-    __asm__ volatile(
-        "mov                    %0, %%"REG_a"       \n\t"
-        "1:                                         \n\t"
-        "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
-        "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
-        "movq    (%2, %%"REG_a",2), %%mm2           \n\t"
-        "movq   8(%2, %%"REG_a",2), %%mm3           \n\t"
-        "psrlw                  $8, %%mm0           \n\t"
-        "psrlw                  $8, %%mm1           \n\t"
-        "psrlw                  $8, %%mm2           \n\t"
-        "psrlw                  $8, %%mm3           \n\t"
-        "packuswb            %%mm1, %%mm0           \n\t"
-        "packuswb            %%mm3, %%mm2           \n\t"
-        "movq                %%mm0, (%3, %%"REG_a") \n\t"
-        "movq                %%mm2, (%4, %%"REG_a") \n\t"
-        "add                    $8, %%"REG_a"       \n\t"
-        " js                    1b                  \n\t"
-        : : "g" ((x86_reg)-width), "r" (src1+width*2), "r" (src2+width*2), "r" (dstU+width), "r" (dstV+width)
-        : "%"REG_a
-    );
-}
-
 /* This is almost identical to the previous, end exists only because
  * yuy2ToY/UV)(dst, src+1, ...) would have 100% unaligned accesses. */
 static void RENAME(uyvyToY)(uint8_t *dst, const uint8_t *src,
@@ -1700,33 +1674,6 @@ static void RENAME(uyvyToUV)(uint8_t *dstU, uint8_t *dstV,
     assert(src1 == src2);
 }
 
-static void RENAME(BEToUV)(uint8_t *dstU, uint8_t *dstV,
-                           const uint8_t *src1, const uint8_t *src2,
-                           int width, uint32_t *unused)
-{
-    __asm__ volatile(
-        "movq "MANGLE(bm01010101)", %%mm4           \n\t"
-        "mov                    %0, %%"REG_a"       \n\t"
-        "1:                                         \n\t"
-        "movq    (%1, %%"REG_a",2), %%mm0           \n\t"
-        "movq   8(%1, %%"REG_a",2), %%mm1           \n\t"
-        "movq    (%2, %%"REG_a",2), %%mm2           \n\t"
-        "movq   8(%2, %%"REG_a",2), %%mm3           \n\t"
-        "pand                %%mm4, %%mm0           \n\t"
-        "pand                %%mm4, %%mm1           \n\t"
-        "pand                %%mm4, %%mm2           \n\t"
-        "pand                %%mm4, %%mm3           \n\t"
-        "packuswb            %%mm1, %%mm0           \n\t"
-        "packuswb            %%mm3, %%mm2           \n\t"
-        "movq                %%mm0, (%3, %%"REG_a") \n\t"
-        "movq                %%mm2, (%4, %%"REG_a") \n\t"
-        "add                    $8, %%"REG_a"       \n\t"
-        " js                    1b                  \n\t"
-        : : "g" ((x86_reg)-width), "r" (src1+width*2), "r" (src2+width*2), "r" (dstU+width), "r" (dstV+width)
-        : "%"REG_a
-    );
-}
-
 static av_always_inline void RENAME(nvXXtoUV)(uint8_t *dst1, uint8_t *dst2,
                                               const uint8_t *src, int width)
 {
@@ -2268,6 +2215,7 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
         }
     }
 
+    if (c->scalingBpp == 8) {
 #if !COMPILE_TEMPLATE_MMX2
     c->hScale       = RENAME(hScale      );
 #endif /* !COMPILE_TEMPLATE_MMX2 */
@@ -2285,6 +2233,7 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
 #if COMPILE_TEMPLATE_MMX2
     }
 #endif /* COMPILE_TEMPLATE_MMX2 */
+    }
 
 #if !COMPILE_TEMPLATE_MMX2
     switch(srcFormat) {
@@ -2292,12 +2241,6 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
         case PIX_FMT_UYVY422  : c->chrToYV12 = RENAME(uyvyToUV); break;
         case PIX_FMT_NV12     : c->chrToYV12 = RENAME(nv12ToUV); break;
         case PIX_FMT_NV21     : c->chrToYV12 = RENAME(nv21ToUV); break;
-        case PIX_FMT_YUV420P16BE:
-        case PIX_FMT_YUV422P16BE:
-        case PIX_FMT_YUV444P16BE: c->chrToYV12 = RENAME(BEToUV); break;
-        case PIX_FMT_YUV420P16LE:
-        case PIX_FMT_YUV422P16LE:
-        case PIX_FMT_YUV444P16LE: c->chrToYV12 = RENAME(LEToUV); break;
         default: break;
     }
 #endif /* !COMPILE_TEMPLATE_MMX2 */
@@ -2312,16 +2255,8 @@ static av_cold void RENAME(sws_init_swScale)(SwsContext *c)
     switch (srcFormat) {
 #if !COMPILE_TEMPLATE_MMX2
     case PIX_FMT_YUYV422  :
-    case PIX_FMT_YUV420P16BE:
-    case PIX_FMT_YUV422P16BE:
-    case PIX_FMT_YUV444P16BE:
-    case PIX_FMT_Y400A    :
-    case PIX_FMT_GRAY16BE : c->lumToYV12 = RENAME(yuy2ToY); break;
-    case PIX_FMT_UYVY422  :
-    case PIX_FMT_YUV420P16LE:
-    case PIX_FMT_YUV422P16LE:
-    case PIX_FMT_YUV444P16LE:
-    case PIX_FMT_GRAY16LE : c->lumToYV12 = RENAME(uyvyToY); break;
+    case PIX_FMT_Y400A    : c->lumToYV12 = RENAME(yuy2ToY); break;
+    case PIX_FMT_UYVY422  : c->lumToYV12 = RENAME(uyvyToY); break;
 #endif /* !COMPILE_TEMPLATE_MMX2 */
     case PIX_FMT_BGR24    : c->lumToYV12 = RENAME(bgr24ToY); break;
     case PIX_FMT_RGB24    : c->lumToYV12 = RENAME(rgb24ToY); break;
diff --git a/tests/ref/lavfi/pixdesc b/tests/ref/lavfi/pixdesc
index de13e94..8695b1d 100644
--- a/tests/ref/lavfi/pixdesc
+++ b/tests/ref/lavfi/pixdesc
@@ -1,8 +1,8 @@
 abgr                037bf9df6a765520ad6d490066bf4b89
 argb                c442a8261c2265a07212ef0f72e35f5a
 bgr24               0d0cb38ab3fa0b2ec0865c14f78b217b
-bgr48be             4ba0ff7fc9e011ea264610ad1585bb1f
-bgr48le             d022bfdd6a07d5dcc693799322a386b4
+bgr48be             74dedaaacae8fd1ef46e05f78cf29d62
+bgr48le             0eb7d30801eac6058814bddd330b3c76
 bgr4_byte           50d23cc82d9dcef2fd12adb81fb9b806
 bgr555be            49f01b1f1f0c84fd9e776dd34cc3c280
 bgr555le            378d6ac4223651a1adcbf94a3d0d807b
@@ -18,8 +18,8 @@ monow               9251497f3b0634f1165d12d5a289d943
 nv12                e0af357888584d36eec5aa0f673793ef
 nv21                9a3297f3b34baa038b1f37cb202b512f
 rgb24               b41eba9651e1b5fe386289b506188105
-rgb48be             460b6de89b156290a12d3941db8bd731
-rgb48le             cd93cb34d15996987367dabda3a10128
+rgb48be             e3bc84c9af376fb6d0f0293cc7b713a6
+rgb48le             f51c0e71638a822458329abb2f4052c7
 rgb4_byte           c93ba89b74c504e7f5ae9d9ab1546c73
 rgb555be            912a62c5e53bfcbac2a0340e10973cf2
 rgb555le            a937a0fc764fb57dc1b3af87cba0273c
@@ -38,12 +38,12 @@ yuv420p16le         2d59c4f1d0314a5a957a7cfc4b6fabcc
 yuv420p9be          ce880fa07830e5297c22acf6e20555ce
 yuv420p9le          16543fda8f87d94a6cf857d2e8d4461a
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
-yuv422p16be         5499502e1c29534a158a1fe60e889f60
-yuv422p16le         e3d61fde6978591596bc36b914386623
+yuv422p16be         86ad3447f97969ce095afeef81fa9abf
+yuv422p16le         a53a9b451f4a81eeae33362c1bbd07f8
 yuv440p             5a064afe2b453bb52cdb3f176b1aa1cf
 yuv444p             0a98447b78fd476aa39686da6a74fa2e
-yuv444p16be         ea602a24b8e6969679265078bd8607b6
-yuv444p16le         1262a0dc57ee147967fc896d04206313
+yuv444p16be         95db370ae765dd3d10b7def14704ae73
+yuv444p16le         36b969179b5ad9d312a0d1e1cd6bc402
 yuva420p            a29884f3f3dfe1e00b961bc17bef3d47
 yuvj420p            32eec78ba51857b16ce9b813a49b7189
 yuvj422p            0dfa0ed434f73be51428758c69e082cb
diff --git a/tests/ref/lavfi/pixfmts_copy b/tests/ref/lavfi/pixfmts_copy
index de13e94..8695b1d 100644
--- a/tests/ref/lavfi/pixfmts_copy
+++ b/tests/ref/lavfi/pixfmts_copy
@@ -1,8 +1,8 @@
 abgr                037bf9df6a765520ad6d490066bf4b89
 argb                c442a8261c2265a07212ef0f72e35f5a
 bgr24               0d0cb38ab3fa0b2ec0865c14f78b217b
-bgr48be             4ba0ff7fc9e011ea264610ad1585bb1f
-bgr48le             d022bfdd6a07d5dcc693799322a386b4
+bgr48be             74dedaaacae8fd1ef46e05f78cf29d62
+bgr48le             0eb7d30801eac6058814bddd330b3c76
 bgr4_byte           50d23cc82d9dcef2fd12adb81fb9b806
 bgr555be            49f01b1f1f0c84fd9e776dd34cc3c280
 bgr555le            378d6ac4223651a1adcbf94a3d0d807b
@@ -18,8 +18,8 @@ monow               9251497f3b0634f1165d12d5a289d943
 nv12                e0af357888584d36eec5aa0f673793ef
 nv21                9a3297f3b34baa038b1f37cb202b512f
 rgb24               b41eba9651e1b5fe386289b506188105
-rgb48be             460b6de89b156290a12d3941db8bd731
-rgb48le             cd93cb34d15996987367dabda3a10128
+rgb48be             e3bc84c9af376fb6d0f0293cc7b713a6
+rgb48le             f51c0e71638a822458329abb2f4052c7
 rgb4_byte           c93ba89b74c504e7f5ae9d9ab1546c73
 rgb555be            912a62c5e53bfcbac2a0340e10973cf2
 rgb555le            a937a0fc764fb57dc1b3af87cba0273c
@@ -38,12 +38,12 @@ yuv420p16le         2d59c4f1d0314a5a957a7cfc4b6fabcc
 yuv420p9be          ce880fa07830e5297c22acf6e20555ce
 yuv420p9le          16543fda8f87d94a6cf857d2e8d4461a
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
-yuv422p16be         5499502e1c29534a158a1fe60e889f60
-yuv422p16le         e3d61fde6978591596bc36b914386623
+yuv422p16be         86ad3447f97969ce095afeef81fa9abf
+yuv422p16le         a53a9b451f4a81eeae33362c1bbd07f8
 yuv440p             5a064afe2b453bb52cdb3f176b1aa1cf
 yuv444p             0a98447b78fd476aa39686da6a74fa2e
-yuv444p16be         ea602a24b8e6969679265078bd8607b6
-yuv444p16le         1262a0dc57ee147967fc896d04206313
+yuv444p16be         95db370ae765dd3d10b7def14704ae73
+yuv444p16le         36b969179b5ad9d312a0d1e1cd6bc402
 yuva420p            a29884f3f3dfe1e00b961bc17bef3d47
 yuvj420p            32eec78ba51857b16ce9b813a49b7189
 yuvj422p            0dfa0ed434f73be51428758c69e082cb
diff --git a/tests/ref/lavfi/pixfmts_crop b/tests/ref/lavfi/pixfmts_crop
index af42cd3..c7b5952 100644
--- a/tests/ref/lavfi/pixfmts_crop
+++ b/tests/ref/lavfi/pixfmts_crop
@@ -1,8 +1,8 @@
 abgr                cd761690872843d1b7ab0c695393c751
 argb                2ec6ef18769bcd651c2e8904d5a3ee67
 bgr24               3450fd00cf1493d1ded75544d82ba3ec
-bgr48be             90cb5d373a1123432d63c6a10c101afa
-bgr48le             9371f54ceda9010f1199e86f4930ac3f
+bgr48be             a9a7d177cef0914d3f1d266f00dff676
+bgr48le             b475d1b529ed80c728ddbacd22d35281
 bgr4_byte           2f6ac3cdd4676ab4e2982bdf0664945b
 bgr555be            d3a7c273604723adeb7e5f5dd1c4272b
 bgr555le            d22442fc13b464f9ba455b08df4e981f
@@ -14,8 +14,8 @@ gray                8c4850e66562a587a292dc728a65ea4a
 gray16be            daa5a6b98fb4a280c57c57bff1a2ab5a
 gray16le            84f5ea7259073edcb893113b42213c8e
 rgb24               3b90ed64b687d3dc186c6ef521dc71a8
-rgb48be             a808128041a1962deaa8620c7448feba
-rgb48le             ce92d02cc322608d5be377cb1940677b
+rgb48be             b8f9fd6aaa24d75275ee2f8b8a7b9e55
+rgb48le             3e52e831a040f086c3ae983241172cce
 rgb4_byte           6958029f73c6cdfed4f71020d816f027
 rgb555be            41a7d1836837bc90f2cae19a9c9df3b3
 rgb555le            eeb78f8ce6186fba55c941469e60ba67
@@ -29,12 +29,12 @@ yuv420p             bfea0188ddd4889787c403caae119cc7
 yuv420p16be         8365eff38b8c329aeb95fc605fa229bb
 yuv420p16le         5e8dd38d973d5854abe1ad4efad20cc1
 yuv422p             f2f930a91fe00d4252c4720b5ecd8961
-yuv422p16be         167e4338811a7d272925a4c6417d60da
-yuv422p16le         3359395d5875d581fa1e975013d30114
+yuv422p16be         6647fe1c381c148f8207c988c0e22bf0
+yuv422p16le         e1548c9dc51202db38a9625c8954203f
 yuv440p             2472417d980e395ad6843cbb8b633b29
 yuv444p             1f151980486848c96bc5585ced99003e
-yuv444p16be         d69280c2856865d2ea94bd5292aac1c6
-yuv444p16le         33f43e030bedf9723be4f63c3e9fc80e
+yuv444p16be         ac3b159f8c858fcdf475a8c024ee79b6
+yuv444p16le         9a6863bfc5faee206065c11dc994bf0c
 yuva420p            7536753dfbc7932560fb50c921369a0e
 yuvj420p            21f891093006d42d7683b0e1d773a657
 yuvj422p            9a43d474c407590ad8f213880586b45e
diff --git a/tests/ref/lavfi/pixfmts_hflip b/tests/ref/lavfi/pixfmts_hflip
index 3a3dbf0..7f4342f 100644
--- a/tests/ref/lavfi/pixfmts_hflip
+++ b/tests/ref/lavfi/pixfmts_hflip
@@ -1,8 +1,8 @@
 abgr                49468c6c9ceee5d52b08b1270a909323
 argb                50ba9f16c6475530602f2983278b82d0
 bgr24               cc53d2011d097972db0d22756c3699e3
-bgr48be             11641cf0f4516a9aed98f7872720f801
-bgr48le             b5440734eed128554dd9f83b34ba582f
+bgr48be             90374bc92471f1bd4931d71ef8b73f50
+bgr48le             696f628d0dd32121e60a0d61ac47d6e6
 bgr4_byte           aac987e7d1a6a96477cfc0b48a4285de
 bgr555be            bc07265898440116772200390d70c092
 bgr555le            ccee08679bac84a1f960c6c9070c5538
@@ -14,8 +14,8 @@ gray                03efcb4ab52a24c0af0e03cfd26c9377
 gray16be            9bcbca979601ddc4869f846f08f3d1dd
 gray16le            c1b8965adcc7f847ee343149ff507073
 rgb24               754f1722fc738590cc407ac65749bfe8
-rgb48be             10743e1577dc3198dbbc7c0b3b8f429e
-rgb48le             dd945a44f39119221407bf7a04f1bc49
+rgb48be             2397b9d3c296ac15f8a2325a703f81c7
+rgb48le             527043c72546d8b4bb1ce2dea4b294c3
 rgb4_byte           c8a3f995fcf3e0919239ea2c413ddc29
 rgb555be            045ce8607d3910586f4d97481dda8632
 rgb555le            8778ee0cf58ce9ad1d99a1eca9f95e87
@@ -29,12 +29,12 @@ yuv420p             2d5c80f9ba2ddd85b2aeda3564cc7d64
 yuv420p16be         758b0c1e2113b15e7afde48da4e4d024
 yuv420p16le         480ccd951dcb806bc875d307e02e50a0
 yuv422p             6e728f4eb9eae287c224f396d84be6ea
-yuv422p16be         a05d43cd62b790087bd37083174557de
-yuv422p16le         6954abebcbc62d81068d58d0c62bdd5b
+yuv422p16be         9dbe0af0eb877987611cf04bfa577202
+yuv422p16le         2d8f37231110177cc5e1b61c8cb4b163
 yuv440p             a99e2b57ed601f39852715c9d675d0d3
 yuv444p             947e47f7bb5fdccc659d19b7df2b6fc3
-yuv444p16be         e5ef45bc3d2f5b0b2542d5151340c382
-yuv444p16le         70793e3d66d0c23a0cdedabe9c24c2a7
+yuv444p16be         debc96a7ec4fec0a412f9d8995bc48a2
+yuv444p16le         5b5e1348a631fc2206bb7ff851a52687
 yuva420p            d83ec0c01498189f179ec574918185f1
 yuvj420p            df3aaaec3bb157c3bde5f0365af30f4f
 yuvj422p            d113871528d510a192797af59df9c05c
diff --git a/tests/ref/lavfi/pixfmts_null b/tests/ref/lavfi/pixfmts_null
index de13e94..8695b1d 100644
--- a/tests/ref/lavfi/pixfmts_null
+++ b/tests/ref/lavfi/pixfmts_null
@@ -1,8 +1,8 @@
 abgr                037bf9df6a765520ad6d490066bf4b89
 argb                c442a8261c2265a07212ef0f72e35f5a
 bgr24               0d0cb38ab3fa0b2ec0865c14f78b217b
-bgr48be             4ba0ff7fc9e011ea264610ad1585bb1f
-bgr48le             d022bfdd6a07d5dcc693799322a386b4
+bgr48be             74dedaaacae8fd1ef46e05f78cf29d62
+bgr48le             0eb7d30801eac6058814bddd330b3c76
 bgr4_byte           50d23cc82d9dcef2fd12adb81fb9b806
 bgr555be            49f01b1f1f0c84fd9e776dd34cc3c280
 bgr555le            378d6ac4223651a1adcbf94a3d0d807b
@@ -18,8 +18,8 @@ monow               9251497f3b0634f1165d12d5a289d943
 nv12                e0af357888584d36eec5aa0f673793ef
 nv21                9a3297f3b34baa038b1f37cb202b512f
 rgb24               b41eba9651e1b5fe386289b506188105
-rgb48be             460b6de89b156290a12d3941db8bd731
-rgb48le             cd93cb34d15996987367dabda3a10128
+rgb48be             e3bc84c9af376fb6d0f0293cc7b713a6
+rgb48le             f51c0e71638a822458329abb2f4052c7
 rgb4_byte           c93ba89b74c504e7f5ae9d9ab1546c73
 rgb555be            912a62c5e53bfcbac2a0340e10973cf2
 rgb555le            a937a0fc764fb57dc1b3af87cba0273c
@@ -38,12 +38,12 @@ yuv420p16le         2d59c4f1d0314a5a957a7cfc4b6fabcc
 yuv420p9be          ce880fa07830e5297c22acf6e20555ce
 yuv420p9le          16543fda8f87d94a6cf857d2e8d4461a
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
-yuv422p16be         5499502e1c29534a158a1fe60e889f60
-yuv422p16le         e3d61fde6978591596bc36b914386623
+yuv422p16be         86ad3447f97969ce095afeef81fa9abf
+yuv422p16le         a53a9b451f4a81eeae33362c1bbd07f8
 yuv440p             5a064afe2b453bb52cdb3f176b1aa1cf
 yuv444p             0a98447b78fd476aa39686da6a74fa2e
-yuv444p16be         ea602a24b8e6969679265078bd8607b6
-yuv444p16le         1262a0dc57ee147967fc896d04206313
+yuv444p16be         95db370ae765dd3d10b7def14704ae73
+yuv444p16le         36b969179b5ad9d312a0d1e1cd6bc402
 yuva420p            a29884f3f3dfe1e00b961bc17bef3d47
 yuvj420p            32eec78ba51857b16ce9b813a49b7189
 yuvj422p            0dfa0ed434f73be51428758c69e082cb
diff --git a/tests/ref/lavfi/pixfmts_scale b/tests/ref/lavfi/pixfmts_scale
index 670efe1..de41952 100644
--- a/tests/ref/lavfi/pixfmts_scale
+++ b/tests/ref/lavfi/pixfmts_scale
@@ -1,8 +1,8 @@
 abgr                d894cb97f6c80eb21bdbe8a4eea62d86
 argb                54346f2b2eef10919e0f247241df3b24
 bgr24               570f8d6b51a838aed022ef67535f6bdc
-bgr48be             fcc0f2dbf45d325f84f816c74cbeeebe
-bgr48le             3f9c2b23eed3b8d196d1c14b38ce50f5
+bgr48be             07f7a0cc34feb3646434d47c0cec8cee
+bgr48le             9abd2c3a66088e6c9078232064eba61e
 bgr4_byte           ee1d35a7baf8e9016891929a2f565c0b
 bgr555be            de8901c1358834fddea060fcb3a67beb
 bgr555le            36b745067197f9ca8c1731cac51329c9
@@ -11,15 +11,15 @@ bgr565le            3a514a298c6161a071ddf9963c06509d
 bgr8                7f007fa6c153a16e808a9c51605a4016
 bgra                a5e7040f9a80cccd65e5acf2ca09ace5
 gray                d7786a7d9d99ac74230cc045cab5632c
-gray16be            af39ce3a497f6734b157c8b94544f537
-gray16le            7ac1b788bcc472010df7a97e762485e0
+gray16be            bba98532da29a31599df2feec3b08e3e
+gray16le            30267f127d5734c4767f3944f1729a33
 monob               88c4c050758e64d120f50c7eff694381
 monow               d31772ebaa877fc2a78565937f7f9673
 nv12                4676d59db43d657dc12841f6bc3ab452
 nv21                69c699510ff1fb777b118ebee1002f14
 rgb24               514692e28e8ff6860e415ce4fcf6eb8c
-rgb48be             1894cd30dabcd3180518e4d5f09f25e7
-rgb48le             1354e6e27ce3c1d4d4989ee56030c94b
+rgb48be             f18841c19fc6d9c817a3095f557b9bc5
+rgb48le             819e7b8acd8965ba57ba46198a5cc9bf
 rgb4_byte           d81ffd3add95842a618eec81024f0b5c
 rgb555be            4607309f9f217d51cbb53d13b84b4537
 rgb555le            a350ef1dc2c9688ed49e7ba018843795
@@ -31,19 +31,19 @@ uyvy422             314bd486277111a95d9369b944fa0400
 yuv410p             7df8f6d69b56a8dcb6c7ee908e5018b5
 yuv411p             1143e7c5cc28fe0922b051b17733bc4c
 yuv420p             fdad2d8df8985e3d17e73c71f713cb14
-yuv420p10be         6d335e75b553da590135cf8bb999610c
-yuv420p10le         d510ddbabefd03ef39ec943fcb51b709
-yuv420p16be         29a0265764530070f5cd3251cc01f66a
-yuv420p16le         6f3a265b084a78baec229238d9f7945f
-yuv420p9be          ec4983b7a949c0472110a7a2c58e278a
-yuv420p9le          c136dce5913a722eee44ab72cff664b2
+yuv420p10be         8b5ad855229840a8fa87786cab83f856
+yuv420p10le         5264e87921d47b17411578d9c92672b1
+yuv420p16be         f6ef3ba90f238b467c7e72ade927083d
+yuv420p16le         faf6aab3b1c16e8afbe160686dd360e0
+yuv420p9be          c14bf746d161face61e1f39b491bf7ef
+yuv420p9le          59457f9a51768bf1d4342238d50a9be3
 yuv422p             918e37701ee7377d16a8a6c119c56a40
-yuv422p16be         ef3e865fc1d0c68977c735323c50af6e
-yuv422p16le         428a9b96214c09cb5a983ce36d6961ff
+yuv422p16be         837945d3a771366a5a72a4ed095a4f53
+yuv422p16le         b8292ae9b52eb7afc3d8b93e8fd895b4
 yuv440p             461503fdb9b90451020aa3b25ddf041c
 yuv444p             81b2eba962d12e8d64f003ac56f6faf2
-yuv444p16be         99a3738c70c8fbdc5a0e4ad4bf50648d
-yuv444p16le         385d0cc5240d62da0871915be5d86f0a
+yuv444p16be         cc7460f76477aa4b4c33442f67c06a89
+yuv444p16le         9a5ed60d68c0a4a5155f9d376174cdf7
 yuva420p            8673a9131fb47de69788863f93a50eb7
 yuvj420p            30427bd6caf5bda93a173dbebe759e09
 yuvj422p            fc8288f64fd149573f73cf8da05d8e6d
diff --git a/tests/ref/lavfi/pixfmts_vflip b/tests/ref/lavfi/pixfmts_vflip
index 0383ad9..89502d1 100644
--- a/tests/ref/lavfi/pixfmts_vflip
+++ b/tests/ref/lavfi/pixfmts_vflip
@@ -1,8 +1,8 @@
 abgr                25e72e9dbd01ab00727c976d577f7be5
 argb                19869bf1a5ac0b6af4d8bbe2c104533c
 bgr24               89108a4ba00201f79b75b9305c42352d
-bgr48be             ed82382da09b64a8e04728fcf76e6814
-bgr48le             0f1f135608c2ff24d26d03e939fc2112
+bgr48be             908b4edb525fd154a95a3744c4ab5420
+bgr48le             796c2072d6fa13a091f5c5b175417ed5
 bgr4_byte           407fcf564ed764c38e1d748f700ab921
 bgr555be            f739d2519f7e9d494359bf67a3821537
 bgr555le            bd7b3ec4d684dfad075d89a606cb8b74
@@ -18,8 +18,8 @@ monow               ff9869d067ecb94eb9d90c9750c31fea
 nv12                046f00f598ce14d9854a3534a5c99114
 nv21                01ea369dd2d0d3ed7451dc5c8d61497f
 rgb24               eaefabc168d0b14576bab45bc1e56e1e
-rgb48be             4e0c384163ebab06a08e74637beb02bc
-rgb48le             a77bfeefcd96750cf0e1917a2e2bf1e7
+rgb48be             8e347deca2902e7dc1ece261322577d8
+rgb48le             2034e485f946e4064b5fb9be09865e55
 rgb4_byte           8c6ff02df0b06dd2d574836c3741b2a2
 rgb555be            40dc33cfb5cf56aac1c5a290ac486c36
 rgb555le            4f8eaad29a17e0f8e9d8ab743e76b999
@@ -38,12 +38,12 @@ yuv420p16le         0f609e588e5a258644ef85170d70e030
 yuv420p9be          be40ec975fb2873891643cbbbddbc3b0
 yuv420p9le          7e606310d3f5ff12badf911e8f333471
 yuv422p             d7f5cb44d9b0210d66d6a8762640ab34
-yuv422p16be         9bd8f8c961822b586fa4cf992be54acc
-yuv422p16le         9c4a1239605c7952b736ac3130163f14
+yuv422p16be         8cdfbddf2dd4c44c3efef4ee00170eba
+yuv422p16le         a2f421f6a1af950544081c1797de01ae
 yuv440p             876385e96165acf51271b20e5d85a416
 yuv444p             9c3c667d1613b72d15bc6d851c5eb8f7
-yuv444p16be         0f4afa4a4aacf4bb6b87641abde71ea9
-yuv444p16le         8f31557bc52adfe00ae8b40a9b8c23f8
+yuv444p16be         b092690d22f0b26360fbf5cfd739be17
+yuv444p16le         18768b4ddca92d06f9713fef467276a9
 yuva420p            c705d1cf061d8c6580ac690b55f92276
 yuvj420p            41fd02b204da0ab62452cd14b595e2e4
 yuvj422p            7f6ca9bc1812cde02036d7d29a7cce43



More information about the ffmpeg-cvslog mailing list