swscale: fix YUV420P 9/10bit support.

Fix handling of input if not in native endianness, and add support for
9/10-bit output. This allows us to force endianness of YUV420P 9/10bit
in the H264/10bit fate tests, which should fix them on big-endian
systems.

13 files changed, 202 insertions, 65 deletions

diff --git a/libavutil/pixfmt.h b/libavutil/pixfmt.h
index 533eb9f706..d88775f462 100644
--- a/libavutil/pixfmt.h
+++ b/libavutil/pixfmt.h
@@ -128,10 +128,10 @@ enum PixelFormat {
     PIX_FMT_VDPAU_MPEG4,  ///< MPEG4 HW decoding with VDPAU, data[0] contains a vdpau_render_state struct which contains the bitstream of the slices as well as various fields extracted from headers
     PIX_FMT_DXVA2_VLD,    ///< HW decoding through DXVA2, Picture.data[3] contains a LPDIRECT3DSURFACE9 pointer
 
-    PIX_FMT_RGB444BE,  ///< packed RGB 4:4:4, 16bpp, (msb)4A 4R 4G 4B(lsb), big-endian, most significant bits to 0
     PIX_FMT_RGB444LE,  ///< packed RGB 4:4:4, 16bpp, (msb)4A 4R 4G 4B(lsb), little-endian, most significant bits to 0
-    PIX_FMT_BGR444BE,  ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), big-endian, most significant bits to 1
+    PIX_FMT_RGB444BE,  ///< packed RGB 4:4:4, 16bpp, (msb)4A 4R 4G 4B(lsb), big-endian, most significant bits to 0
     PIX_FMT_BGR444LE,  ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), little-endian, most significant bits to 1
+    PIX_FMT_BGR444BE,  ///< packed BGR 4:4:4, 16bpp, (msb)4A 4B 4G 4R(lsb), big-endian, most significant bits to 1
     PIX_FMT_Y400A,     ///< 8bit gray, 8bit alpha
     PIX_FMT_BGR48BE,   ///< packed RGB 16:16:16, 48bpp, 16B, 16G, 16R, the 2-byte value for each R/G/B component is stored as big-endian
     PIX_FMT_BGR48LE,   ///< packed RGB 16:16:16, 48bpp, 16B, 16G, 16R, the 2-byte value for each R/G/B component is stored as little-endian
diff --git a/libswscale/ppc/swscale_template.c b/libswscale/ppc/swscale_template.c
index 3e40c3f0a6..e69656ca15 100644
--- a/libswscale/ppc/swscale_template.c
+++ b/libswscale/ppc/swscale_template.c
@@ -639,7 +639,7 @@ static int RENAME(swScale)(SwsContext *c, const uint8_t* src[], int srcStride[],
             } else 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 (is16BPS(dstFormat)) {
+                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
                     yuv2yuvX16inC(
                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
@@ -716,7 +716,7 @@ static int RENAME(swScale)(SwsContext *c, const uint8_t* src[], int srcStride[],
             } 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)) {
+                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
                     yuv2yuvX16inC(
                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
diff --git a/libswscale/swscale.c b/libswscale/swscale.c
index 2830f26ce5..b63a3868c5 100644
--- a/libswscale/swscale.c
+++ b/libswscale/swscale.c
@@ -212,10 +212,11 @@ DECLARE_ALIGNED(8, const uint8_t, dither_8x8_220)[8][8]={
 static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
                                                     const int16_t *chrFilter, const int16_t **chrSrc, int chrFilterSize,
                                                     const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest,
-                                                    int dstW, int chrDstW, int big_endian)
+                                                    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;
 
     for (i = 0; i < dstW; i++) {
         int val = 1 << 10;
@@ -225,9 +226,9 @@ static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, co
             val += lumSrc[j][i] * lumFilter[j];
 
         if (big_endian) {
-            AV_WB16(&dest[i], av_clip_uint16(val >> 11));
+            AV_WB16(&dest[i], av_clip_uint16(val >> shift));
         } else {
-            AV_WL16(&dest[i], av_clip_uint16(val >> 11));
+            AV_WL16(&dest[i], av_clip_uint16(val >> shift));
         }
     }
 
@@ -243,11 +244,11 @@ static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, co
             }
 
             if (big_endian) {
-                AV_WB16(&uDest[i], av_clip_uint16(u >> 11));
-                AV_WB16(&vDest[i], av_clip_uint16(v >> 11));
+                AV_WB16(&uDest[i], av_clip_uint16(u >> shift));
+                AV_WB16(&vDest[i], av_clip_uint16(v >> shift));
             } else {
-                AV_WL16(&uDest[i], av_clip_uint16(u >> 11));
-                AV_WL16(&vDest[i], av_clip_uint16(v >> 11));
+                AV_WL16(&uDest[i], av_clip_uint16(u >> shift));
+                AV_WL16(&vDest[i], av_clip_uint16(v >> shift));
             }
         }
     }
@@ -261,9 +262,9 @@ static av_always_inline void yuv2yuvX16inC_template(const int16_t *lumFilter, co
                 val += alpSrc[j][i] * lumFilter[j];
 
             if (big_endian) {
-                AV_WB16(&aDest[i], av_clip_uint16(val >> 11));
+                AV_WB16(&aDest[i], av_clip_uint16(val >> shift));
             } else {
-                AV_WL16(&aDest[i], av_clip_uint16(val >> 11));
+                AV_WL16(&aDest[i], av_clip_uint16(val >> shift));
             }
         }
     }
@@ -274,19 +275,28 @@ static inline void yuv2yuvX16inC(const int16_t *lumFilter, const int16_t **lumSr
                                  const int16_t **alpSrc, uint16_t *dest, uint16_t *uDest, uint16_t *vDest, uint16_t *aDest, int dstW, int chrDstW,
                                  enum PixelFormat dstFormat)
 {
-    if (isBE(dstFormat)) {
-        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
-                               chrFilter, chrSrc, chrFilterSize,
-                               alpSrc,
-                               dest, uDest, vDest, aDest,
-                               dstW, chrDstW, 1);
+#define conv16(bits) \
+    if (isBE(dstFormat)) { \
+        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \
+                               chrFilter, chrSrc, chrFilterSize, \
+                               alpSrc, \
+                               dest, uDest, vDest, aDest, \
+                               dstW, chrDstW, 1, bits); \
+    } else { \
+        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize, \
+                               chrFilter, chrSrc, chrFilterSize, \
+                               alpSrc, \
+                               dest, uDest, vDest, aDest, \
+                               dstW, chrDstW, 0, bits); \
+    }
+    if (is16BPS(dstFormat)) {
+        conv16(16);
+    } else if (av_pix_fmt_descriptors[dstFormat].comp[0].depth_minus1 == 8) {
+        conv16(9);
     } else {
-        yuv2yuvX16inC_template(lumFilter, lumSrc, lumFilterSize,
-                               chrFilter, chrSrc, chrFilterSize,
-                               alpSrc,
-                               dest, uDest, vDest, aDest,
-                               dstW, chrDstW, 0);
+        conv16(10);
     }
+#undef conv16
 }
 
 static inline void yuv2yuvXinC(const int16_t *lumFilter, const int16_t **lumSrc, int lumFilterSize,
@@ -1669,25 +1679,124 @@ static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[
                 length*=2;
             fillPlane(dst[plane], dstStride[plane], length, height, y, (plane==3) ? 255 : 128);
         } else {
-            if(isNBPS(c->srcFormat)) {
-                const int depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
-                uint16_t *srcPtr2 = (uint16_t*)srcPtr;
+            if(is9_OR_10BPS(c->srcFormat)) {
+                const int src_depth = av_pix_fmt_descriptors[c->srcFormat].comp[plane].depth_minus1+1;
+                const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
+                const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
 
                 if (is16BPS(c->dstFormat)) {
                     uint16_t *dstPtr2 = (uint16_t*)dstPtr;
-                    for (i = 0; i < height; i++) {
-                        for (j = 0; j < length; j++)
-                            dstPtr2[j] = (srcPtr2[j]<<(16-depth)) | (srcPtr2[j]>>(2*depth-16));
-                        dstPtr2 += dstStride[plane]/2;
-                        srcPtr2 += srcStride[plane]/2;
+#define COPY9_OR_10TO16(rfunc, wfunc) \
+                    for (i = 0; i < height; i++) { \
+                        for (j = 0; j < length; j++) { \
+                            int srcpx = rfunc(&srcPtr2[j]); \
+                            wfunc(&dstPtr2[j], (srcpx<<(16-src_depth)) | (srcpx>>(2*src_depth-16))); \
+                        } \
+                        dstPtr2 += dstStride[plane]/2; \
+                        srcPtr2 += srcStride[plane]/2; \
+                    }
+                    if (isBE(c->dstFormat)) {
+                        if (isBE(c->srcFormat)) {
+                            COPY9_OR_10TO16(AV_RB16, AV_WB16);
+                        } else {
+                            COPY9_OR_10TO16(AV_RL16, AV_WB16);
+                        }
+                    } else {
+                        if (isBE(c->srcFormat)) {
+                            COPY9_OR_10TO16(AV_RB16, AV_WL16);
+                        } else {
+                            COPY9_OR_10TO16(AV_RL16, AV_WL16);
+                        }
+                    }
+                } else if (is9_OR_10BPS(c->dstFormat)) {
+                    uint16_t *dstPtr2 = (uint16_t*)dstPtr;
+#define COPY9_OR_10TO9_OR_10(loop) \
+                    for (i = 0; i < height; i++) { \
+                        for (j = 0; j < length; j++) { \
+                            loop; \
+                        } \
+                        dstPtr2 += dstStride[plane]/2; \
+                        srcPtr2 += srcStride[plane]/2; \
+                    }
+#define COPY9_OR_10TO9_OR_10_2(rfunc, wfunc) \
+                    if (dst_depth > src_depth) { \
+                        COPY9_OR_10TO9_OR_10(int srcpx = rfunc(&srcPtr2[j]); \
+                            wfunc(&dstPtr2[j], (srcpx << 1) | (srcpx >> 9))); \
+                    } else if (dst_depth < src_depth) { \
+                        COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]) >> 1)); \
+                    } else { \
+                        COPY9_OR_10TO9_OR_10(wfunc(&dstPtr2[j], rfunc(&srcPtr2[j]))); \
+                    }
+                    if (isBE(c->dstFormat)) {
+                        if (isBE(c->srcFormat)) {
+                            COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WB16);
+                        } else {
+                            COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WB16);
+                        }
+                    } else {
+                        if (isBE(c->srcFormat)) {
+                            COPY9_OR_10TO9_OR_10_2(AV_RB16, AV_WL16);
+                        } else {
+                            COPY9_OR_10TO9_OR_10_2(AV_RL16, AV_WL16);
+                        }
                     }
                 } else {
                     // FIXME Maybe dither instead.
-                    for (i = 0; i < height; i++) {
-                        for (j = 0; j < length; j++)
-                            dstPtr[j] = srcPtr2[j]>>(depth-8);
-                        dstPtr  += dstStride[plane];
-                        srcPtr2 += srcStride[plane]/2;
+#define COPY9_OR_10TO8(rfunc) \
+                    for (i = 0; i < height; i++) { \
+                        for (j = 0; j < length; j++) { \
+                            dstPtr[j] = rfunc(&srcPtr2[j])>>(src_depth-8); \
+                        } \
+                        dstPtr  += dstStride[plane]; \
+                        srcPtr2 += srcStride[plane]/2; \
+                    }
+                    if (isBE(c->srcFormat)) {
+                        COPY9_OR_10TO8(AV_RB16);
+                    } else {
+                        COPY9_OR_10TO8(AV_RL16);
+                    }
+                }
+            } else if(is9_OR_10BPS(c->dstFormat)) {
+                const int dst_depth = av_pix_fmt_descriptors[c->dstFormat].comp[plane].depth_minus1+1;
+                uint16_t *dstPtr2 = (uint16_t*)dstPtr;
+
+                if (is16BPS(c->srcFormat)) {
+                    const uint16_t *srcPtr2 = (const uint16_t*)srcPtr;
+#define COPY16TO9_OR_10(rfunc, wfunc) \
+                    for (i = 0; i < height; i++) { \
+                        for (j = 0; j < length; j++) { \
+                            wfunc(&dstPtr2[j], rfunc(&srcPtr2[j])>>(16-dst_depth)); \
+                        } \
+                        dstPtr2 += dstStride[plane]/2; \
+                        srcPtr2 += srcStride[plane]/2; \
+                    }
+                    if (isBE(c->dstFormat)) {
+                        if (isBE(c->srcFormat)) {
+                            COPY16TO9_OR_10(AV_RB16, AV_WB16);
+                        } else {
+                            COPY16TO9_OR_10(AV_RL16, AV_WB16);
+                        }
+                    } else {
+                        if (isBE(c->srcFormat)) {
+                            COPY16TO9_OR_10(AV_RB16, AV_WL16);
+                        } else {
+                            COPY16TO9_OR_10(AV_RL16, AV_WL16);
+                        }
+                    }
+                } else /* 8bit */ {
+#define COPY8TO9_OR_10(wfunc) \
+                    for (i = 0; i < height; i++) { \
+                        for (j = 0; j < length; j++) { \
+                            const int srcpx = srcPtr[j]; \
+                            wfunc(&dstPtr2[j], (srcpx<<(dst_depth-8)) | (srcpx >> (16-dst_depth))); \
+                        } \
+                        dstPtr2 += dstStride[plane]/2; \
+                        srcPtr  += srcStride[plane]; \
+                    }
+                    if (isBE(c->dstFormat)) {
+                        COPY8TO9_OR_10(AV_WB16);
+                    } else {
+                        COPY8TO9_OR_10(AV_WL16);
                     }
                 }
             } else if(is16BPS(c->srcFormat) && !is16BPS(c->dstFormat)) {
diff --git a/libswscale/swscale_internal.h b/libswscale/swscale_internal.h
index 2d40215ea9..2369546cb7 100644
--- a/libswscale/swscale_internal.h
+++ b/libswscale/swscale_internal.h
@@ -354,7 +354,7 @@ const char *sws_format_name(enum PixelFormat format);
         || (x)==PIX_FMT_YUV422P16BE \
         || (x)==PIX_FMT_YUV444P16BE \
     )
-#define isNBPS(x)       (           \
+#define is9_OR_10BPS(x) (           \
            (x)==PIX_FMT_YUV420P9LE  \
         || (x)==PIX_FMT_YUV420P9BE  \
         || (x)==PIX_FMT_YUV420P10LE \
diff --git a/libswscale/swscale_template.c b/libswscale/swscale_template.c
index 81a8d66277..fe872561cb 100644
--- a/libswscale/swscale_template.c
+++ b/libswscale/swscale_template.c
@@ -164,6 +164,8 @@ static inline void LEToUV_c(uint8_t *dstU, uint8_t *dstV, const uint8_t *src1,
                             const uint8_t *src2, long width, uint32_t *unused)
 {
     int i;
+    // FIXME I don't think this code is right for YUV444/422, since then h is not subsampled so
+    // we need to skip each second pixel. Same for BEToUV.
     for (i=0; i<width; i++) {
         dstU[i]= src1[2*i + 1];
         dstV[i]= src2[2*i + 1];
@@ -226,8 +228,8 @@ static inline void nv21ToUV_c(uint8_t *dstU, uint8_t *dstV,
 }
 
 // FIXME Maybe dither instead.
-#define YUV_NBPS(depth) \
-static inline void yuv ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
+#define YUV_NBPS(depth, endianness, rfunc) \
+static inline void endianness ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
                                           const uint8_t *_srcU, const uint8_t *_srcV, \
                                           long width, uint32_t *unused) \
 { \
@@ -235,21 +237,23 @@ static inline void yuv ## depth ## ToUV_c(uint8_t *dstU, uint8_t *dstV, \
     const uint16_t *srcU = (const uint16_t*)_srcU; \
     const uint16_t *srcV = (const uint16_t*)_srcV; \
     for (i = 0; i < width; i++) { \
-        dstU[i] = srcU[i]>>(depth-8); \
-        dstV[i] = srcV[i]>>(depth-8); \
+        dstU[i] = rfunc(&srcU[i])>>(depth-8); \
+        dstV[i] = rfunc(&srcV[i])>>(depth-8); \
     } \
 } \
 \
-static inline void yuv ## depth ## ToY_c(uint8_t *dstY, const uint8_t *_srcY, long width, uint32_t *unused) \
+static inline void endianness ## depth ## ToY_c(uint8_t *dstY, const uint8_t *_srcY, long width, uint32_t *unused) \
 { \
     int i; \
     const uint16_t *srcY = (const uint16_t*)_srcY; \
     for (i = 0; i < width; i++) \
-        dstY[i] = srcY[i]>>(depth-8); \
+        dstY[i] = rfunc(&srcY[i])>>(depth-8); \
 } \
 
-YUV_NBPS( 9)
-YUV_NBPS(10)
+YUV_NBPS( 9, LE, AV_RL16)
+YUV_NBPS( 9, BE, AV_RB16)
+YUV_NBPS(10, LE, AV_RL16)
+YUV_NBPS(10, BE, AV_RB16)
 
 static inline void bgr24ToY_c(uint8_t *dst, const uint8_t *src,
                               long width, uint32_t *unused)
@@ -666,7 +670,7 @@ static int swScale_c(SwsContext *c, const uint8_t* src[], int srcStride[],
             } else 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 (is16BPS(dstFormat)) {
+                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
                     yuv2yuvX16inC(
                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
@@ -743,7 +747,7 @@ static int swScale_c(SwsContext *c, const uint8_t* src[], int srcStride[],
             } 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)) {
+                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
                     yuv2yuvX16inC(
                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
@@ -816,10 +820,10 @@ 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; break;
-        case PIX_FMT_YUV420P9BE:
-        case PIX_FMT_YUV420P9LE: c->chrToYV12 = yuv9ToUV_c; break;
-        case PIX_FMT_YUV420P10BE:
-        case PIX_FMT_YUV420P10LE: c->chrToYV12 = yuv10ToUV_c; break;
+        case PIX_FMT_YUV420P9BE: c->chrToYV12 = BE9ToUV_c; break;
+        case PIX_FMT_YUV420P9LE: c->chrToYV12 = LE9ToUV_c; break;
+        case PIX_FMT_YUV420P10BE: c->chrToYV12 = BE10ToUV_c; break;
+        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;
@@ -866,10 +870,10 @@ static void sws_init_swScale_c(SwsContext *c)
     c->lumToYV12 = NULL;
     c->alpToYV12 = NULL;
     switch (srcFormat) {
-    case PIX_FMT_YUV420P9BE:
-    case PIX_FMT_YUV420P9LE: c->lumToYV12 = yuv9ToY_c; break;
-    case PIX_FMT_YUV420P10BE:
-    case PIX_FMT_YUV420P10LE: c->lumToYV12 = yuv10ToY_c; break;
+    case PIX_FMT_YUV420P9BE: c->lumToYV12 = BE9ToY_c; break;
+    case PIX_FMT_YUV420P9LE: c->lumToYV12 = LE9ToY_c; break;
+    case PIX_FMT_YUV420P10BE: c->lumToYV12 = BE10ToY_c; break;
+    case PIX_FMT_YUV420P10LE: c->lumToYV12 = LE10ToY_c; break;
     case PIX_FMT_YUYV422  :
     case PIX_FMT_YUV420P16BE:
     case PIX_FMT_YUV422P16BE:
diff --git a/libswscale/utils.c b/libswscale/utils.c
index 4f9f269731..818d014159 100644
--- a/libswscale/utils.c
+++ b/libswscale/utils.c
@@ -143,9 +143,13 @@ int sws_isSupportedInput(enum PixelFormat pix_fmt)
         || (x)==PIX_FMT_GRAY8       \
         || (x)==PIX_FMT_YUV410P     \
         || (x)==PIX_FMT_YUV440P     \
+        || (x)==PIX_FMT_YUV420P9LE    \
+        || (x)==PIX_FMT_YUV420P10LE   \
         || (x)==PIX_FMT_YUV420P16LE   \
         || (x)==PIX_FMT_YUV422P16LE   \
         || (x)==PIX_FMT_YUV444P16LE   \
+        || (x)==PIX_FMT_YUV420P9BE    \
+        || (x)==PIX_FMT_YUV420P10BE   \
         || (x)==PIX_FMT_YUV420P16BE   \
         || (x)==PIX_FMT_YUV422P16BE   \
         || (x)==PIX_FMT_YUV444P16BE   \
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index b1009cb4e9..d719721693 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -2611,7 +2611,7 @@ static int RENAME(swScale)(SwsContext *c, const uint8_t* src[], int srcStride[],
             } else 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 (is16BPS(dstFormat)) {
+                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
                     yuv2yuvX16inC(
                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
@@ -2688,7 +2688,7 @@ static int RENAME(swScale)(SwsContext *c, const uint8_t* src[], int srcStride[],
             } 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)) {
+                if (is16BPS(dstFormat) || is9_OR_10BPS(dstFormat)) {
                     yuv2yuvX16inC(
                                   vLumFilter+dstY*vLumFilterSize   , lumSrcPtr, vLumFilterSize,
                                   vChrFilter+chrDstY*vChrFilterSize, chrSrcPtr, vChrFilterSize,
diff --git a/tests/fate/h264.mak b/tests/fate/h264.mak
index 9cb669c1a0..5e6ff50e84 100644
--- a/tests/fate/h264.mak
+++ b/tests/fate/h264.mak
@@ -308,13 +308,13 @@ fate-h264-conformance-frext-hpcvfl_bcrm_a: CMD = framecrc  -i $(SAMPLES)/h264-co
 fate-h264-conformance-frext-hpcvflnl_bcrm_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/HPCVFLNL_BRCM_A.264 -vsync 0
 fate-h264-conformance-frext-hpcvmolq_brcm_b: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/HPCVMOLQ_BRCM_B.264
 fate-h264-conformance-frext-hpcvnl_brcm_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/HPCVNL_BRCM_A.264
-fate-h264-conformance-frext-pph10i1_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I1_Panasonic_A.264
-fate-h264-conformance-frext-pph10i2_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I2_Panasonic_A.264
-fate-h264-conformance-frext-pph10i3_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I3_Panasonic_A.264
-fate-h264-conformance-frext-pph10i4_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I4_Panasonic_A.264
-fate-h264-conformance-frext-pph10i5_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I5_Panasonic_A.264
-fate-h264-conformance-frext-pph10i6_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I6_Panasonic_A.264
-fate-h264-conformance-frext-pph10i7_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I7_Panasonic_A.264
+fate-h264-conformance-frext-pph10i1_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I1_Panasonic_A.264 -pix_fmt yuv420p10le
+fate-h264-conformance-frext-pph10i2_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I2_Panasonic_A.264 -pix_fmt yuv420p10le
+fate-h264-conformance-frext-pph10i3_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I3_Panasonic_A.264 -pix_fmt yuv420p10le
+fate-h264-conformance-frext-pph10i4_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I4_Panasonic_A.264 -pix_fmt yuv420p10le
+fate-h264-conformance-frext-pph10i5_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I5_Panasonic_A.264 -pix_fmt yuv420p10le
+fate-h264-conformance-frext-pph10i6_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I6_Panasonic_A.264 -pix_fmt yuv420p10le
+fate-h264-conformance-frext-pph10i7_panasonic_a: CMD = framecrc  -i $(SAMPLES)/h264-conformance/FRext/PPH10I7_Panasonic_A.264 -pix_fmt yuv420p10le
 fate-h264-conformance-hcbp2_hhi_a: CMD = framecrc  -vsync 0 -strict 1 -i $(SAMPLES)/h264-conformance/HCBP2_HHI_A.264
 fate-h264-conformance-hcmp1_hhi_a: CMD = framecrc  -vsync 0 -strict 1 -i $(SAMPLES)/h264-conformance/HCMP1_HHI_A.264
 fate-h264-conformance-ls_sva_d: CMD = framecrc  -i $(SAMPLES)/h264-conformance/LS_SVA_D.264
diff --git a/tests/ref/lavfi/pixdesc_le b/tests/ref/lavfi/pixdesc_le
index 860f97d5b6..2078ae1abc 100644
--- a/tests/ref/lavfi/pixdesc_le
+++ b/tests/ref/lavfi/pixdesc_le
@@ -27,8 +27,12 @@ uyvy422             adcf64516a19fce44df77082bdb16291
 yuv410p             2d9225153c83ee1132397d619d94d1b3
 yuv411p             8b298af3e43348ca1b11eb8a3252ac6c
 yuv420p             eba2f135a08829387e2f698ff72a2939
+yuv420p10be         7605e266c088d0fcf68c7b27c3ceff5f
+yuv420p10le         4228ee628c6deec123a13b9784516cc7
 yuv420p16be         16c009a235cd52b74791a895423152a3
 yuv420p16le         2d59c4f1d0314a5a957a7cfc4b6fabcc
+yuv420p9be          ce880fa07830e5297c22acf6e20555ce
+yuv420p9le          16543fda8f87d94a6cf857d2e8d4461a
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
 yuv422p16be         5499502e1c29534a158a1fe60e889f60
 yuv422p16le         e3d61fde6978591596bc36b914386623
diff --git a/tests/ref/lavfi/pixfmts_copy_le b/tests/ref/lavfi/pixfmts_copy_le
index 860f97d5b6..2078ae1abc 100644
--- a/tests/ref/lavfi/pixfmts_copy_le
+++ b/tests/ref/lavfi/pixfmts_copy_le
@@ -27,8 +27,12 @@ uyvy422             adcf64516a19fce44df77082bdb16291
 yuv410p             2d9225153c83ee1132397d619d94d1b3
 yuv411p             8b298af3e43348ca1b11eb8a3252ac6c
 yuv420p             eba2f135a08829387e2f698ff72a2939
+yuv420p10be         7605e266c088d0fcf68c7b27c3ceff5f
+yuv420p10le         4228ee628c6deec123a13b9784516cc7
 yuv420p16be         16c009a235cd52b74791a895423152a3
 yuv420p16le         2d59c4f1d0314a5a957a7cfc4b6fabcc
+yuv420p9be          ce880fa07830e5297c22acf6e20555ce
+yuv420p9le          16543fda8f87d94a6cf857d2e8d4461a
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
 yuv422p16be         5499502e1c29534a158a1fe60e889f60
 yuv422p16le         e3d61fde6978591596bc36b914386623
diff --git a/tests/ref/lavfi/pixfmts_null_le b/tests/ref/lavfi/pixfmts_null_le
index 860f97d5b6..2078ae1abc 100644
--- a/tests/ref/lavfi/pixfmts_null_le
+++ b/tests/ref/lavfi/pixfmts_null_le
@@ -27,8 +27,12 @@ uyvy422             adcf64516a19fce44df77082bdb16291
 yuv410p             2d9225153c83ee1132397d619d94d1b3
 yuv411p             8b298af3e43348ca1b11eb8a3252ac6c
 yuv420p             eba2f135a08829387e2f698ff72a2939
+yuv420p10be         7605e266c088d0fcf68c7b27c3ceff5f
+yuv420p10le         4228ee628c6deec123a13b9784516cc7
 yuv420p16be         16c009a235cd52b74791a895423152a3
 yuv420p16le         2d59c4f1d0314a5a957a7cfc4b6fabcc
+yuv420p9be          ce880fa07830e5297c22acf6e20555ce
+yuv420p9le          16543fda8f87d94a6cf857d2e8d4461a
 yuv422p             c9bba4529821d796a6ab09f6a5fd355a
 yuv422p16be         5499502e1c29534a158a1fe60e889f60
 yuv422p16le         e3d61fde6978591596bc36b914386623
diff --git a/tests/ref/lavfi/pixfmts_scale_le b/tests/ref/lavfi/pixfmts_scale_le
index 49f68a82e4..cbccb41cbd 100644
--- a/tests/ref/lavfi/pixfmts_scale_le
+++ b/tests/ref/lavfi/pixfmts_scale_le
@@ -27,8 +27,12 @@ uyvy422             314bd486277111a95d9369b944fa0400
 yuv410p             7df8f6d69b56a8dcb6c7ee908e5018b5
 yuv411p             1143e7c5cc28fe0922b051b17733bc4c
 yuv420p             fdad2d8df8985e3d17e73c71f713cb14
+yuv420p10be         5051128ca208d89595f7672b1707340b
+yuv420p10le         77895bf65e70ad2ca021702fff55c8fc
 yuv420p16be         29a0265764530070f5cd3251cc01f66a
 yuv420p16le         6f3a265b084a78baec229238d9f7945f
+yuv420p9be          02de6b37dc8a631ce2367b535670c40c
+yuv420p9le          cddfbaf8e2a61aa5ea09fb396bcbc872
 yuv422p             918e37701ee7377d16a8a6c119c56a40
 yuv422p16be         ef3e865fc1d0c68977c735323c50af6e
 yuv422p16le         428a9b96214c09cb5a983ce36d6961ff
diff --git a/tests/ref/lavfi/pixfmts_vflip_le b/tests/ref/lavfi/pixfmts_vflip_le
index 72fde29f56..698921a2a2 100644
--- a/tests/ref/lavfi/pixfmts_vflip_le
+++ b/tests/ref/lavfi/pixfmts_vflip_le
@@ -27,8 +27,12 @@ uyvy422             ffbd36720c77398d9a0d03ce2625928f
 yuv410p             7bfb39d7afb49d6a6173e6b23ae321eb
 yuv411p             4a90048cc3a65fac150e53289700efe1
 yuv420p             2e6d6062e8cad37fb3ab2c433b55f382
+yuv420p10be         df97d20b3b4a10c174d4360552c4160d
+yuv420p10le         4b5249208602b941332945c926f80ae9
 yuv420p16be         539076782902664a8acf381bf4f713e8
 yuv420p16le         0f609e588e5a258644ef85170d70e030
+yuv420p9be          be40ec975fb2873891643cbbbddbc3b0
+yuv420p9le          7e606310d3f5ff12badf911e8f333471
 yuv422p             d7f5cb44d9b0210d66d6a8762640ab34
 yuv422p16be         9bd8f8c961822b586fa4cf992be54acc
 yuv422p16le         9c4a1239605c7952b736ac3130163f14