H420 functionality

R=harryjin@google.com
BUG=libyuv:488

Review URL: https://webrtc-codereview.appspot.com/54869004 .

README.chromium

 Name: libyuv
 URL: http://code.google.com/p/libyuv/
-Version: 1478
+Version: 1479
 License: BSD
 License File: LICENSE
 

include/libyuv/convert_argb.h

@@ -165,6 +165,38 @@ int J422ToABGR(const uint8* src_y, int src_stride_y,
                uint8* dst_abgr, int dst_stride_abgr,
                int width, int height);
 
+// Convert H420 to ARGB.
+LIBYUV_API
+int H420ToARGB(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height);
+
+// Convert H422 to ARGB.
+LIBYUV_API
+int H422ToARGB(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height);
+
+// Convert H420 to ABGR.
+LIBYUV_API
+int H420ToABGR(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_abgr, int dst_stride_abgr,
+               int width, int height);
+
+// Convert H422 to ABGR.
+LIBYUV_API
+int H422ToABGR(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_abgr, int dst_stride_abgr,
+               int width, int height);
+
 // BGRA little endian (argb in memory) to ARGB.
 LIBYUV_API
 int BGRAToARGB(const uint8* src_frame, int src_stride_frame,

include/libyuv/row.h

@@ -101,6 +101,8 @@ extern "C" {
 #define HAS_J400TOARGBROW_SSE2
 #define HAS_J422TOARGBROW_SSSE3
 #define HAS_J422TOABGRROW_SSSE3
+#define HAS_H422TOARGBROW_SSSE3
+#define HAS_H422TOABGRROW_SSSE3
 #define HAS_MERGEUVROW_SSE2
 #define HAS_MIRRORROW_SSE2
 #define HAS_MIRRORROW_SSSE3
@@ -238,6 +240,8 @@ extern "C" {
 #define HAS_INTERPOLATEROW_AVX2
 #define HAS_J422TOARGBROW_AVX2
 #define HAS_J422TOABGRROW_AVX2
+#define HAS_H422TOARGBROW_AVX2
+#define HAS_H422TOABGRROW_AVX2
 #define HAS_MERGEUVROW_AVX2
 #define HAS_MIRRORROW_AVX2
 #define HAS_SPLITUVROW_AVX2
@@ -1047,6 +1051,16 @@ void J422ToABGRRow_C(const uint8* src_y,
                      const uint8* src_v,
                      uint8* dst_argb,
                      int width);
+void H422ToARGBRow_C(const uint8* src_y,
+                     const uint8* src_u,
+                     const uint8* src_v,
+                     uint8* dst_argb,
+                     int width);
+void H422ToABGRRow_C(const uint8* src_y,
+                     const uint8* src_u,
+                     const uint8* src_v,
+                     uint8* dst_argb,
+                     int width);
 void I422ToBGRARow_C(const uint8* src_y,
                      const uint8* src_u,
                      const uint8* src_v,
@@ -1220,6 +1234,26 @@ void J422ToABGRRow_AVX2(const uint8* src_y,
                         const uint8* src_v,
                         uint8* dst_argb,
                         int width);
+void H422ToARGBRow_SSSE3(const uint8* src_y,
+                         const uint8* src_u,
+                         const uint8* src_v,
+                         uint8* dst_argb,
+                         int width);
+void H422ToABGRRow_SSSE3(const uint8* src_y,
+                         const uint8* src_u,
+                         const uint8* src_v,
+                         uint8* dst_argb,
+                         int width);
+void H422ToARGBRow_AVX2(const uint8* src_y,
+                        const uint8* src_u,
+                        const uint8* src_v,
+                        uint8* dst_argb,
+                        int width);
+void H422ToABGRRow_AVX2(const uint8* src_y,
+                        const uint8* src_u,
+                        const uint8* src_v,
+                        uint8* dst_argb,
+                        int width);
 void I422ToBGRARow_SSSE3(const uint8* src_y,
                          const uint8* src_u,
                          const uint8* src_v,
@@ -1394,6 +1428,26 @@ void J422ToABGRRow_Any_AVX2(const uint8* src_y,
                             const uint8* src_v,
                             uint8* dst_argb,
                             int width);
+void H422ToARGBRow_Any_SSSE3(const uint8* src_y,
+                             const uint8* src_u,
+                             const uint8* src_v,
+                             uint8* dst_argb,
+                             int width);
+void H422ToABGRRow_Any_SSSE3(const uint8* src_y,
+                             const uint8* src_u,
+                             const uint8* src_v,
+                             uint8* dst_argb,
+                             int width);
+void H422ToARGBRow_Any_AVX2(const uint8* src_y,
+                            const uint8* src_u,
+                            const uint8* src_v,
+                            uint8* dst_argb,
+                            int width);
+void H422ToABGRRow_Any_AVX2(const uint8* src_y,
+                            const uint8* src_u,
+                            const uint8* src_v,
+                            uint8* dst_argb,
+                            int width);
 void I422ToBGRARow_Any_SSSE3(const uint8* src_y,
                              const uint8* src_u,
                              const uint8* src_v,

include/libyuv/version.h

@@ -11,6 +11,6 @@
 #ifndef INCLUDE_LIBYUV_VERSION_H_  // NOLINT
 #define INCLUDE_LIBYUV_VERSION_H_
 
-#define LIBYUV_VERSION 1478
+#define LIBYUV_VERSION 1479
 
 #endif  // INCLUDE_LIBYUV_VERSION_H_  NOLINT

include/libyuv/video_common.h

@@ -62,7 +62,7 @@ enum FourCC {
 
   // 2 Secondary YUV formats: row biplanar.
   FOURCC_M420 = FOURCC('M', '4', '2', '0'),
-  FOURCC_Q420 = FOURCC('Q', '4', '2', '0'), // deprecated.
+  FOURCC_Q420 = FOURCC('Q', '4', '2', '0'),  // deprecated.
 
   // 9 Primary RGB formats: 4 32 bpp, 2 24 bpp, 3 16 bpp.
   FOURCC_ARGB = FOURCC('A', 'R', 'G', 'B'),
@@ -90,7 +90,8 @@ enum FourCC {
   FOURCC_YV24 = FOURCC('Y', 'V', '2', '4'),
   FOURCC_YU12 = FOURCC('Y', 'U', '1', '2'),  // Linux version of I420.
   FOURCC_J420 = FOURCC('J', '4', '2', '0'),
-  FOURCC_J400 = FOURCC('J', '4', '0', '0'),
+  FOURCC_J400 = FOURCC('J', '4', '0', '0'),  // unofficial fourcc
+  FOURCC_H420 = FOURCC('H', '4', '2', '0'),  // unofficial fourcc
 
   // 14 Auxiliary aliases.  CanonicalFourCC() maps these to canonical fourcc.
   FOURCC_IYUV = FOURCC('I', 'Y', 'U', 'V'),  // Alias for I420.
@@ -150,6 +151,7 @@ enum FourCCBpp {
   FOURCC_BPP_YU12 = 12,
   FOURCC_BPP_J420 = 12,
   FOURCC_BPP_J400 = 8,
+  FOURCC_BPP_H420 = 12,
   FOURCC_BPP_MJPG = 0,  // 0 means unknown.
   FOURCC_BPP_H264 = 0,
   FOURCC_BPP_IYUV = 12,

source/convert_argb.cc

@@ -1544,6 +1544,298 @@ int J422ToABGR(const uint8* src_y, int src_stride_y,
   return 0;
 }
 
+// Convert H420 to ARGB.
+LIBYUV_API
+int H420ToARGB(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height) {
+  int y;
+  void (*H422ToARGBRow)(const uint8* y_buf,
+                        const uint8* u_buf,
+                        const uint8* v_buf,
+                        uint8* rgb_buf,
+                        int width) = H422ToARGBRow_C;
+  if (!src_y || !src_u || !src_v || !dst_argb ||
+      width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+#if defined(HAS_H422TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    H422ToARGBRow = H422ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToARGBRow = H422ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_H422TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    H422ToARGBRow = H422ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      H422ToARGBRow = H422ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_H422TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    H422ToARGBRow = H422ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToARGBRow = H422ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_H422TOARGBROW_MIPS_DSPR2)
+  if (TestCpuFlag(kCpuHasMIPS_DSPR2) && IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_y, 4) && IS_ALIGNED(src_stride_y, 4) &&
+      IS_ALIGNED(src_u, 2) && IS_ALIGNED(src_stride_u, 2) &&
+      IS_ALIGNED(src_v, 2) && IS_ALIGNED(src_stride_v, 2) &&
+      IS_ALIGNED(dst_argb, 4) && IS_ALIGNED(dst_stride_argb, 4)) {
+    H422ToARGBRow = H422ToARGBRow_MIPS_DSPR2;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    H422ToARGBRow(src_y, src_u, src_v, dst_argb, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert H422 to ARGB.
+LIBYUV_API
+int H422ToARGB(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_argb, int dst_stride_argb,
+               int width, int height) {
+  int y;
+  void (*H422ToARGBRow)(const uint8* y_buf,
+                        const uint8* u_buf,
+                        const uint8* v_buf,
+                        uint8* rgb_buf,
+                        int width) = H422ToARGBRow_C;
+  if (!src_y || !src_u || !src_v ||
+      !dst_argb ||
+      width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_argb = dst_argb + (height - 1) * dst_stride_argb;
+    dst_stride_argb = -dst_stride_argb;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width &&
+      src_stride_u * 2 == width &&
+      src_stride_v * 2 == width &&
+      dst_stride_argb == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_argb = 0;
+  }
+#if defined(HAS_H422TOARGBROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    H422ToARGBRow = H422ToARGBRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToARGBRow = H422ToARGBRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_H422TOARGBROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    H422ToARGBRow = H422ToARGBRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      H422ToARGBRow = H422ToARGBRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_H422TOARGBROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    H422ToARGBRow = H422ToARGBRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToARGBRow = H422ToARGBRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_H422TOARGBROW_MIPS_DSPR2)
+  if (TestCpuFlag(kCpuHasMIPS_DSPR2) && IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_y, 4) && IS_ALIGNED(src_stride_y, 4) &&
+      IS_ALIGNED(src_u, 2) && IS_ALIGNED(src_stride_u, 2) &&
+      IS_ALIGNED(src_v, 2) && IS_ALIGNED(src_stride_v, 2) &&
+      IS_ALIGNED(dst_argb, 4) && IS_ALIGNED(dst_stride_argb, 4)) {
+    H422ToARGBRow = H422ToARGBRow_MIPS_DSPR2;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    H422ToARGBRow(src_y, src_u, src_v, dst_argb, width);
+    dst_argb += dst_stride_argb;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
+// Convert H420 to ABGR.
+LIBYUV_API
+int H420ToABGR(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_abgr, int dst_stride_abgr,
+               int width, int height) {
+  int y;
+  void (*H422ToABGRRow)(const uint8* y_buf,
+                        const uint8* u_buf,
+                        const uint8* v_buf,
+                        uint8* rgb_buf,
+                        int width) = H422ToABGRRow_C;
+  if (!src_y || !src_u || !src_v || !dst_abgr ||
+      width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_abgr = dst_abgr + (height - 1) * dst_stride_abgr;
+    dst_stride_abgr = -dst_stride_abgr;
+  }
+#if defined(HAS_H422TOABGRROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    H422ToABGRRow = H422ToABGRRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToABGRRow = H422ToABGRRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_H422TOABGRROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    H422ToABGRRow = H422ToABGRRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      H422ToABGRRow = H422ToABGRRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_H422TOABGRROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    H422ToABGRRow = H422ToABGRRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToABGRRow = H422ToABGRRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_H422TOABGRROW_MIPS_DSPR2)
+  if (TestCpuFlag(kCpuHasMIPS_DSPR2) && IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_y, 4) && IS_ALIGNED(src_stride_y, 4) &&
+      IS_ALIGNED(src_u, 2) && IS_ALIGNED(src_stride_u, 2) &&
+      IS_ALIGNED(src_v, 2) && IS_ALIGNED(src_stride_v, 2) &&
+      IS_ALIGNED(dst_abgr, 4) && IS_ALIGNED(dst_stride_abgr, 4)) {
+    H422ToABGRRow = H422ToABGRRow_MIPS_DSPR2;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    H422ToABGRRow(src_y, src_u, src_v, dst_abgr, width);
+    dst_abgr += dst_stride_abgr;
+    src_y += src_stride_y;
+    if (y & 1) {
+      src_u += src_stride_u;
+      src_v += src_stride_v;
+    }
+  }
+  return 0;
+}
+
+// Convert H422 to ABGR.
+LIBYUV_API
+int H422ToABGR(const uint8* src_y, int src_stride_y,
+               const uint8* src_u, int src_stride_u,
+               const uint8* src_v, int src_stride_v,
+               uint8* dst_abgr, int dst_stride_abgr,
+               int width, int height) {
+  int y;
+  void (*H422ToABGRRow)(const uint8* y_buf,
+                        const uint8* u_buf,
+                        const uint8* v_buf,
+                        uint8* rgb_buf,
+                        int width) = H422ToABGRRow_C;
+  if (!src_y || !src_u || !src_v ||
+      !dst_abgr ||
+      width <= 0 || height == 0) {
+    return -1;
+  }
+  // Negative height means invert the image.
+  if (height < 0) {
+    height = -height;
+    dst_abgr = dst_abgr + (height - 1) * dst_stride_abgr;
+    dst_stride_abgr = -dst_stride_abgr;
+  }
+  // Coalesce rows.
+  if (src_stride_y == width &&
+      src_stride_u * 2 == width &&
+      src_stride_v * 2 == width &&
+      dst_stride_abgr == width * 4) {
+    width *= height;
+    height = 1;
+    src_stride_y = src_stride_u = src_stride_v = dst_stride_abgr = 0;
+  }
+#if defined(HAS_H422TOABGRROW_SSSE3)
+  if (TestCpuFlag(kCpuHasSSSE3)) {
+    H422ToABGRRow = H422ToABGRRow_Any_SSSE3;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToABGRRow = H422ToABGRRow_SSSE3;
+    }
+  }
+#endif
+#if defined(HAS_H422TOABGRROW_AVX2)
+  if (TestCpuFlag(kCpuHasAVX2)) {
+    H422ToABGRRow = H422ToABGRRow_Any_AVX2;
+    if (IS_ALIGNED(width, 16)) {
+      H422ToABGRRow = H422ToABGRRow_AVX2;
+    }
+  }
+#endif
+#if defined(HAS_H422TOABGRROW_NEON)
+  if (TestCpuFlag(kCpuHasNEON)) {
+    H422ToABGRRow = H422ToABGRRow_Any_NEON;
+    if (IS_ALIGNED(width, 8)) {
+      H422ToABGRRow = H422ToABGRRow_NEON;
+    }
+  }
+#endif
+#if defined(HAS_H422TOABGRROW_MIPS_DSPR2)
+  if (TestCpuFlag(kCpuHasMIPS_DSPR2) && IS_ALIGNED(width, 4) &&
+      IS_ALIGNED(src_y, 4) && IS_ALIGNED(src_stride_y, 4) &&
+      IS_ALIGNED(src_u, 2) && IS_ALIGNED(src_stride_u, 2) &&
+      IS_ALIGNED(src_v, 2) && IS_ALIGNED(src_stride_v, 2) &&
+      IS_ALIGNED(dst_abgr, 4) && IS_ALIGNED(dst_stride_abgr, 4)) {
+    H422ToABGRRow = H422ToABGRRow_MIPS_DSPR2;
+  }
+#endif
+
+  for (y = 0; y < height; ++y) {
+    H422ToABGRRow(src_y, src_u, src_v, dst_abgr, width);
+    dst_abgr += dst_stride_abgr;
+    src_y += src_stride_y;
+    src_u += src_stride_u;
+    src_v += src_stride_v;
+  }
+  return 0;
+}
+
 #ifdef __cplusplus
 }  // extern "C"
 }  // namespace libyuv

source/row_any.cc

@@ -46,6 +46,8 @@ ANY31(I422ToARGBRow_Any_SSSE3, I422ToARGBRow_SSSE3, 1, 0, 4, 7)
 ANY31(I422ToABGRRow_Any_SSSE3, I422ToABGRRow_SSSE3, 1, 0, 4, 7)
 ANY31(J422ToARGBRow_Any_SSSE3, J422ToARGBRow_SSSE3, 1, 0, 4, 7)
 ANY31(J422ToABGRRow_Any_SSSE3, J422ToABGRRow_SSSE3, 1, 0, 4, 7)
+ANY31(H422ToARGBRow_Any_SSSE3, H422ToARGBRow_SSSE3, 1, 0, 4, 7)
+ANY31(H422ToABGRRow_Any_SSSE3, H422ToABGRRow_SSSE3, 1, 0, 4, 7)
 #endif
 #ifdef HAS_I444TOARGBROW_SSSE3
 ANY31(I444ToARGBRow_Any_SSSE3, I444ToARGBRow_SSSE3, 0, 0, 4, 7)
@@ -72,6 +74,12 @@ ANY31(J422ToARGBRow_Any_AVX2, J422ToARGBRow_AVX2, 1, 0, 4, 15)
 #ifdef HAS_J422TOABGRROW_AVX2
 ANY31(J422ToABGRRow_Any_AVX2, J422ToABGRRow_AVX2, 1, 0, 4, 15)
 #endif
+#ifdef HAS_H422TOARGBROW_AVX2
+ANY31(H422ToARGBRow_Any_AVX2, H422ToARGBRow_AVX2, 1, 0, 4, 15)
+#endif
+#ifdef HAS_H422TOABGRROW_AVX2
+ANY31(H422ToABGRRow_Any_AVX2, H422ToABGRRow_AVX2, 1, 0, 4, 15)
+#endif
 #ifdef HAS_I422TOARGBROW_AVX2
 ANY31(I422ToARGBRow_Any_AVX2, I422ToARGBRow_AVX2, 1, 0, 4, 15)
 #endif

source/row_common.cc

@@ -1010,7 +1010,7 @@ void J400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int width) {
 #define VG 52 /* round(0.813 * 64) */
 #define VR -102 /* round(-1.596 * 64) */
 
-// Bias values to subtract 16 from Y and 128 from U and V.
+// Bias values to subtract 16 from Y and 128 from U and V, with rounding.
 #define BB (UB * 128 + YGB)
 #define BG (UG * 128 + VG * 128 + YGB)
 #define BR (VR * 128 + YGB)
@@ -1058,7 +1058,7 @@ static __inline void YPixel(uint8 y, uint8* b, uint8* g, uint8* r) {
 #define VGJ 46 /* round(0.71414  * 64) */
 #define VRJ -90 /* round(-1.40200 * 64) */
 
-// Bias values to subtract 16 from Y and 128 from U and V.
+// Bias values to round Y and subtract 128 from U and V.
 #define BBJ (UBJ * 128 + YGBJ)
 #define BGJ (UGJ * 128 + VGJ * 128 + YGBJ)
 #define BRJ (VRJ * 128 + YGBJ)
@@ -1082,6 +1082,46 @@ static __inline void YuvJPixel(uint8 y, uint8 u, uint8 v,
 #undef BGJ
 #undef BRJ
 
+// BT.709 YUV to RGB reference
+// *  R = Y                - V * -1.28033
+// *  G = Y - U *  0.21482 - V *  0.38059
+// *  B = Y - U * -2.12798
+
+// Y contribution to R,G,B.  Scale and bias.
+// TODO(fbarchard): Consider moving constants into a common header.
+#define YGH 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
+#define YGBH 32  /* 64 / 2 */
+
+// U and V contributions to R,G,B.
+#define UBH -128 /* max(-128, round(-2.12798 * 64)) */
+#define UGH 14 /* round(0.21482 * 64) */
+#define VGH 24 /* round(0.38059  * 64) */
+#define VRH -82 /* round(-1.28033 * 64) */
+
+// Bias values to round, and subtract 128 from U and V.
+#define BBH (UBH * 128 + YGBH)
+#define BGH (UGH * 128 + VGH * 128 + YGBH)
+#define BRH (VRH * 128 + YGBH)
+
+// C reference code that mimics the YUV assembly.
+static __inline void YuvHPixel(uint8 y, uint8 u, uint8 v,
+                               uint8* b, uint8* g, uint8* r) {
+  uint32 y1 = (uint32)(y * 0x0101 * YGH) >> 16;
+  *b = Clamp((int32)(-(u * UBH) + y1 + BBH) >> 6);
+  *g = Clamp((int32)(-(v * VGH + u * UGH) + y1 + BGH) >> 6);
+  *r = Clamp((int32)(-(v * VRH) + y1 + BRH) >> 6);
+}
+
+#undef YGH
+#undef YGBH
+#undef UBH
+#undef UGH
+#undef VGH
+#undef VRH
+#undef BBH
+#undef BGH
+#undef BRH
+
 #if !defined(LIBYUV_DISABLE_NEON) && \
     (defined(__ARM_NEON__) || defined(__aarch64__) || defined(LIBYUV_NEON))
 // C mimic assembly.
@@ -1204,6 +1244,58 @@ void J422ToABGRRow_C(const uint8* src_y,
   }
 }
 
+// TODO(fbarchard): replace with common matrix function.
+void H422ToARGBRow_C(const uint8* src_y,
+                     const uint8* src_u,
+                     const uint8* src_v,
+                     uint8* rgb_buf,
+                     int width) {
+  int x;
+  for (x = 0; x < width - 1; x += 2) {
+    YuvHPixel(src_y[0], src_u[0], src_v[0],
+              rgb_buf + 0, rgb_buf + 1, rgb_buf + 2);
+    rgb_buf[3] = 255;
+    YuvHPixel(src_y[1], src_u[0], src_v[0],
+              rgb_buf + 4, rgb_buf + 5, rgb_buf + 6);
+    rgb_buf[7] = 255;
+    src_y += 2;
+    src_u += 1;
+    src_v += 1;
+    rgb_buf += 8;  // Advance 2 pixels.
+  }
+  if (width & 1) {
+    YuvHPixel(src_y[0], src_u[0], src_v[0],
+              rgb_buf + 0, rgb_buf + 1, rgb_buf + 2);
+    rgb_buf[3] = 255;
+  }
+}
+
+void H422ToABGRRow_C(const uint8* src_y,
+                     const uint8* src_u,
+                     const uint8* src_v,
+                     uint8* rgb_buf,
+                     int width) {
+  int x;
+  for (x = 0; x < width - 1; x += 2) {
+    YuvHPixel(src_y[0], src_u[0], src_v[0],
+              rgb_buf + 2, rgb_buf + 1, rgb_buf + 0);
+    rgb_buf[3] = 255;
+    YuvHPixel(src_y[1], src_u[0], src_v[0],
+              rgb_buf + 6, rgb_buf + 5, rgb_buf + 4);
+    rgb_buf[7] = 255;
+    src_y += 2;
+    src_u += 1;
+    src_v += 1;
+    rgb_buf += 8;  // Advance 2 pixels.
+  }
+  if (width & 1) {
+    YuvHPixel(src_y[0], src_u[0], src_v[0],
+              rgb_buf + 2, rgb_buf + 1, rgb_buf + 0);
+    rgb_buf[3] = 255;
+  }
+}
+
+
 void I422ToRGB24Row_C(const uint8* src_y,
                       const uint8* src_u,
                       const uint8* src_v,
@@ -2183,6 +2275,7 @@ void I422ToUYVYRow_C(const uint8* src_y,
 
 extern struct YuvConstants kYuvConstants;
 extern struct YuvConstants kYuvJConstants;
+extern struct YuvConstants kYuvHConstants;
 
 #define ANYYUV(NAMEANY, ANY_SIMD, YUVCONSTANTS)                                \
     void NAMEANY(const uint8* y_buf,                                           \
@@ -2196,18 +2289,22 @@ extern struct YuvConstants kYuvJConstants;
 #ifdef HAS_I422TOARGBMATRIXROW_SSSE3
 ANYYUV(I422ToARGBRow_SSSE3, I422ToARGBMatrixRow_SSSE3, kYuvConstants)
 ANYYUV(J422ToARGBRow_SSSE3, I422ToARGBMatrixRow_SSSE3, kYuvJConstants)
+ANYYUV(H422ToARGBRow_SSSE3, I422ToARGBMatrixRow_SSSE3, kYuvHConstants)
 #endif
 #ifdef HAS_I422TOARGBMATRIXROW_AVX2
 ANYYUV(I422ToARGBRow_AVX2, I422ToARGBMatrixRow_AVX2, kYuvConstants)
 ANYYUV(J422ToARGBRow_AVX2, I422ToARGBMatrixRow_AVX2, kYuvJConstants)
+ANYYUV(H422ToARGBRow_AVX2, I422ToARGBMatrixRow_AVX2, kYuvHConstants)
 #endif
 #ifdef HAS_I422TOABGRMATRIXROW_SSSE3
 ANYYUV(I422ToABGRRow_SSSE3, I422ToABGRMatrixRow_SSSE3, kYuvConstants)
 ANYYUV(J422ToABGRRow_SSSE3, I422ToABGRMatrixRow_SSSE3, kYuvJConstants)
+ANYYUV(H422ToABGRRow_SSSE3, I422ToABGRMatrixRow_SSSE3, kYuvHConstants)
 #endif
 #ifdef HAS_I422TOABGRMATRIXROW_AVX2
 ANYYUV(I422ToABGRRow_AVX2, I422ToABGRMatrixRow_AVX2, kYuvConstants)
 ANYYUV(J422ToABGRRow_AVX2, I422ToABGRMatrixRow_AVX2, kYuvJConstants)
+ANYYUV(H422ToABGRRow_AVX2, I422ToABGRMatrixRow_AVX2, kYuvHConstants)
 #endif
 
 // Maximum temporary width for wrappers to process at a time, in pixels.

source/row_gcc.cc

@@ -1429,6 +1429,57 @@ YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
 #undef BGJ
 #undef BRJ
 
+// BT.709 YUV to RGB reference
+// *  R = Y                - V * -1.28033
+// *  G = Y - U *  0.21482 - V *  0.38059
+// *  B = Y - U * -2.12798
+
+// Y contribution to R,G,B.  Scale and bias.
+// TODO(fbarchard): Consider moving constants into a common header.
+#define YGH 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
+#define YGBH 32  /* 64 / 2 */
+
+// U and V contributions to R,G,B.
+#define UBH -128 /* max(-128, round(-2.12798 * 64)) */
+#define UGH 14 /* round(0.21482 * 64) */
+#define VGH 24 /* round(0.38059  * 64) */
+#define VRH -82 /* round(-1.28033 * 64) */
+
+// Bias values to round, and subtract 128 from U and V.
+#define BBH (UBH * 128 + YGBH)
+#define BGH (UGH * 128 + VGH * 128 + YGBH)
+#define BRH (VRH * 128 + YGBH)
+
+// BT.709 constants for YUV to RGB.
+YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
+  { UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0,
+    UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0 },
+  { UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
+    UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
+    UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
+    UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH },
+  { 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH,
+    0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH },
+  { BBH, BBH, BBH, BBH, BBH, BBH, BBH, BBH,
+    BBH, BBH, BBH, BBH, BBH, BBH, BBH, BBH },
+  { BGH, BGH, BGH, BGH, BGH, BGH, BGH, BGH,
+    BGH, BGH, BGH, BGH, BGH, BGH, BGH, BGH },
+  { BRH, BRH, BRH, BRH, BRH, BRH, BRH, BRH,
+    BRH, BRH, BRH, BRH, BRH, BRH, BRH, BRH },
+  { YGH, YGH, YGH, YGH, YGH, YGH, YGH, YGH,
+    YGH, YGH, YGH, YGH, YGH, YGH, YGH, YGH }
+};
+
+#undef YGH
+#undef YGBH
+#undef UBH
+#undef UGH
+#undef VGH
+#undef VRH
+#undef BBH
+#undef BGH
+#undef BRH
+
 // Read 8 UV from 411
 #define READYUV444                                                             \
     "movq       " MEMACCESS([u_buf]) ",%%xmm0                   \n"            \

source/row_win.cc

@@ -143,6 +143,57 @@ YuvConstants SIMD_ALIGNED(kYuvJConstants) = {
 #undef BGJ
 #undef BRJ
 
+// BT.709 YUV to RGB reference
+// *  R = Y                - V * -1.28033
+// *  G = Y - U *  0.21482 - V *  0.38059
+// *  B = Y - U * -2.12798
+
+// Y contribution to R,G,B.  Scale and bias.
+// TODO(fbarchard): Consider moving constants into a common header.
+#define YGH 16320 /* round(1.000 * 64 * 256 * 256 / 257) */
+#define YGBH 32  /* 64 / 2 */
+
+// U and V contributions to R,G,B.
+#define UBH -128 /* max(-128, round(-2.12798 * 64)) */
+#define UGH 14 /* round(0.21482 * 64) */
+#define VGH 24 /* round(0.38059  * 64) */
+#define VRH -82 /* round(-1.28033 * 64) */
+
+// Bias values to round, and subtract 128 from U and V.
+#define BBH (UBH * 128 + YGBH)
+#define BGH (UGH * 128 + VGH * 128 + YGBH)
+#define BRH (VRH * 128 + YGBH)
+
+// BT.709 constants for YUV to RGB.
+YuvConstants SIMD_ALIGNED(kYuvHConstants) = {
+  { UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0,
+    UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0, UBH, 0 },
+  { UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
+    UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
+    UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH,
+    UGH, VGH, UGH, VGH, UGH, VGH, UGH, VGH },
+  { 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH,
+    0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH, 0, VRH },
+  { BBH, BBH, BBH, BBH, BBH, BBH, BBH, BBH,
+    BBH, BBH, BBH, BBH, BBH, BBH, BBH, BBH },
+  { BGH, BGH, BGH, BGH, BGH, BGH, BGH, BGH,
+    BGH, BGH, BGH, BGH, BGH, BGH, BGH, BGH },
+  { BRH, BRH, BRH, BRH, BRH, BRH, BRH, BRH,
+    BRH, BRH, BRH, BRH, BRH, BRH, BRH, BRH },
+  { YGH, YGH, YGH, YGH, YGH, YGH, YGH, YGH,
+    YGH, YGH, YGH, YGH, YGH, YGH, YGH, YGH }
+};
+
+#undef YGH
+#undef YGBH
+#undef UBH
+#undef UGH
+#undef VGH
+#undef VRH
+#undef BBH
+#undef BGH
+#undef BRH
+
 // 64 bit
 #if defined(_M_X64)
 

unit_test/convert_test.cc

@@ -493,6 +493,8 @@ TEST_F(libyuvTest, FMT_PLANAR##To##FMT_B##N) {                                 \
 TESTPLANARTOB(I420, 2, 2, ARGB, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(J420, 2, 2, ARGB, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(J420, 2, 2, ABGR, 4, 4, 1, 2, ARGB, 4)
+TESTPLANARTOB(H420, 2, 2, ARGB, 4, 4, 1, 2, ARGB, 4)
+TESTPLANARTOB(H420, 2, 2, ABGR, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(I420, 2, 2, BGRA, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(I420, 2, 2, ABGR, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(I420, 2, 2, RGBA, 4, 4, 1, 2, ARGB, 4)
@@ -504,6 +506,8 @@ TESTPLANARTOB(I420, 2, 2, ARGB4444, 2, 2, 1, 17, ARGB, 4)
 TESTPLANARTOB(I422, 2, 1, ARGB, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(J422, 2, 1, ARGB, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(J422, 2, 1, ABGR, 4, 4, 1, 2, ARGB, 4)
+TESTPLANARTOB(H422, 2, 1, ARGB, 4, 4, 1, 2, ARGB, 4)
+TESTPLANARTOB(H422, 2, 1, ABGR, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(I422, 2, 1, BGRA, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(I422, 2, 1, ABGR, 4, 4, 1, 2, ARGB, 4)
 TESTPLANARTOB(I422, 2, 1, RGBA, 4, 4, 1, 2, ARGB, 4)
@@ -516,8 +520,6 @@ TESTPLANARTOB(I422, 2, 1, UYVY, 2, 4, 1, 0, ARGB, 4)
 TESTPLANARTOB(I420, 2, 2, I400, 1, 1, 1, 0, ARGB, 4)
 TESTPLANARTOB(J420, 2, 2, J400, 1, 1, 1, 0, ARGB, 4)
 
-
-
 #define TESTQPLANARTOBI(FMT_PLANAR, SUBSAMP_X, SUBSAMP_Y, FMT_B, BPP_B, ALIGN, \
                        YALIGN, W1280, DIFF, N, NEG, OFF)                       \
 TEST_F(libyuvTest, FMT_PLANAR##To##FMT_B##N) {                                 \