diff --git a/include/convert.h b/include/convert.h
index dd51024a9..731f624cd 100644
--- a/include/convert.h
+++ b/include/convert.h
@@ -14,128 +14,96 @@
 
 #include "basic_types.h"
 
-namespace libyuv
-{
-class Convert {
- public:
-  static int
-  I420ToRGB24(const uint8* src_yplane, int src_ystride,
-              const uint8* src_uplane, int src_ustride,
-              const uint8* src_vplane, int src_vstride,
-              uint8* dst_frame, int dst_stride,
-              int src_width, int src_height);
+namespace libyuv {
 
-  static int
-  I420ToARGB(const uint8* src_yplane, int src_ystride,
-             const uint8* src_uplane, int src_ustride,
-             const uint8* src_vplane, int src_vstride,
-             uint8* dst_frame, int dst_stride,
-             int src_width, int src_height);
+int
+I420ToRGB24(const uint8* src_yplane, int src_ystride,
+            const uint8* src_uplane, int src_ustride,
+            const uint8* src_vplane, int src_vstride,
+            uint8* dst_frame, int dst_stride,
+            int src_width, int src_height);
 
-  static int
-  I420ToARGB4444(const uint8* src_yplane, int src_ystride,
-                 const uint8* src_uplane, int src_ustride,
-                 const uint8* src_vplane, int src_vstride,
-                 uint8* dst_frame, int dst_stride,
-                 int src_width, int src_height);
-
-  static int
-  I420ToRGB565(const uint8* src_yplane, int src_ystride,
+int
+I420ToARGB4444(const uint8* src_yplane, int src_ystride,
                const uint8* src_uplane, int src_ustride,
                const uint8* src_vplane, int src_vstride,
                uint8* dst_frame, int dst_stride,
                int src_width, int src_height);
 
-  static int
-  I420ToARGB1555(const uint8* src_yplane, int src_ystride,
-                 const uint8* src_uplane, int src_ustride,
-                 const uint8* src_vplane, int src_vstride,
-                 uint8* dst_frame, int dst_stride,
-                 int src_width, int src_height);
-
-   static int
-  I420ToYUY2(const uint8* src_yplane, int src_ystride,
+int
+I420ToRGB565(const uint8* src_yplane, int src_ystride,
              const uint8* src_uplane, int src_ustride,
              const uint8* src_vplane, int src_vstride,
              uint8* dst_frame, int dst_stride,
              int src_width, int src_height);
 
-  static int
-  I420ToUYVY(const uint8* src_yplane, int src_ystride,
-             const uint8* src_uplane, int src_ustride,
-             const uint8* src_vplane, int src_vstride,
-             uint8* dst_frame, int dst_stride,
-             int src_width, int src_height);
-  static int
-  UYVYToI420(const uint8* src_frame, int src_stride,
-             uint8* dst_yplane, int dst_ystride,
-             uint8* dst_uplane, int dst_ustride,
-             uint8* dst_vplane, int dst_vstride,
-             int src_width, int src_height);
+int
+I420ToARGB1555(const uint8* src_yplane, int src_ystride,
+               const uint8* src_uplane, int src_ustride,
+               const uint8* src_vplane, int src_vstride,
+               uint8* dst_frame, int dst_stride,
+               int src_width, int src_height);
 
-  static int
-  RGB24ToARGB(const uint8* src_frame, int src_stride,
-              uint8* dst_frame, int dst_stride,
-              int src_width, int src_height);
+int
+I420ToYUY2(const uint8* src_yplane, int src_ystride,
+           const uint8* src_uplane, int src_ustride,
+           const uint8* src_vplane, int src_vstride,
+           uint8* dst_frame, int dst_stride,
+           int src_width, int src_height);
 
-  static int
-  RGB24ToI420(const uint8* src_frame, int src_stride,
-              uint8* dst_yplane, int dst_ystride,
-              uint8* dst_uplane, int dst_ustride,
-              uint8* dst_vplane, int dst_vstride,
-              int src_width, int src_height);
+int
+I420ToUYVY(const uint8* src_yplane, int src_ystride,
+           const uint8* src_uplane, int src_ustride,
+           const uint8* src_vplane, int src_vstride,
+           uint8* dst_frame, int dst_stride,
+           int src_width, int src_height);
 
-  static int
-  RAWToI420(const uint8* src_frame, int src_stride,
+int
+RGB24ToARGB(const uint8* src_frame, int src_stride,
+            uint8* dst_frame, int dst_stride,
+            int src_width, int src_height);
+
+int
+RGB24ToI420(const uint8* src_frame, int src_stride,
             uint8* dst_yplane, int dst_ystride,
             uint8* dst_uplane, int dst_ustride,
             uint8* dst_vplane, int dst_vstride,
             int src_width, int src_height);
 
-  static int
-  ABGRToI420(const uint8* src_frame, int src_stride,
-             uint8* dst_yplane, int dst_ystride,
-             uint8* dst_uplane, int dst_ustride,
-             uint8* dst_vplane, int dst_vstride,
-             int src_width, int src_height);
+int
+RAWToI420(const uint8* src_frame, int src_stride,
+          uint8* dst_yplane, int dst_ystride,
+          uint8* dst_uplane, int dst_ustride,
+          uint8* dst_vplane, int dst_vstride,
+          int src_width, int src_height);
 
-  static int
-  I420ToABGR(const uint8* src_yplane, int src_ystride,
-             const uint8* src_uplane, int src_ustride,
-             const uint8* src_vplane, int src_vstride,
+int
+ABGRToI420(const uint8* src_frame, int src_stride,
+           uint8* dst_yplane, int dst_ystride,
+           uint8* dst_uplane, int dst_ustride,
+           uint8* dst_vplane, int dst_vstride,
+           int src_width, int src_height);
+
+int
+BGRAToI420(const uint8* src_frame, int src_stride,
+           uint8* dst_yplane, int dst_ystride,
+           uint8* dst_uplane, int dst_ustride,
+           uint8* dst_vplane, int dst_vstride,
+           int src_width, int src_height);
+
+int
+ARGBToI420(const uint8* src_frame, int src_stride,
+           uint8* dst_yplane, int dst_ystride,
+           uint8* dst_uplane, int dst_ustride,
+           uint8* dst_vplane, int dst_vstride,
+           int src_width, int src_height);
+
+int
+NV12ToRGB565(const uint8* src_yplane, int src_ystride,
+             const  uint8* src_uvplane, int src_uvstride,
              uint8* dst_frame, int dst_stride,
              int src_width, int src_height);
 
-  static int
-  I420ToBGRA(const uint8* src_yplane, int src_ystride,
-             const uint8* src_uplane, int src_ustride,
-             const uint8* src_vplane, int src_vstride,
-             uint8* dst_frame, int dst_stride,
-             int src_width, int src_height);
-
-  static int`
-  BGRAToI420(const uint8* src_frame, int src_stride,
-             uint8* dst_yplane, int dst_ystride,
-             uint8* dst_uplane, int dst_ustride,
-             uint8* dst_vplane, int dst_vstride,
-             int src_width, int src_height);
-
-  int static
-  ARGBToI420(const uint8* src_frame, int src_stride,
-             uint8* dst_yplane, int dst_ystride,
-             uint8* dst_uplane, int dst_ustride,
-             uint8* dst_vplane, int dst_vstride,
-             int src_width, int src_height);
-
-  static int
-  NV12ToRGB565(const uint8* src_yplane, int src_ystride,
-               const  uint8* src_uvplane, int src_uvstride,
-               uint8* dst_frame, int dst_stride,
-               int src_width, int src_height);
-
-  DISALLOW_IMPLICIT_CONSTRUCTORS(Convert);
-
-};
 } //  namespace libyuv
 
 #endif // LIBYUV_INCLUDE_CONVERT_H_
diff --git a/include/format_conversion.h b/include/format_conversion.h
index e93532046..e73d4e50a 100644
--- a/include/format_conversion.h
+++ b/include/format_conversion.h
@@ -8,8 +8,9 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
-#ifndef LIBYUV_INCLUDE_FORMAT_CONVERSION_H_
-#define LIBYUV_INCLUDE_FORMAT_CONVERSION_H_
+
+#ifndef LIBYUV_INCLUDE_FORMATCONVERSION_H_
+#define LIBYUV_INCLUDE_FORMATCONVERSION_H_
 
 #include "basic_types.h"
 
@@ -23,14 +24,18 @@ void BayerRGBToI420(const uint8* src_bayer, int src_pitch_bayer,
                     uint8* dst_v, int dst_pitch_v,
                     int width, int height);
 
-// Converts any 32 bit ARGB to any Bayer RGB format.
-void RGB32ToBayerRGB(const uint8* src_rgb, int src_pitch_rgb,
-                     uint32 src_fourcc_rgb,
-                     uint8* dst_bayer, int dst_pitch_bayer,
-                     uint32 dst_fourcc_bayer,
-                     int width, int height);
+// Converts any Bayer RGB format to ARGB.
+void BayerRGBToARGB(const uint8* src_bayer, int src_pitch_bayer,
+                    uint32 src_fourcc_bayer,
+                    uint8* dst_rgb, int dst_pitch_rgb,
+                    int width, int height);
 
+// Converts ARGB to any Bayer RGB format.
+void ARGBToBayerRGB(const uint8* src_rgb, int src_pitch_rgb,
+                    uint8* dst_bayer, int dst_pitch_bayer,
+                    uint32 dst_fourcc_bayer,
+                    int width, int height);
 
 }  // namespace libyuv
 
-#endif  // LIBYUV_INCLUDE_FORMAT_CONVERSION_H_
+#endif  // LIBYUV_INCLUDE_FORMATCONVERSION_H_
diff --git a/include/general.h b/include/general.h
index 253f27eeb..1a5b48380 100644
--- a/include/general.h
+++ b/include/general.h
@@ -9,50 +9,128 @@
  */
 
 
-/*
- * General operations on YUV images.
- */
-
-#ifndef LIBYUV_INCLUDE_GENERAL_H_
-#define LIBYUV_INCLUDE_GENERAL_H_
+#ifndef LIBYUV_INCLUDE_PLANAR_FUNCTIONS_H_
+#define LIBYUV_INCLUDE_PLANAR_FUNCTIONS_H_
 
 #include "basic_types.h"
 
 namespace libyuv {
 
-// Supported rotation
-enum VideoRotationMode
-{
-  kRotateNone = 0,
-  kRotateClockwise = 90,
-  kRotateCounterClockwise = -90,
-  kRotate180 = 180,
-};
+// Copy I420 to I420.
+void I420Copy(const uint8* src_y, int src_pitch_y,
+              const uint8* src_u, int src_pitch_u,
+              const uint8* src_v, int src_pitch_v,
+              uint8* dst_y, int dst_pitch_y,
+              uint8* dst_u, int dst_pitch_u,
+              uint8* dst_v, int dst_pitch_v,
+              int width, int height);
 
-// I420  mirror
-int
-I420Mirror(const uint8* src_yplane, int src_ystride,
-           const uint8* src_uplane, int src_ustride,
-           const uint8* src_vplane, int src_vstride,
-           uint8* dst_yplane, int dst_ystride,
-           uint8* dst_uplane, int dst_ustride,
-           uint8* dst_vplane, int dst_vstride,
-           int width, int height);
+// Convert I422 to I420.  Used by MJPG.
+void I422ToI420(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-// Crop/Pad I420 frame to match required dimensions.
-int
-I420CropPad(const uint8* src_frame, int src_width,
-           int src_height, uint8* dst_frame,
-           int dst_width, int dst_height);
+// Convert M420 to I420.
+void M420ToI420(const uint8* src_m420, int src_pitch_m420,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-// I420 Crop - make a center cut
-int
-I420Cut(uint8* frame,
-        int src_width, int src_height,
-        int dst_width, int dst_height);
+// Convert Q420 to I420.
+void Q420ToI420(const uint8* src_y, int src_pitch_y,
+                const uint8* src_yuy2, int src_pitch_yuy2,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
+// Convert NV12 to I420.  Also used for NV21.
+void NV12ToI420(const uint8* src_y,
+                const uint8* src_uv, int src_pitch,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-} // namespace libyuv
+// Convert YUY2 to I420.
+void YUY2ToI420(const uint8* src_yuy2, int src_pitch_yuy2,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
+// Convert UYVY to I420.
+void UYVYToI420(const uint8* src_uyvy, int src_pitch_uyvy,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-#endif // LIBYUV_INCLUDE_GENERAL_H_
+// Convert I420 to ARGB.
+void I420ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert I420 to BGRA.
+void I420ToBGRA(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert I420 to ABGR.
+void I420ToABGR(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert I422 to ARGB.
+void I422ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert I444 to ARGB.
+void I444ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert I400 to ARGB.
+void I400ToARGB(const uint8* src_y, int src_pitch_y,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert I400 to ARGB.
+void I400ToARGB_Reference(const uint8* src_y, int src_pitch_y,
+                          uint8* dst_argb, int dst_pitch_argb,
+                          int width, int height);
+
+// Convert RAW to ARGB.
+void RAWToARGB(const uint8* src_raw, int src_pitch_raw,
+               uint8* dst_argb, int dst_pitch_argb,
+               int width, int height);
+
+// Convert BG24 to ARGB.
+void BG24ToARGB(const uint8* src_bg24, int src_pitch_bg24,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+// Convert ABGR to ARGB.
+void ABGRToARGB(const uint8* src_abgr, int src_pitch_abgr,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
+
+}  // namespace libyuv
+
+#endif  // LIBYUV_INCLUDE_PLANAR_FUNCTIONS_H_
diff --git a/include/planar_functions.h b/include/planar_functions.h
index a078b6762..1a5b48380 100644
--- a/include/planar_functions.h
+++ b/include/planar_functions.h
@@ -16,112 +16,120 @@
 
 namespace libyuv {
 
-class PlanarFunctions {
- public:
+// Copy I420 to I420.
+void I420Copy(const uint8* src_y, int src_pitch_y,
+              const uint8* src_u, int src_pitch_u,
+              const uint8* src_v, int src_pitch_v,
+              uint8* dst_y, int dst_pitch_y,
+              uint8* dst_u, int dst_pitch_u,
+              uint8* dst_v, int dst_pitch_v,
+              int width, int height);
 
-    // Copy I420 to I420.
-    static void I420Copy(const uint8* src_y, int src_pitch_y,
-                         const uint8* src_u, int src_pitch_u,
-                         const uint8* src_v, int src_pitch_v,
-                         uint8* dst_y, int dst_pitch_y,
-                         uint8* dst_u, int dst_pitch_u,
-                         uint8* dst_v, int dst_pitch_v,
-                         int width, int height);
+// Convert I422 to I420.  Used by MJPG.
+void I422ToI420(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-    // Convert I422 to I420.  Used by MJPG.
-    static void I422ToI420(const uint8* src_y, int src_pitch_y,
-                           const uint8* src_u, int src_pitch_u,
-                           const uint8* src_v, int src_pitch_v,
-                           uint8* dst_y, int dst_pitch_y,
-                           uint8* dst_u, int dst_pitch_u,
-                           uint8* dst_v, int dst_pitch_v,
-                           int width, int height);
+// Convert M420 to I420.
+void M420ToI420(const uint8* src_m420, int src_pitch_m420,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-    // Convert M420 to I420.
-    static void M420ToI420(const uint8* src_m420, int src_pitch_m420,
-                           uint8* dst_y, int dst_pitch_y,
-                           uint8* dst_u, int dst_pitch_u,
-                           uint8* dst_v, int dst_pitch_v,
-                           int width, int height);
+// Convert Q420 to I420.
+void Q420ToI420(const uint8* src_y, int src_pitch_y,
+                const uint8* src_yuy2, int src_pitch_yuy2,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-    // Convert Q420 to I420.
-    static void Q420ToI420(const uint8* src_y, int src_pitch_y,
-                           const uint8* src_yuy2, int src_pitch_yuy2,
-                           uint8* dst_y, int dst_pitch_y,
-                           uint8* dst_u, int dst_pitch_u,
-                           uint8* dst_v, int dst_pitch_v,
-                           int width, int height);
+// Convert NV12 to I420.  Also used for NV21.
+void NV12ToI420(const uint8* src_y,
+                const uint8* src_uv, int src_pitch,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-    // Convert NV12 to I420.  Also used for NV21.
-    static void NV12ToI420(const uint8* src_y,
-                           const uint8* src_uv, int src_pitch,
-                           uint8* dst_y, int dst_pitch_y,
-                           uint8* dst_u, int dst_pitch_u,
-                           uint8* dst_v, int dst_pitch_v,
-                           int width, int height);
+// Convert YUY2 to I420.
+void YUY2ToI420(const uint8* src_yuy2, int src_pitch_yuy2,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-    // Convert YUY2 to I420.
-    static void YUY2ToI420(const uint8* src_yuy2, int src_pitch_yuy2,
-                           uint8* dst_y, int dst_pitch_y,
-                           uint8* dst_u, int dst_pitch_u,
-                           uint8* dst_v, int dst_pitch_v,
-                           int width, int height);
+// Convert UYVY to I420.
+void UYVYToI420(const uint8* src_uyvy, int src_pitch_uyvy,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height);
 
-    // Convert UYVY to I420.
-    static void UYVYToI420(const uint8* src_uyvy, int src_pitch_uyvy,
-                           uint8* dst_y, int dst_pitch_y,
-                           uint8* dst_u, int dst_pitch_u,
-                           uint8* dst_v, int dst_pitch_v,
-                           int width, int height);
+// Convert I420 to ARGB.
+void I420ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    // Convert I420 to ARGB.
-    static void I420ToARGB(const uint8* src_y, int src_pitch_y,
-                           const uint8* src_u, int src_pitch_u,
-                           const uint8* src_v, int src_pitch_v,
-                           uint8* dst_argb, int dst_pitch_argb,
-                           int width, int height);
+// Convert I420 to BGRA.
+void I420ToBGRA(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    // Convert I422 to ARGB.
-    static void I422ToARGB(const uint8* src_y, int src_pitch_y,
-                           const uint8* src_u, int src_pitch_u,
-                           const uint8* src_v, int src_pitch_v,
-                           uint8* dst_argb, int dst_pitch_argb,
-                           int width, int height);
+// Convert I420 to ABGR.
+void I420ToABGR(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    // Convert I444 to ARGB.
-    static void I444ToARGB(const uint8* src_y, int src_pitch_y,
-                           const uint8* src_u, int src_pitch_u,
-                           const uint8* src_v, int src_pitch_v,
-                           uint8* dst_argb, int dst_pitch_argb,
-                           int width, int height);
+// Convert I422 to ARGB.
+void I422ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    // Convert I400 to ARGB.
-    static void I400ToARGB(const uint8* src_y, int src_pitch_y,
-                           uint8* dst_argb, int dst_pitch_argb,
-                           int width, int height);
+// Convert I444 to ARGB.
+void I444ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    // Convert I400 to ARGB.
-    static void I400ToARGB_Reference(const uint8* src_y, int src_pitch_y,
-                                     uint8* dst_argb, int dst_pitch_argb,
-                                     int width, int height);
+// Convert I400 to ARGB.
+void I400ToARGB(const uint8* src_y, int src_pitch_y,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    // Convert RAW to ARGB.
-    static void RAWToARGB(const uint8* src_raw, int src_pitch_raw,
+// Convert I400 to ARGB.
+void I400ToARGB_Reference(const uint8* src_y, int src_pitch_y,
                           uint8* dst_argb, int dst_pitch_argb,
                           int width, int height);
 
-    // Convert BG24 to ARGB.
-    static void BG24ToARGB(const uint8* src_bg24, int src_pitch_bg24,
-                           uint8* dst_argb, int dst_pitch_argb,
-                           int width, int height);
+// Convert RAW to ARGB.
+void RAWToARGB(const uint8* src_raw, int src_pitch_raw,
+               uint8* dst_argb, int dst_pitch_argb,
+               int width, int height);
 
-    // Convert ABGR to ARGB.
-    static void ABGRToARGB(const uint8* src_abgr, int src_pitch_abgr,
-                           uint8* dst_argb, int dst_pitch_argb,
-                           int width, int height);
+// Convert BG24 to ARGB.
+void BG24ToARGB(const uint8* src_bg24, int src_pitch_bg24,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
-    DISALLOW_IMPLICIT_CONSTRUCTORS(PlanarFunctions);
-  };
+// Convert ABGR to ARGB.
+void ABGRToARGB(const uint8* src_abgr, int src_pitch_abgr,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height);
 
 }  // namespace libyuv
 
diff --git a/include/scale.h b/include/scale.h
index 57b2c18b9..9cef9bce8 100644
--- a/include/scale.h
+++ b/include/scale.h
@@ -12,7 +12,6 @@
 #ifndef LIBYUV_INCLUDE_SCALE_H_
 #define LIBYUV_INCLUDE_SCALE_H_
 
-
 #include "basic_types.h"
 
 #if defined(_MSC_VER)
@@ -21,43 +20,32 @@
 #define ALIGN16(var) var __attribute__((aligned(16)))
 #endif
 
+namespace libyuv {
 
-namespace libyuv
-{
+// Scales a YUV 4:2:0 image from the input width and height to the
+// output width and height. If outh_offset is nonzero, the image is
+// offset by that many pixels and stretched to (outh - outh_offset * 2)
+// pixels high, instead of outh.
+// If interpolate is not set, a simple nearest-neighbor algorithm is
+// used. This produces basic (blocky) quality at the fastest speed.
+// If interpolate is set, interpolation is used to produce a better
+// quality image, at the expense of speed.
+// Returns true if successful.
+bool Scale(const uint8 *in, int32 inw, int32 inh,
+           uint8 *out, int32 outw, int32 outh, int32 outh_offset,
+           bool interpolate);
 
-class YuvScaler {
- public:
-  // Scales a YUV 4:2:0 image from the input width and height to the
-  // output width and height. If outh_offset is nonzero, the image is
-  // offset by that many pixels and stretched to (outh - outh_offset * 2)
-  // pixels high, instead of outh.
-  // If interpolate is not set, a simple nearest-neighbor algorithm is
-  // used. This produces basic (blocky) quality at the fastest speed.
-  // If interpolate is set, interpolation is used to produce a better
-  // quality image, at the expense of speed.
-  // Returns true if successful.
-  static bool Scale(const uint8 *in, int32 inw, int32 inh,
-                    uint8 *out, int32 outw, int32 outh, int32 outh_offset,
-                    bool interpolate);
+// Same, but specified in terms of each plane location and stride.
+bool Scale(const uint8 *inY, const uint8 *inU, const uint8 *inV,
+           int32 istrideY, int32 istrideU, int32 istrideV,
+           int32 iwidth, int32 iheight,
+           uint8 *outY, uint8 *outU, uint8 *outV,
+           int32 ostrideY, int32 ostrideU, int32 ostrideV,
+           int32 owidth, int32 oheight,
+           bool interpolate);
 
-  // Same, but specified in terms of each plane location and stride.
-  static bool Scale(const uint8 *inY, const uint8 *inU, const uint8 *inV,
-                    int32 istrideY, int32 istrideU, int32 istrideV,
-                    int32 iwidth, int32 iheight,
-                    uint8 *outY, uint8 *outU, uint8 *outV,
-                    int32 ostrideY, int32 ostrideU, int32 ostrideV,
-                    int32 owidth, int32 oheight,
-                    bool interpolate);
-
-  // For testing, allow disabling of optimizations.
-  static void SetUseReferenceImpl(bool use) { use_reference_impl_ = use; }
-
- private:
-
-  static bool use_reference_impl_;
-
-  DISALLOW_IMPLICIT_CONSTRUCTORS(YuvScaler);
-};
+// For testing, allow disabling of optimizations.
+void SetUseReferenceImpl(bool use);
 
 } // namespace libyuv
 
diff --git a/source/convert.cc b/source/convert.cc
index 70821bc62..4e4394a09 100644
--- a/source/convert.cc
+++ b/source/convert.cc
@@ -37,11 +37,11 @@ void *memcpy_8(void * dest, const void * src, size_t n);
 
 
 int
-Convert::I420ToRGB24(const uint8* src_yplane, int src_ystride,
-                     const uint8* src_uplane, int src_ustride,
-                     const uint8* src_vplane, int src_vstride,
-                     uint8* dst_frame, int dst_stride,
-                     int src_width, int src_height)
+I420ToRGB24(const uint8* src_yplane, int src_ystride,
+            const uint8* src_uplane, int src_ustride,
+            const uint8* src_vplane, int src_vstride,
+            uint8* dst_frame, int dst_stride,
+            int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
       dst_frame == NULL)
@@ -106,168 +106,13 @@ Convert::I420ToRGB24(const uint8* src_yplane, int src_ystride,
   return 0;
 }
 
-
-int
-Convert::I420ToARGB(const uint8* src_yplane, int src_ystride,
-                    const uint8* src_uplane, int src_ustride,
-                    const uint8* src_vplane, int src_vstride,
-                    uint8* dst_frame, int dst_stride,
-                    int src_width, int src_height)
-{
-  if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
-      dst_frame == NULL){
-      return -1;
-    }
-
-  uint8* out1 = dst_frame;
-  uint8* out2 = out1 + dst_stride * 4;
-  const uint8 *y1,*y2, *u, *v;
-  y1 = src_yplane;
-  y2 = src_yplane + src_ystride;
-  u = src_uplane;
-  v = src_vplane;
-  int h, w;
-  int tmpR, tmpG, tmpB;
-
-  for (h = ((src_height + 1) >> 1); h > 0; h--){
-    // Do 2 rows at the time
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Vertical and horizontal sub-sampling
-
-      tmpR = (int32)((mapYc[y1[0]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y1[0]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y1[0]] + mapUcb[u[0]] + 128) >> 8);
-      out1[0] = Clip(tmpB);
-      out1[1] = Clip(tmpG);
-      out1[2] = Clip(tmpR);
-      out1[3] = 0xff;
-
-      tmpR = (int32)((mapYc[y1[1]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y1[1]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y1[1]] + mapUcb[u[0]] + 128) >> 8);
-      out1[4] = Clip(tmpB);
-      out1[5] = Clip(tmpG);
-      out1[6] = Clip(tmpR);
-      out1[7] = 0xff;
-
-      tmpR = (int32)((mapYc[y2[0]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y2[0]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y2[0]] + mapUcb[u[0]] + 128) >> 8);
-      out2[0] = Clip(tmpB);
-      out2[1] = Clip(tmpG);
-      out2[2] = Clip(tmpR);
-      out2[3] = 0xff;
-
-      tmpR = (int32)((mapYc[y2[1]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y2[1]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y2[1]] + mapUcb[u[0]] + 128) >> 8);
-      out2[4] = Clip(tmpB);
-      out2[5] = Clip(tmpG);
-      out2[6] = Clip(tmpR);
-      out2[7] = 0xff;
-
-      out1 += 8;
-      out2 += 8;
-      y1 += 2;
-      y2 += 2;
-      u++;
-      v++;
-    }
-  y1 += 2 * src_ystride - src_width;
-  y2 += 2 * src_ystride - src_width;
-  u += src_ustride - ((src_width + 1) >> 1);
-  v += src_vstride - ((src_width + 1) >> 1);
-  out1 += (2 * dst_stride - src_width) * 4;
-  out2 += (2 * dst_stride - src_width) * 4;
-  } // end height for
-  return 0;
-}
-
-
-int
-Convert::I420ToBGRA(const uint8* src_yplane, int src_ystride,
-                    const uint8* src_uplane, int src_ustride,
-                    const uint8* src_vplane, int src_vstride,
-                    uint8* dst_frame, int dst_stride,
-                    int src_width, int src_height)
-{
-  if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
-      dst_frame == NULL){
-      return -1;
-    }
-
-  uint8* out1 = dst_frame;
-  uint8* out2 = out1 + dst_stride * 4;
-  const uint8 *y1,*y2, *u, *v;
-  y1 = src_yplane;
-  y2 = src_yplane + src_ystride;
-  u = src_uplane;
-  v = src_vplane;
-  int h, w;
-  int tmpR, tmpG, tmpB;
-
-  for (h = ((src_height + 1) >> 1); h > 0; h--){
-    // Do 2 rows at the time
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Vertical and horizontal sub-sampling
-
-      tmpR = (int32)((mapYc[y1[0]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y1[0]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y1[0]] + mapUcb[u[0]] + 128) >> 8);
-      out1[0] = 0xff;
-      out1[1] = Clip(tmpR);
-      out1[2] = Clip(tmpG);
-      out1[3] = Clip(tmpB);
-
-      tmpR = (int32)((mapYc[y1[1]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y1[1]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y1[1]] + mapUcb[u[0]] + 128) >> 8);
-      out1[4] = 0xff;
-      out1[5] = Clip(tmpR);
-      out1[6] = Clip(tmpG);
-      out1[7] = Clip(tmpB);
-
-      tmpR = (int32)((mapYc[y2[0]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y2[0]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y2[0]] + mapUcb[u[0]] + 128) >> 8);
-      out2[0] = 0xff;
-      out2[1] = Clip(tmpR);
-      out2[2] = Clip(tmpG);
-      out2[3] = Clip(tmpB);
-
-      tmpR = (int32)((mapYc[y2[1]] + mapVcr[v[0]] + 128) >> 8);
-      tmpG = (int32)((mapYc[y2[1]] + mapUcg[u[0]] + mapVcg[v[0]] + 128) >> 8);
-      tmpB = (int32)((mapYc[y2[1]] + mapUcb[u[0]] + 128) >> 8);
-      out2[4] = 0xff;
-      out2[5] = Clip(tmpR);
-      out2[6] = Clip(tmpG);
-      out2[7] = Clip(tmpB);
-
-      out1 += 8;
-      out2 += 8;
-      y1 += 2;
-      y2 += 2;
-      u++;
-      v++;
-    }
-  y1 += 2 * src_ystride - src_width;
-  y2 += 2 * src_ystride - src_width;
-  u += src_ustride - ((src_width + 1) >> 1);
-  v += src_vstride - ((src_width + 1) >> 1);
-  out1 += (2 * dst_stride - src_width) * 4;
-  out2 += (2 * dst_stride - src_width) * 4;
-  } // end height for
-  return 0;
-}
-
-
 // Little Endian...
 int
-Convert::I420ToARGB4444(const uint8* src_yplane, int src_ystride,
-                        const uint8* src_uplane, int src_ustride,
-                        const uint8* src_vplane, int src_vstride,
-                        uint8* dst_frame, int dst_stride,
-                        int src_width, int src_height)
+I420ToARGB4444(const uint8* src_yplane, int src_ystride,
+               const uint8* src_uplane, int src_ustride,
+               const uint8* src_vplane, int src_vstride,
+               uint8* dst_frame, int dst_stride,
+               int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
       dst_frame == NULL){
@@ -332,11 +177,11 @@ Convert::I420ToARGB4444(const uint8* src_yplane, int src_ystride,
 
 
 int
-Convert::I420ToRGB565(const uint8* src_yplane, int src_ystride,
-                      const uint8* src_uplane, int src_ustride,
-                      const uint8* src_vplane, int src_vstride,
-                      uint8* dst_frame, int dst_stride,
-                      int src_width, int src_height)
+I420ToRGB565(const uint8* src_yplane, int src_ystride,
+             const uint8* src_uplane, int src_ustride,
+             const uint8* src_vplane, int src_vstride,
+             uint8* dst_frame, int dst_stride,
+             int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
       dst_frame == NULL){
@@ -414,11 +259,11 @@ Convert::I420ToRGB565(const uint8* src_yplane, int src_ystride,
 
 
 int
-Convert::I420ToARGB1555(const uint8* src_yplane, int src_ystride,
-                        const uint8* src_uplane, int src_ustride,
-                        const uint8* src_vplane, int src_vstride,
-                        uint8* dst_frame, int dst_stride,
-                        int src_width, int src_height)
+I420ToARGB1555(const uint8* src_yplane, int src_ystride,
+               const uint8* src_uplane, int src_ustride,
+               const uint8* src_vplane, int src_vstride,
+               uint8* dst_frame, int dst_stride,
+               int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
       dst_frame == NULL){
@@ -485,11 +330,11 @@ Convert::I420ToARGB1555(const uint8* src_yplane, int src_ystride,
 
 
 int
-Convert::I420ToYUY2(const uint8* src_yplane, int src_ystride,
-                    const uint8* src_uplane, int src_ustride,
-                    const uint8* src_vplane, int src_vstride,
-                    uint8* dst_frame, int dst_stride,
-                    int src_width, int src_height)
+I420ToYUY2(const uint8* src_yplane, int src_ystride,
+           const uint8* src_uplane, int src_ustride,
+           const uint8* src_vplane, int src_vstride,
+           uint8* dst_frame, int dst_stride,
+           int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
       dst_frame == NULL){
@@ -601,11 +446,11 @@ Convert::I420ToYUY2(const uint8* src_yplane, int src_ystride,
 }
 
 int
-Convert::I420ToUYVY(const uint8* src_yplane, int src_ystride,
-                    const uint8* src_uplane, int src_ustride,
-                    const uint8* src_vplane, int src_vstride,
-                    uint8* dst_frame, int dst_stride,
-                    int src_width, int src_height)
+I420ToUYVY(const uint8* src_yplane, int src_ystride,
+           const uint8* src_uplane, int src_ustride,
+           const uint8* src_vplane, int src_vstride,
+           uint8* dst_frame, int dst_stride,
+           int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
       dst_frame == NULL){
@@ -717,10 +562,10 @@ loop0:
 
 
 int
-Convert::NV12ToRGB565(const uint8* src_yplane, int src_ystride,
-                      const  uint8* src_uvplane, int src_uvstride,
-                      uint8* dst_frame, int dst_stride,
-                      int src_width, int src_height)
+NV12ToRGB565(const uint8* src_yplane, int src_ystride,
+             const  uint8* src_uvplane, int src_uvstride,
+             uint8* dst_frame, int dst_stride,
+             int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uvplane == NULL || dst_frame == NULL){
      return -1;
@@ -788,11 +633,11 @@ Convert::NV12ToRGB565(const uint8* src_yplane, int src_ystride,
 }
 
 int
-Convert::I420ToABGR(const uint8* src_yplane, int src_ystride,
-                    const uint8* src_uplane, int src_ustride,
-                    const uint8* src_vplane, int src_vstride,
-                    uint8* dst_frame, int dst_stride,
-                    int src_width, int src_height)
+I420ToABGR(const uint8* src_yplane, int src_ystride,
+           const uint8* src_uplane, int src_ustride,
+           const uint8* src_vplane, int src_vstride,
+           uint8* dst_frame, int dst_stride,
+           int src_width, int src_height)
 {
   if (src_yplane == NULL || src_uplane == NULL || src_vplane == NULL ||
        dst_frame == NULL){
@@ -872,56 +717,10 @@ Convert::I420ToABGR(const uint8* src_yplane, int src_ystride,
   return 0;
 }
 
-
 int
-Convert::UYVYToI420(const uint8* src_frame, int src_stride,
-                    uint8* dst_yplane, int dst_ystride,
-                    uint8* dst_uplane, int dst_ustride,
-                    uint8* dst_vplane, int dst_vstride,
-                    int src_width, int src_height)
-{
-  if (dst_yplane == NULL || dst_uplane == NULL || dst_vplane == NULL ||
-      src_frame == NULL){
-     return -1;
-  }
-
-  int i, j;
-  uint8* outI = dst_yplane;
-  uint8* outU = dst_uplane;
-  uint8* outV = dst_vplane;
-
-  // U0Y0V0Y1..U2Y2V2Y3.....
-
-  for (i = 0; i < ((src_height + 1) >> 1); i++){
-    for (j = 0; j < ((src_width + 1) >> 1); j++){
-      outI[0] = src_frame[1];
-      *outU = src_frame[0];
-      outI[1] = src_frame[3];
-      *outV = src_frame[2];
-      src_frame += 4;
-      outI += 2;
-      outU++;
-      outV++;
-    }
-    for (j = 0; j < ((src_width + 1) >> 1); j++)
-    {
-      outI[0] = src_frame[1];
-      outI[1] = src_frame[3];
-      src_frame += 4;
-      outI += 2;
-    }
-    outI += dst_ystride - src_width;
-    outU += dst_ustride - ((src_width + 1) << 1);
-    outV += dst_vstride - ((src_width + 1) << 1);
-  }
-  return 0;
-}
-
-
-int
-Convert::RGB24ToARGB(const uint8* src_frame, int src_stride,
-                     uint8* dst_frame, int dst_stride,
-                     int src_width, int src_height)
+RGB24ToARGB(const uint8* src_frame, int src_stride,
+            uint8* dst_frame, int dst_stride,
+            int src_width, int src_height)
 {
   if (src_frame == NULL || dst_frame == NULL){
     return -1;
@@ -936,7 +735,7 @@ Convert::RGB24ToARGB(const uint8* src_frame, int src_stride,
   {
     for (j = 0; j < src_width; j++)
     {
-      offset = j*4;
+      offset = j * 4;
       outFrame[0 + offset] = inFrame[0];
       outFrame[1 + offset] = inFrame[1];
       outFrame[2 + offset] = inFrame[2];
@@ -949,326 +748,190 @@ Convert::RGB24ToARGB(const uint8* src_frame, int src_stride,
   return 0;
 }
 
+// ARGBToI420Row_C etc row functions use the following macro, generating
+// code with RGB offsets/strides different for each version.  Less error
+// prone than duplicating the code.
+// template could be used, but macro method works for C and asm and this is
+// performance critical code.
 
-int
-Convert::RGB24ToI420(const uint8* src_frame, int src_stride,
+#define MAKEROWRGBTOI420(NAME,R,G,B,BPP) \
+static void \
+NAME(const uint8* src_row0, const uint8* src_row1, \
+         uint8* dst_yplane0, uint8* dst_yplane1, \
+         uint8* dst_uplane, \
+         uint8* dst_vplane, \
+         int src_width) { \
+  for (int x = 0; x < src_width - 1; x += 2) { \
+    dst_yplane0[0] = (uint8)((src_row0[R] * 66 + \
+                              src_row0[G] * 129 + \
+                              src_row0[B] * 25 + 128) >> 8) + 16; \
+    dst_yplane0[1] = (uint8)((src_row0[R + BPP] * 66 + \
+                              src_row0[G + BPP] * 129 + \
+                              src_row0[B + BPP] * 25 + 128) >> 8) + 16; \
+    dst_yplane1[0] = (uint8)((src_row1[R] * 66 + \
+                              src_row1[G] * 129 + \
+                              src_row1[B] * 25 + 128) >> 8) + 16; \
+    dst_yplane1[1] = (uint8)((src_row1[R + BPP] * 66 + \
+                              src_row1[G + BPP] * 129 + \
+                              src_row1[B + BPP] * 25 + 128) >> 8) + 16; \
+    dst_uplane[0] = (uint8)(((src_row0[R] + src_row0[R + BPP] + \
+                              src_row1[R] + src_row1[R + BPP]) * -38 + \
+                             (src_row0[G] + src_row0[G + BPP] + \
+                              src_row1[G] + src_row1[G + BPP]) * -74 + \
+                             (src_row0[B] + src_row0[B + BPP] + \
+                              src_row1[B] + src_row1[B + BPP]) * 112 + \
+                              + 512) >> 10) + 128; \
+    dst_vplane[0] = (uint8)(((src_row0[R] + src_row0[R + BPP] + \
+                              src_row1[R] + src_row1[R + BPP]) * 112 + \
+                             (src_row0[G] + src_row0[G + BPP] + \
+                              src_row1[G] + src_row1[G + BPP]) * -94 + \
+                             (src_row0[B] + src_row0[B + BPP] + \
+                              src_row1[B] + src_row1[B + BPP]) * -18 + \
+                              + 512) >> 10) + 128; \
+    dst_yplane0 += 2; \
+    dst_yplane1 += 2; \
+    ++dst_uplane; \
+    ++dst_vplane; \
+    src_row0 += BPP * 2; \
+    src_row1 += BPP * 2; \
+  } \
+  if (src_width & 1) { \
+    dst_yplane0[0] = (uint8)((src_row0[R] * 66 + \
+                              src_row0[G] * 129 + \
+                              src_row0[B] * 25 + 128) >> 8) + 16; \
+    dst_yplane1[0] = (uint8)((src_row1[R] * 66 + \
+                              src_row1[G] * 129 + \
+                              src_row1[B] * 25 + 128) >> 8) + 16; \
+    dst_uplane[0] = (uint8)(((src_row0[R] + \
+                              src_row1[R]) * -38 + \
+                             (src_row0[G] + \
+                              src_row1[G]) * -74 + \
+                             (src_row0[B] + \
+                              src_row1[B]) * 112 + \
+                              + 256) >> 9) + 128; \
+    dst_vplane[0] = (uint8)(((src_row0[R] + \
+                              src_row1[R]) * 112 + \
+                             (src_row0[G] + \
+                              src_row1[G]) * -94 + \
+                             (src_row0[B] + \
+                              src_row1[B]) * -18 + \
+                              + 256) >> 9) + 128; \
+  } \
+}
+
+// Generate variations of RGBToI420.  Parameters are r,g,b offsets within a
+// pixel, and number of bytes per pixel.
+MAKEROWRGBTOI420(ARGBToI420Row_C, 2, 1, 0, 4)
+MAKEROWRGBTOI420(BGRAToI420Row_C, 1, 2, 3, 4)
+MAKEROWRGBTOI420(ABGRToI420Row_C, 0, 1, 2, 4)
+MAKEROWRGBTOI420(RGB24ToI420Row_C, 2, 1, 0, 3)
+MAKEROWRGBTOI420(RAWToI420Row_C, 0, 1, 2, 3)
+
+static int RGBToI420(const uint8* src_frame, int src_stride,
                      uint8* dst_yplane, int dst_ystride,
                      uint8* dst_uplane, int dst_ustride,
                      uint8* dst_vplane, int dst_vstride,
-                     int src_width, int src_height)
-{
+                     int src_width, int src_height,
+                     void (*RGBToI420Row)(const uint8* src_row0,
+                                          const uint8* src_row1,
+                                          uint8* dst_yplane0,
+                                          uint8* dst_yplane1,
+                                          uint8* dst_uplane,
+                                          uint8* dst_vplane,
+                                          int src_width)) {
   if (src_frame == NULL || dst_yplane == NULL ||
-      dst_vplane == NULL || dst_vplane == NULL)
-    return -1;
-
-  uint8* yStartPtr;
-  uint8* yStartPtr2;
-  uint8* uStartPtr;
-  uint8* vStartPtr;
-  const uint8* inpPtr;
-  const uint8* inpPtr2;
-  int h, w;
-
-  yStartPtr = dst_yplane;
-  yStartPtr2 = yStartPtr + dst_ystride;
-  uStartPtr = dst_uplane;
-  vStartPtr = dst_vplane;
-  inpPtr = src_frame + src_stride * src_height * 3 - 3 * src_height;
-  inpPtr2 = inpPtr - 3 * src_stride;
-
-  for (h = 0; h < ((src_height + 1) >> 1); h++ ){
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Y
-      yStartPtr[0] =  (uint8)((66 * inpPtr[2] + 129 * inpPtr[1]
-                                      + 25 * inpPtr[0] + 128) >> 8) + 16;
-      yStartPtr2[0] = (uint8)((66 * inpPtr2[2] + 129 * inpPtr2[1]
-                                      + 25 * inpPtr2[0] + 128) >> 8) + 16;
-      // Moving to next column
-      yStartPtr[1] = (uint8)((66 * inpPtr[5] + 129 * inpPtr[4]
-                                     + 25 * inpPtr[3]  + 128) >> 8) + 16;
-      yStartPtr2[1] = (uint8)((66 * inpPtr2[5] + 129 * inpPtr2[4]
-                                      + 25 * inpPtr2[3] + 128) >> 8 ) + 16;
-      // U
-      uStartPtr[0] = (uint8)((-38 * inpPtr[2] - 74 * inpPtr[1] +
-                                      112 * inpPtr[0] + 128) >> 8) + 128;
-      // V
-      vStartPtr[0] = (uint8)((112 * inpPtr[2] -94 * inpPtr[1] -
-                                     18 * inpPtr[0] + 128) >> 8) + 128;
-
-      yStartPtr += 2;
-      yStartPtr2 += 2;
-      uStartPtr++;
-      vStartPtr++;
-      inpPtr += 6;
-      inpPtr2 += 6;
-    } // end for w
-    yStartPtr += dst_ystride + dst_ystride - src_width;
-    yStartPtr2 += dst_ystride + dst_ystride - src_width;
-    uStartPtr += dst_ustride + dst_ustride - ((src_width + 1) >> 1);
-    vStartPtr += dst_vstride + dst_vstride - ((src_width + 1) >> 1);
-    inpPtr -= 3 * (2 * src_stride + src_width);
-    inpPtr2 -= 3 * (2 * src_stride + src_width);
-  } // end for h
-  return 0;
-}
-
-int
-Convert::RAWToI420(const uint8* src_frame, int src_stride,
-                   uint8* dst_yplane, int dst_ystride,
-                   uint8* dst_uplane, int dst_ustride,
-                   uint8* dst_vplane, int dst_vstride,
-                   int src_width, int src_height)
-{
-  if (src_frame == NULL || dst_yplane == NULL ||
-      dst_vplane == NULL || dst_vplane == NULL)
-    return -1;
-
-  uint8* yStartPtr;
-  uint8* yStartPtr2;
-  uint8* uStartPtr;
-  uint8* vStartPtr;
-  const uint8* inpPtr;
-  const uint8* inpPtr2;
-  int h, w;
-
-  yStartPtr = dst_yplane;
-  yStartPtr2 = yStartPtr + dst_ystride;
-  uStartPtr = dst_uplane;
-  vStartPtr = dst_vplane;
-  inpPtr = src_frame + src_stride * src_height * 3 - 3 * src_height;
-  inpPtr2 = inpPtr - 3 * src_stride;
-
-  // Same as RGB24 - reverse ordering
-
-  for (h = 0; h < ((src_height + 1) >> 1); h++){
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Y
-      yStartPtr[0] =  (uint8)((66 * inpPtr[0] + 129 * inpPtr[1]
-                                      + 25 * inpPtr[2] + 128) >> 8) + 16;
-      yStartPtr2[0] = (uint8)((66 * inpPtr2[2] + 129 * inpPtr2[1]
-                                      + 25 * inpPtr2[0] + 128) >> 8) + 16;
-      // Moving to next column
-      yStartPtr[1] = (uint8)((66 * inpPtr[3] + 129 * inpPtr[4]
-                                     + 25 * inpPtr[5]  + 128) >> 8) + 16;
-      yStartPtr2[1] = (uint8)((66 * inpPtr2[3] + 129 * inpPtr2[4]
-                                      + 25 * inpPtr2[5] + 128) >> 8 ) + 16;
-      // U
-      uStartPtr[0] = (uint8)((-38 * inpPtr[0] - 74 * inpPtr[1] +
-                                      112 * inpPtr[2] + 128) >> 8) + 128;
-      // V
-      vStartPtr[0] = (uint8)((112 * inpPtr[0] -94 * inpPtr[1] -
-                                     18 * inpPtr[2] + 128) >> 8) + 128;
-
-      yStartPtr += 2;
-      yStartPtr2 += 2;
-      uStartPtr++;
-      vStartPtr++;
-      inpPtr += 6;
-      inpPtr2 += 6;
-    } // end for w
-    yStartPtr += dst_ystride + dst_ystride - src_width;
-    yStartPtr2 += dst_ystride + dst_ystride - src_width;
-    uStartPtr += dst_ustride + dst_ustride - ((src_width + 1) >> 1);
-    vStartPtr += dst_vstride + dst_vstride - ((src_width + 1) >> 1);
-    inpPtr -= 3 * (2 * src_stride + src_width);
-    inpPtr2 -= 3 * (2 * src_stride + src_width);
-  } // end for h
-  return 0;
-}
-
-
-int
-Convert::BGRAToI420(const uint8* src_frame, int src_stride,
-                     uint8* dst_yplane, int dst_ystride,
-                     uint8* dst_uplane, int dst_ustride,
-                     uint8* dst_vplane, int dst_vstride,
-                     int src_width, int src_height)
-{
-  if (src_frame == NULL || dst_yplane == NULL ||
-      dst_vplane == NULL || dst_vplane == NULL)
-    return -1;
-
-  uint8* yStartPtr;
-  uint8* yStartPtr2;
-  uint8* uStartPtr;
-  uint8* vStartPtr;
-  const uint8* inpPtr;
-  const uint8* inpPtr2;
-  int h, w;
-
-  // Assuming RGB in a bottom up orientation.
-  yStartPtr = dst_yplane;
-  yStartPtr2 = yStartPtr + dst_ystride;
-  uStartPtr = dst_uplane;
-  vStartPtr = dst_vplane;
-  inpPtr = src_frame + src_stride * src_height * 3 - 3 * src_height;
-  inpPtr2 = inpPtr - 3 * src_stride;
-
-  for (h = 0; h < ((src_height + 1) >> 1); h++ ){
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Y
-      yStartPtr[0] =  (uint8)((66 * inpPtr[1] + 129 * inpPtr[2]
-                                      + 25 * inpPtr[3] + 128) >> 8) + 16;
-      yStartPtr2[0] = (uint8)((66 * inpPtr2[1] + 129 * inpPtr2[2]
-                                      + 25 * inpPtr2[3] + 128) >> 8) + 16;
-      // Moving to next column
-      yStartPtr[1] = (uint8)((66 * inpPtr[5] + 129 * inpPtr[6]
-                                     + 25 * inpPtr[7]  + 128) >> 8) + 16;
-      yStartPtr2[1] = (uint8)((66 * inpPtr2[5] + 129 * inpPtr2[6]
-                                      + 25 * inpPtr2[7] + 128) >> 8 ) + 16;
-      // U
-      uStartPtr[0] = (uint8)((-38 * inpPtr[1] - 74 * inpPtr[2] +
-                                      112 * inpPtr[3] + 128) >> 8) + 128;
-      // V
-      vStartPtr[0] = (uint8)((112 * inpPtr[1] -94 * inpPtr[2] -
-                                     18 * inpPtr[3] + 128) >> 8) + 128;
-
-      yStartPtr += 2;
-      yStartPtr2 += 2;
-      uStartPtr++;
-      vStartPtr++;
-      inpPtr += 6;
-      inpPtr2 += 6;
-    } // end for w
-    yStartPtr += dst_ystride + dst_ystride - src_width;
-    yStartPtr2 += dst_ystride + dst_ystride - src_width;
-    uStartPtr += dst_ustride + dst_ustride - ((src_width + 1) >> 1);
-    vStartPtr += dst_vstride + dst_vstride - ((src_width + 1) >> 1);
-    inpPtr -= 3 * (2 * src_stride + src_width);
-    inpPtr2 -= 3 * (2 * src_stride + src_width);
-  } // end for h
-  return 0;
-}
-
-
-int
-Convert::ARGBToI420(const uint8* src_frame, int src_stride,
-                     uint8* dst_yplane, int dst_ystride,
-                     uint8* dst_uplane, int dst_ustride,
-                     uint8* dst_vplane, int dst_vstride,
-                     int src_width, int src_height)
-{
-  if (src_frame == NULL || dst_yplane == NULL ||
-      dst_vplane == NULL || dst_vplane == NULL)
-    return -1;
-
-  uint8* yStartPtr;
-  uint8* yStartPtr2;
-  uint8* uStartPtr;
-  uint8* vStartPtr;
-  const uint8* inpPtr;
-  const uint8* inpPtr2;
-  int h, w;
-
-  yStartPtr = dst_yplane;
-  yStartPtr2 = yStartPtr + dst_ystride;
-  uStartPtr = dst_uplane;
-  vStartPtr = dst_vplane;
-  inpPtr = src_frame + src_stride * src_height * 3 - 3 * src_height;
-  inpPtr2 = inpPtr - 3 * src_stride;
-
-  for (h = 0; h < ((src_height + 1) >> 1); h++ ){
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Y
-      yStartPtr[0] =  (uint8)((66 * inpPtr[2] + 129 * inpPtr[1]
-                                      + 25 * inpPtr[0] + 128) >> 8) + 16;
-      yStartPtr2[0] = (uint8)((66 * inpPtr2[2] + 129 * inpPtr2[1]
-                                      + 25 * inpPtr2[0] + 128) >> 8) + 16;
-      // Moving to next column
-      yStartPtr[1] = (uint8)((66 * inpPtr[6] + 129 * inpPtr[5]
-                                     + 25 * inpPtr[4]  + 128) >> 8) + 16;
-      yStartPtr2[1] = (uint8)((66 * inpPtr2[5] + 129 * inpPtr2[4]
-                                      + 25 * inpPtr2[3] + 128) >> 8 ) + 16;
-      // U
-      uStartPtr[0] = (uint8)((-38 * inpPtr[2] - 74 * inpPtr[1] +
-                                      112 * inpPtr[0] + 128) >> 8) + 128;
-      // V
-      vStartPtr[0] = (uint8)((112 * inpPtr[2] -94 * inpPtr[1] -
-                                     18 * inpPtr[0] + 128) >> 8) + 128;
-
-      yStartPtr += 2;
-      yStartPtr2 += 2;
-      uStartPtr++;
-      vStartPtr++;
-      inpPtr += 6;
-      inpPtr2 += 6;
-    } // end for w
-    yStartPtr += dst_ystride + dst_ystride - src_width;
-    yStartPtr2 += dst_ystride + dst_ystride - src_width;
-    uStartPtr += dst_ustride + dst_ustride - ((src_width + 1) >> 1);
-    vStartPtr += dst_vstride + dst_vstride - ((src_width + 1) >> 1);
-    inpPtr -= 3 * (2 * src_stride + src_width);
-    inpPtr2 -= 3 * (2 * src_stride + src_width);
-  } // end for h
-  return 0;
-}
-
-
-int
-Convert::ABGRToI420(const uint8* src_frame, int src_stride,
-                    uint8* dst_yplane, int dst_ystride,
-                    uint8* dst_uplane, int dst_ustride,
-                    uint8* dst_vplane, int dst_vstride,
-                    int src_width, int src_height)
-{
-  if (src_frame == NULL || dst_yplane == NULL ||
-      dst_vplane == NULL || dst_vplane == NULL){
+      dst_vplane == NULL || dst_vplane == NULL) {
     return -1;
   }
-
-  uint8* yStartPtr;
-  uint8* yStartPtr2;
-  uint8* uStartPtr;
-  uint8* vStartPtr;
-  const uint8* inpPtr;
-  const uint8* inpPtr2;
-
-  yStartPtr = dst_yplane;
-  yStartPtr2 = yStartPtr + dst_ystride;
-  uStartPtr = dst_uplane;
-  vStartPtr = dst_vplane;
-  inpPtr = src_frame;
-  inpPtr2 = inpPtr + 4 * src_stride;
-  int h, w;
-
-  // RGBA in memory
-  for (h = 0; h < ((src_height + 1) >> 1); h++){
-    for (w = 0; w < ((src_width + 1) >> 1); w++){
-      // Y
-      yStartPtr[0]  = (uint8)((66 * inpPtr[0] + 129 * inpPtr[1]
-                               + 25 * inpPtr[2] + 128) >> 8) + 16;
-      yStartPtr2[0] = (uint8)((66 * inpPtr2[0] + 129 * inpPtr2[1]
-                               + 25 * inpPtr2[2] + 128) >> 8) + 16;
-      // Moving to next column
-      yStartPtr[1] =  (uint8)((66 * inpPtr[4] + 129 * inpPtr[5]
-                               + 25 * inpPtr[6] + 128) >> 8) + 16;
-      yStartPtr2[1] = (uint8)((66 * inpPtr2[4] + 129 * inpPtr2[5]
-                                + 25 * inpPtr2[6] + 128) >> 8) + 16;
-      // U
-      uStartPtr[0] = (uint8)((-38 * inpPtr[0] - 74 * inpPtr[1]
-                              + 112 * inpPtr[2] + 128) >> 8) + 128;
-      // V
-      vStartPtr[0] = (uint8)((112 * inpPtr[0] - 94 * inpPtr[1]
-                              - 18 * inpPtr[2] + 128) >> 8) + 128;
-
-      yStartPtr += 2;
-      yStartPtr2 += 2;
-      uStartPtr++;
-      vStartPtr++;
-      inpPtr += 8;
-      inpPtr2 += 8;
-    }
-
-    yStartPtr += 2 * dst_ystride - src_width;
-    yStartPtr2 += 2 * dst_ystride - src_width;
-    uStartPtr += dst_ustride + dst_ustride - ((src_width + 1) >> 1);
-    vStartPtr += dst_vstride + dst_vstride - ((src_width + 1) >> 1);
-    inpPtr += 4 * (2 * src_stride - src_width);
-    inpPtr2 += 4 * (2 * src_stride - src_width);
+  if (src_height < 0) {
+    src_height = -src_height;
+    src_frame = src_frame + src_stride * (src_height -1);
+    src_stride = -src_stride;
+  }
+  for (int y = 0; y < src_height - 1; y += 2) {
+    RGBToI420Row(src_frame, src_frame + src_stride,
+                 dst_yplane, dst_yplane + dst_ystride,
+                 dst_uplane, dst_vplane,
+                 src_width);
+    src_frame += src_stride * 2;
+    dst_yplane += dst_ystride * 2;
+    dst_uplane += dst_ustride;
+    dst_vplane += dst_vstride;
+  }
+  if (src_height & 1) {
+    RGBToI420Row(src_frame, src_frame,
+                 dst_yplane, dst_yplane,
+                 dst_uplane, dst_vplane,
+                 src_width);
   }
   return 0;
 }
 
+int
+ARGBToI420(const uint8* src_frame, int src_stride,
+           uint8* dst_yplane, int dst_ystride,
+           uint8* dst_uplane, int dst_ustride,
+           uint8* dst_vplane, int dst_vstride,
+           int src_width, int src_height) {
+  return RGBToI420(src_frame, src_stride,
+                   dst_yplane, dst_ystride,
+                   dst_uplane, dst_ustride,
+                   dst_vplane, dst_vstride,
+                   src_width, src_height, ARGBToI420Row_C);
+}
+
+int
+BGRAToI420(const uint8* src_frame, int src_stride,
+           uint8* dst_yplane, int dst_ystride,
+           uint8* dst_uplane, int dst_ustride,
+           uint8* dst_vplane, int dst_vstride,
+           int src_width, int src_height) {
+  return RGBToI420(src_frame, src_stride,
+                   dst_yplane, dst_ystride,
+                   dst_uplane, dst_ustride,
+                   dst_vplane, dst_vstride,
+                   src_width, src_height, BGRAToI420Row_C);
+}
+
+int
+ABGRToI420(const uint8* src_frame, int src_stride,
+           uint8* dst_yplane, int dst_ystride,
+           uint8* dst_uplane, int dst_ustride,
+           uint8* dst_vplane, int dst_vstride,
+           int src_width, int src_height) {
+  return RGBToI420(src_frame, src_stride,
+                   dst_yplane, dst_ystride,
+                   dst_uplane, dst_ustride,
+                   dst_vplane, dst_vstride,
+                   src_width, src_height, ABGRToI420Row_C);
+}
+
+int
+RGB24ToI420(const uint8* src_frame, int src_stride,
+            uint8* dst_yplane, int dst_ystride,
+            uint8* dst_uplane, int dst_ustride,
+            uint8* dst_vplane, int dst_vstride,
+            int src_width, int src_height) {
+  return RGBToI420(src_frame, src_stride,
+                   dst_yplane, dst_ystride,
+                   dst_uplane, dst_ustride,
+                   dst_vplane, dst_vstride,
+                   src_width, src_height, RGB24ToI420Row_C);
+}
+
+int
+RAWToI420(const uint8* src_frame, int src_stride,
+          uint8* dst_yplane, int dst_ystride,
+          uint8* dst_uplane, int dst_ustride,
+          uint8* dst_vplane, int dst_vstride,
+          int src_width, int src_height) {
+  return RGBToI420(src_frame, src_stride,
+                   dst_yplane, dst_ystride,
+                   dst_uplane, dst_ustride,
+                   dst_vplane, dst_vstride,
+                   src_width, src_height, RAWToI420Row_C);
+}
+
 inline
 uint8 Clip(int32 val)
 {
@@ -1328,4 +991,3 @@ void
 #endif
 
 } // namespace libyuv
-
diff --git a/source/format_conversion.cc b/source/format_conversion.cc
index a7e1f3c60..828015424 100644
--- a/source/format_conversion.cc
+++ b/source/format_conversion.cc
@@ -8,9 +8,6 @@
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
-
-#include "format_conversion.h"
-
 #include <assert.h>
 
 #include "common.h"
@@ -19,6 +16,9 @@
 
 namespace libyuv {
 
+// Most code in here is inspired by the material at
+// http://www.siliconimaging.com/RGB%20Bayer.htm
+
 enum {
   RED = 0,
   BLUE = 1,
@@ -274,6 +274,65 @@ static FORCE_INLINE void InterpolateBayerRGBCenter(uint8* r,
   }
 }
 
+// Converts any Bayer RGB format to ARGB.
+void BayerRGBToARGB(const uint8* src, int src_pitch, uint32 src_fourcc,
+                    uint8* dst, int dst_pitch,
+                    int width, int height) {
+  assert(width % 2 == 0);
+  assert(height % 2 == 0);
+
+  uint32 colour_map = FourCcToBayerPixelColourMap(src_fourcc);
+  int src_row_inc = src_pitch * 2 - width;
+  int dst_row_inc = dst_pitch * 2 - width * 4;
+
+  // Iterate over the 2x2 grids.
+  for (int y1 = 0; y1 < height; y1 += 2) {
+    for (int x1 = 0; x1 < width; x1 += 2) {
+      uint32 colours = colour_map;
+      // Iterate over the four pixels within them.
+      for (int y2 = 0; y2 < 2; ++y2) {
+        for (int x2 = 0; x2 < 2; ++x2) {
+          uint8 r, g, b;
+          // The low-order byte of the colour map is the current colour.
+          uint8 current_colour = static_cast<uint8>(colours);
+          colours >>= 8;
+          Position pos = GetPosition(x1 + x2, y1 + y2, width, height);
+          const uint8* src_pixel = &src[y2 * src_pitch + x2];
+          const uint8* dst_pixel = &dst[y2 * dst_pitch + x2 * 4];
+
+          // Convert from Bayer RGB to regular RGB.
+          if (pos == MIDDLE) {
+            // 99% of the image is the middle.
+            InterpolateBayerRGBCenter(&r, &g, &b,
+                                      src_pixel, src_pitch,
+                                      current_colour);
+          } else if (pos >= LEFT_EDGE) {
+            // Next most frequent is edges.
+            InterpolateBayerRGBEdge(&r, &g, &b,
+                                    src_pixel, src_pitch, pos,
+                                    current_colour);
+          } else {
+            // Last is the corners. There are only 4.
+            InterpolateBayerRGBCorner(&r, &g, &b,
+                                      src_pixel, src_pitch, pos,
+                                      current_colour);
+          }
+
+          // Store ARGB
+          dst[0] = b;
+          dst[1] = g;
+          dst[2] = r;
+          dst[3] = 255u;
+        }
+      }
+      src += 2;
+      dst += 2 * 4;
+    }
+    src += src_row_inc;
+    dst += dst_row_inc;
+  }
+}
+
 // Converts any Bayer RGB format to I420.
 void BayerRGBToI420(const uint8* src, int src_pitch, uint32 src_fourcc,
                     uint8* y, int y_pitch,
@@ -430,17 +489,16 @@ static uint32 GenerateSelector(int select0, int select1) {
          static_cast<uint32>((select1 + 12) << 24);
 }
 
-// Converts any 32 bit ARGB to any Bayer RGB format.
-void RGB32ToBayerRGB(const uint8* src_rgb, int src_pitch_rgb,
-                     uint32 src_fourcc_rgb,
-                     uint8* dst_bayer, int dst_pitch_bayer,
-                     uint32 dst_fourcc_bayer,
-                     int width, int height) {
+// Converts 32 bit ARGB to any Bayer RGB format.
+void ARGBToBayerRGB(const uint8* src_rgb, int src_pitch_rgb,
+                    uint8* dst_bayer, int dst_pitch_bayer,
+                    uint32 dst_fourcc_bayer,
+                    int width, int height) {
   assert(width % 2 == 0);
   void (*ARGBToBayerRow)(const uint8* src_argb,
                          uint8* dst_bayer, uint32 selector, int pix);
 #if defined(HAS_ARGBTOBAYERROW_SSSE3)
-  if (CpuInfo::TestCpuFlag(CpuInfo::kCpuHasSSSE3) &&
+  if (libyuv::CpuInfo::TestCpuFlag(libyuv::CpuInfo::kCpuHasSSSE3) &&
       (width % 4 == 0) &&
       IS_ALIGNED(src_rgb, 16) && (src_pitch_rgb % 16 == 0) &&
       IS_ALIGNED(dst_bayer, 4) && (dst_pitch_bayer % 4 == 0)) {
@@ -451,7 +509,6 @@ void RGB32ToBayerRGB(const uint8* src_rgb, int src_pitch_rgb,
     ARGBToBayerRow = ARGBToBayerRow_C;
   }
 
-  assert(src_fourcc_rgb == FOURCC_ARGB);
   int blue_index = 0;
   int green_index = 1;
   int red_index = 2;
diff --git a/source/planar_functions.cc b/source/planar_functions.cc
index f815e7c53..8b0a06111 100644
--- a/source/planar_functions.cc
+++ b/source/planar_functions.cc
@@ -249,13 +249,13 @@ static void I420CopyPlane2(const uint8* src, int src_pitch_0, int src_pitch_1,
 
 // Helper function to copy yuv data without scaling.  Used
 // by our jpeg conversion callbacks to incrementally fill a yuv image.
-void PlanarFunctions::I420Copy(const uint8* src_y, int src_pitch_y,
-                               const uint8* src_u, int src_pitch_u,
-                               const uint8* src_v, int src_pitch_v,
-                               uint8* dst_y, int dst_pitch_y,
-                               uint8* dst_u, int dst_pitch_u,
-                               uint8* dst_v, int dst_pitch_v,
-                               int width, int height) {
+void I420Copy(const uint8* src_y, int src_pitch_y,
+              const uint8* src_u, int src_pitch_u,
+              const uint8* src_v, int src_pitch_v,
+              uint8* dst_y, int dst_pitch_y,
+              uint8* dst_u, int dst_pitch_u,
+              uint8* dst_v, int dst_pitch_v,
+              int width, int height) {
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
@@ -276,13 +276,13 @@ void PlanarFunctions::I420Copy(const uint8* src_y, int src_pitch_y,
 
 // Helper function to copy yuv data without scaling.  Used
 // by our jpeg conversion callbacks to incrementally fill a yuv image.
-void PlanarFunctions::I422ToI420(const uint8* src_y, int src_pitch_y,
-                                 const uint8* src_u, int src_pitch_u,
-                                 const uint8* src_v, int src_pitch_v,
-                                 uint8* dst_y, int dst_pitch_y,
-                                 uint8* dst_u, int dst_pitch_u,
-                                 uint8* dst_v, int dst_pitch_v,
-                                 int width, int height) {
+void I422ToI420(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height) {
   // Negative height means invert the image.
   if (height < 0) {
     height = -height;
@@ -399,11 +399,11 @@ static void X420ToI420(const uint8* src_y,
 }
 
 // Convert M420 to I420.
-void PlanarFunctions::M420ToI420(const uint8* src_m420, int src_pitch_m420,
-                                 uint8* dst_y, int dst_pitch_y,
-                                 uint8* dst_u, int dst_pitch_u,
-                                 uint8* dst_v, int dst_pitch_v,
-                                 int width, int height) {
+void M420ToI420(const uint8* src_m420, int src_pitch_m420,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height) {
   X420ToI420(src_m420, src_pitch_m420, src_pitch_m420 * 2,
              src_m420 + src_pitch_m420 * 2, src_pitch_m420 * 3,
              dst_y, dst_pitch_y, dst_u, dst_pitch_u, dst_v, dst_pitch_v,
@@ -411,13 +411,13 @@ void PlanarFunctions::M420ToI420(const uint8* src_m420, int src_pitch_m420,
 }
 
 // Convert NV12 to I420.
-void PlanarFunctions::NV12ToI420(const uint8* src_y,
-                                 const uint8* src_uv,
-                                 int src_pitch,
-                                 uint8* dst_y, int dst_pitch_y,
-                                 uint8* dst_u, int dst_pitch_u,
-                                 uint8* dst_v, int dst_pitch_v,
-                                 int width, int height) {
+void NV12ToI420(const uint8* src_y,
+                const uint8* src_uv,
+                int src_pitch,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height) {
   X420ToI420(src_y, src_pitch, src_pitch,
              src_uv, src_pitch,
              dst_y, dst_pitch_y, dst_u, dst_pitch_u, dst_v, dst_pitch_v,
@@ -543,12 +543,12 @@ static void SplitYUY2_C(const uint8* src_yuy2,
 
 // Convert Q420 to I420.
 // Format is rows of YY/YUYV
-void PlanarFunctions::Q420ToI420(const uint8* src_y, int src_pitch_y,
-                                 const uint8* src_yuy2, int src_pitch_yuy2,
-                                 uint8* dst_y, int dst_pitch_y,
-                                 uint8* dst_u, int dst_pitch_u,
-                                 uint8* dst_v, int dst_pitch_v,
-                                 int width, int height) {
+void Q420ToI420(const uint8* src_y, int src_pitch_y,
+                const uint8* src_yuy2, int src_pitch_yuy2,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height) {
   void (*SplitYUY2)(const uint8* src_yuy2,
                     uint8* dst_y, uint8* dst_u, uint8* dst_v, int pix);
 #if defined(HAS_SPLITYUY2_SSE2)
@@ -921,11 +921,11 @@ void UYVYToI420RowY_C(const uint8* src_uyvy,
 }
 
 // Convert YUY2 to I420.
-void PlanarFunctions::YUY2ToI420(const uint8* src_yuy2, int src_pitch_yuy2,
-                                 uint8* dst_y, int dst_pitch_y,
-                                 uint8* dst_u, int dst_pitch_u,
-                                 uint8* dst_v, int dst_pitch_v,
-                                 int width, int height) {
+void YUY2ToI420(const uint8* src_yuy2, int src_pitch_yuy2,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height) {
   void (*YUY2ToI420RowUV)(const uint8* src_yuy2, int src_pitch_yuy2,
                           uint8* dst_u, uint8* dst_v, int pix);
   void (*YUY2ToI420RowY)(const uint8* src_yuy2,
@@ -961,11 +961,11 @@ void PlanarFunctions::YUY2ToI420(const uint8* src_yuy2, int src_pitch_yuy2,
 }
 
 // Convert UYVY to I420.
-void PlanarFunctions::UYVYToI420(const uint8* src_uyvy, int src_pitch_uyvy,
-                                 uint8* dst_y, int dst_pitch_y,
-                                 uint8* dst_u, int dst_pitch_u,
-                                 uint8* dst_v, int dst_pitch_v,
-                                 int width, int height) {
+void UYVYToI420(const uint8* src_uyvy, int src_pitch_uyvy,
+                uint8* dst_y, int dst_pitch_y,
+                uint8* dst_u, int dst_pitch_u,
+                uint8* dst_v, int dst_pitch_v,
+                int width, int height) {
   void (*UYVYToI420RowUV)(const uint8* src_uyvy, int src_pitch_uyvy,
                           uint8* dst_u, uint8* dst_v, int pix);
   void (*UYVYToI420RowY)(const uint8* src_uyvy,
@@ -1002,11 +1002,11 @@ void PlanarFunctions::UYVYToI420(const uint8* src_uyvy, int src_pitch_uyvy,
 
 // Convert I420 to ARGB.
 // TODO(fbarchard): Add SSSE3 version and supply C version for fallback.
-void PlanarFunctions::I420ToARGB(const uint8* src_y, int src_pitch_y,
-                                 const uint8* src_u, int src_pitch_u,
-                                 const uint8* src_v, int src_pitch_v,
-                                 uint8* dst_argb, int dst_pitch_argb,
-                                 int width, int height) {
+void I420ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
   for (int y = 0; y < height; ++y) {
     FastConvertYUVToRGB32Row(src_y, src_u, src_v, dst_argb, width);
     dst_argb += dst_pitch_argb;
@@ -1020,12 +1020,48 @@ void PlanarFunctions::I420ToARGB(const uint8* src_y, int src_pitch_y,
   EMMS();
 }
 
+// Convert I420 to BGRA.
+void I420ToBGRA(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
+  for (int y = 0; y < height; ++y) {
+    FastConvertYUVToBGRARow(src_y, src_u, src_v, dst_argb, width);
+    dst_argb += dst_pitch_argb;
+    src_y += src_pitch_y;
+    if (y & 1) {
+      src_u += src_pitch_u;
+      src_v += src_pitch_v;
+    }
+  }
+  EMMS();
+}
+
+// Convert I420 to BGRA.
+void I420ToABGR(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
+  for (int y = 0; y < height; ++y) {
+    FastConvertYUVToABGRRow(src_y, src_u, src_v, dst_argb, width);
+    dst_argb += dst_pitch_argb;
+    src_y += src_pitch_y;
+    if (y & 1) {
+      src_u += src_pitch_u;
+      src_v += src_pitch_v;
+    }
+  }
+  EMMS();
+}
+
 // Convert I422 to ARGB.
-void PlanarFunctions::I422ToARGB(const uint8* src_y, int src_pitch_y,
-                                 const uint8* src_u, int src_pitch_u,
-                                 const uint8* src_v, int src_pitch_v,
-                                 uint8* dst_argb, int dst_pitch_argb,
-                                 int width, int height) {
+void I422ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
   for (int y = 0; y < height; ++y) {
     FastConvertYUVToRGB32Row(src_y, src_u, src_v, dst_argb, width);
     dst_argb += dst_pitch_argb;
@@ -1038,11 +1074,11 @@ void PlanarFunctions::I422ToARGB(const uint8* src_y, int src_pitch_y,
 }
 
 // Convert I444 to ARGB.
-void PlanarFunctions::I444ToARGB(const uint8* src_y, int src_pitch_y,
-                                 const uint8* src_u, int src_pitch_u,
-                                 const uint8* src_v, int src_pitch_v,
-                                 uint8* dst_argb, int dst_pitch_argb,
-                                 int width, int height) {
+void I444ToARGB(const uint8* src_y, int src_pitch_y,
+                const uint8* src_u, int src_pitch_u,
+                const uint8* src_v, int src_pitch_v,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
   for (int y = 0; y < height; ++y) {
     FastConvertYUV444ToRGB32Row(src_y, src_u, src_v, dst_argb, width);
     dst_argb += dst_pitch_argb;
@@ -1055,9 +1091,9 @@ void PlanarFunctions::I444ToARGB(const uint8* src_y, int src_pitch_y,
 }
 
 // Convert I400 to ARGB.
-void PlanarFunctions::I400ToARGB_Reference(const uint8* src_y, int src_pitch_y,
-                                           uint8* dst_argb, int dst_pitch_argb,
-                                           int width, int height) {
+void I400ToARGB_Reference(const uint8* src_y, int src_pitch_y,
+                          uint8* dst_argb, int dst_pitch_argb,
+                          int width, int height) {
   for (int y = 0; y < height; ++y) {
     FastConvertYToRGB32Row(src_y, dst_argb, width);
     dst_argb += dst_pitch_argb;
@@ -1147,9 +1183,9 @@ static void I400ToARGBRow_C(const uint8* src_y, uint8* dst_argb, int pix) {
 }
 
 // Convert I400 to ARGB.
-void PlanarFunctions::I400ToARGB(const uint8* src_y, int src_pitch_y,
-                                 uint8* dst_argb, int dst_pitch_argb,
-                                 int width, int height) {
+void I400ToARGB(const uint8* src_y, int src_pitch_y,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
   void (*I400ToARGBRow)(const uint8* src_y, uint8* dst_argb, int pix);
 #if defined(HAS_I400TOARGBROW_SSE2)
   if (libyuv::CpuInfo::TestCpuFlag(libyuv::CpuInfo::kCpuHasSSE2) &&
@@ -1182,9 +1218,9 @@ static void RAWToARGBRow_C(const uint8* src_raw, uint8* dst_argb, int pix) {
 }
 
 // Convert RAW to ARGB.
-void PlanarFunctions::RAWToARGB(const uint8* src_raw, int src_pitch_raw,
-                                uint8* dst_argb, int dst_pitch_argb,
-                                int width, int height) {
+void RAWToARGB(const uint8* src_raw, int src_pitch_raw,
+               uint8* dst_argb, int dst_pitch_argb,
+               int width, int height) {
   for (int y = 0; y < height; ++y) {
     RAWToARGBRow_C(src_raw, dst_argb, width);
     src_raw += src_pitch_raw;
@@ -1204,9 +1240,9 @@ static void BG24ToARGBRow_C(const uint8* src_bg24, uint8* dst_argb, int pix) {
 }
 
 // Convert BG24 to ARGB.
-void PlanarFunctions::BG24ToARGB(const uint8* src_bg24, int src_pitch_bg24,
-                                 uint8* dst_argb, int dst_pitch_argb,
-                                 int width, int height) {
+void BG24ToARGB(const uint8* src_bg24, int src_pitch_bg24,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
   for (int y = 0; y < height; ++y) {
     BG24ToARGBRow_C(src_bg24, dst_argb, width);
     src_bg24 += src_pitch_bg24;
@@ -1226,9 +1262,9 @@ static void ABGRToARGBRow_C(const uint8* src_abgr, uint8* dst_argb, int pix) {
 }
 
 // Convert ABGR to ARGB.
-void PlanarFunctions::ABGRToARGB(const uint8* src_abgr, int src_pitch_abgr,
-                                 uint8* dst_argb, int dst_pitch_argb,
-                                 int width, int height) {
+void ABGRToARGB(const uint8* src_abgr, int src_pitch_abgr,
+                uint8* dst_argb, int dst_pitch_argb,
+                int width, int height) {
   for (int y = 0; y < height; ++y) {
     ABGRToARGBRow_C(src_abgr, dst_argb, width);
     src_abgr += src_pitch_abgr;
diff --git a/source/row.h b/source/row.h
index 48bb8440e..67119b553 100644
--- a/source/row.h
+++ b/source/row.h
@@ -14,14 +14,24 @@
 #include "basic_types.h"
 
 extern "C" {
-// Can only do 1x.
-// This is the second fastest of the scalers.
 void FastConvertYUVToRGB32Row(const uint8* y_buf,
                               const uint8* u_buf,
                               const uint8* v_buf,
                               uint8* rgb_buf,
                               int width);
 
+void FastConvertYUVToBGRARow(const uint8* y_buf,
+                             const uint8* u_buf,
+                             const uint8* v_buf,
+                             uint8* rgb_buf,
+                             int width);
+
+void FastConvertYUVToABGRRow(const uint8* y_buf,
+                             const uint8* u_buf,
+                             const uint8* v_buf,
+                             uint8* rgb_buf,
+                             int width);
+
 void FastConvertYUV444ToRGB32Row(const uint8* y_buf,
                                  const uint8* u_buf,
                                  const uint8* v_buf,
@@ -39,8 +49,12 @@ void FastConvertYToRGB32Row(const uint8* y_buf,
 #endif
 #ifdef OSX
 extern SIMD_ALIGNED(const int16 kCoefficientsRgbY[768][4]);
+extern SIMD_ALIGNED(const int16 kCoefficientsBgraY[768][4]);
+extern SIMD_ALIGNED(const int16 kCoefficientsAbgrY[768][4]);
 #else
 extern SIMD_ALIGNED(const int16 _kCoefficientsRgbY[768][4]);
+extern SIMD_ALIGNED(const int16 _kCoefficientsBgraY[768][4]);
+extern SIMD_ALIGNED(const int16 _kCoefficientsAbgrY[768][4]);
 #endif
 
 // Method to force C version.
diff --git a/source/row_posix.cc b/source/row_posix.cc
index 44c89dabd..999e8ce1a 100644
--- a/source/row_posix.cc
+++ b/source/row_posix.cc
@@ -55,6 +55,84 @@ void FastConvertYUVToRGB32Row(const uint8* y_buf,  // rdi
 );
 }
 
+void FastConvertYUVToBGRARow(const uint8* y_buf,  // rdi
+                             const uint8* u_buf,  // rsi
+                             const uint8* v_buf,  // rdx
+                             uint8* rgb_buf,      // rcx
+                             int width) {         // r8
+  asm(
+"1:"
+  "movzb  (%1),%%r10\n"
+  "lea    1(%1),%1\n"
+  "movzb  (%2),%%r11\n"
+  "lea    1(%2),%2\n"
+  "movq   2048(%5,%%r10,8),%%xmm0\n"
+  "movzb  (%0),%%r10\n"
+  "movq   4096(%5,%%r11,8),%%xmm1\n"
+  "movzb  0x1(%0),%%r11\n"
+  "paddsw %%xmm1,%%xmm0\n"
+  "movq   (%5,%%r10,8),%%xmm2\n"
+  "lea    2(%0),%0\n"
+  "movq   (%5,%%r11,8),%%xmm3\n"
+  "paddsw %%xmm0,%%xmm2\n"
+  "paddsw %%xmm0,%%xmm3\n"
+  "shufps $0x44,%%xmm3,%%xmm2\n"
+  "psraw  $0x6,%%xmm2\n"
+  "packuswb %%xmm2,%%xmm2\n"
+  "movq   %%xmm2,0x0(%3)\n"
+  "lea    8(%3),%3\n"
+  "sub    $0x2,%4\n"
+  "ja     1b\n"
+  :
+  : "r"(y_buf),  // %0
+    "r"(u_buf),  // %1
+    "r"(v_buf),  // %2
+    "r"(rgb_buf),  // %3
+    "r"(width),  // %4
+    "r" (_kCoefficientsBgraY)  // %5
+  : "memory", "r10", "r11", "xmm0", "xmm1", "xmm2", "xmm3"
+);
+}
+
+void FastConvertYUVToABGRRow(const uint8* y_buf,  // rdi
+                             const uint8* u_buf,  // rsi
+                             const uint8* v_buf,  // rdx
+                             uint8* rgb_buf,      // rcx
+                             int width) {         // r8
+  asm(
+"1:"
+  "movzb  (%1),%%r10\n"
+  "lea    1(%1),%1\n"
+  "movzb  (%2),%%r11\n"
+  "lea    1(%2),%2\n"
+  "movq   2048(%5,%%r10,8),%%xmm0\n"
+  "movzb  (%0),%%r10\n"
+  "movq   4096(%5,%%r11,8),%%xmm1\n"
+  "movzb  0x1(%0),%%r11\n"
+  "paddsw %%xmm1,%%xmm0\n"
+  "movq   (%5,%%r10,8),%%xmm2\n"
+  "lea    2(%0),%0\n"
+  "movq   (%5,%%r11,8),%%xmm3\n"
+  "paddsw %%xmm0,%%xmm2\n"
+  "paddsw %%xmm0,%%xmm3\n"
+  "shufps $0x44,%%xmm3,%%xmm2\n"
+  "psraw  $0x6,%%xmm2\n"
+  "packuswb %%xmm2,%%xmm2\n"
+  "movq   %%xmm2,0x0(%3)\n"
+  "lea    8(%3),%3\n"
+  "sub    $0x2,%4\n"
+  "ja     1b\n"
+  :
+  : "r"(y_buf),  // %0
+    "r"(u_buf),  // %1
+    "r"(v_buf),  // %2
+    "r"(rgb_buf),  // %3
+    "r"(width),  // %4
+    "r" (_kCoefficientsAbgrY)  // %5
+  : "memory", "r10", "r11", "xmm0", "xmm1", "xmm2", "xmm3"
+);
+}
+
 void FastConvertYUV444ToRGB32Row(const uint8* y_buf,  // rdi
                                  const uint8* u_buf,  // rsi
                                  const uint8* v_buf,  // rdx
@@ -166,6 +244,98 @@ void FastConvertYUVToRGB32Row(const uint8* y_buf,
   "ret\n"
 );
 
+void FastConvertYUVToBGRARow(const uint8* y_buf,
+                              const uint8* u_buf,
+                              const uint8* v_buf,
+                              uint8* rgb_buf,
+                              int width);
+  asm(
+  ".text\n"
+#if defined(OSX) || defined(IOS)
+  ".globl _FastConvertYUVToBGRARow\n"
+"_FastConvertYUVToBGRARow:\n"
+#else
+  ".global FastConvertYUVToBGRARow\n"
+"FastConvertYUVToBGRARow:\n"
+#endif
+  "pusha\n"
+  "mov    0x24(%esp),%edx\n"
+  "mov    0x28(%esp),%edi\n"
+  "mov    0x2c(%esp),%esi\n"
+  "mov    0x30(%esp),%ebp\n"
+  "mov    0x34(%esp),%ecx\n"
+
+"1:"
+  "movzbl (%edi),%eax\n"
+  "lea    1(%edi),%edi\n"
+  "movzbl (%esi),%ebx\n"
+  "lea    1(%esi),%esi\n"
+  "movq   _kCoefficientsBgraY+2048(,%eax,8),%mm0\n"
+  "movzbl (%edx),%eax\n"
+  "paddsw _kCoefficientsBgraY+4096(,%ebx,8),%mm0\n"
+  "movzbl 0x1(%edx),%ebx\n"
+  "movq   _kCoefficientsBgraY(,%eax,8),%mm1\n"
+  "lea    2(%edx),%edx\n"
+  "movq   _kCoefficientsBgraY(,%ebx,8),%mm2\n"
+  "paddsw %mm0,%mm1\n"
+  "paddsw %mm0,%mm2\n"
+  "psraw  $0x6,%mm1\n"
+  "psraw  $0x6,%mm2\n"
+  "packuswb %mm2,%mm1\n"
+  "movntq %mm1,0x0(%ebp)\n"
+  "lea    8(%ebp),%ebp\n"
+  "sub    $0x2,%ecx\n"
+  "ja     1b\n"
+  "popa\n"
+  "ret\n"
+);
+
+void FastConvertYUVToABGRRow(const uint8* y_buf,
+                              const uint8* u_buf,
+                              const uint8* v_buf,
+                              uint8* rgb_buf,
+                              int width);
+  asm(
+  ".text\n"
+#if defined(OSX) || defined(IOS)
+  ".globl _FastConvertYUVToABGRRow\n"
+"_FastConvertYUVToABGRRow:\n"
+#else
+  ".global FastConvertYUVToABGRRow\n"
+"FastConvertYUVToABGRRow:\n"
+#endif
+  "pusha\n"
+  "mov    0x24(%esp),%edx\n"
+  "mov    0x28(%esp),%edi\n"
+  "mov    0x2c(%esp),%esi\n"
+  "mov    0x30(%esp),%ebp\n"
+  "mov    0x34(%esp),%ecx\n"
+
+"1:"
+  "movzbl (%edi),%eax\n"
+  "lea    1(%edi),%edi\n"
+  "movzbl (%esi),%ebx\n"
+  "lea    1(%esi),%esi\n"
+  "movq   _kCoefficientsAbgrY+2048(,%eax,8),%mm0\n"
+  "movzbl (%edx),%eax\n"
+  "paddsw _kCoefficientsAbgrY+4096(,%ebx,8),%mm0\n"
+  "movzbl 0x1(%edx),%ebx\n"
+  "movq   _kCoefficientsAbgrY(,%eax,8),%mm1\n"
+  "lea    2(%edx),%edx\n"
+  "movq   _kCoefficientsAbgrY(,%ebx,8),%mm2\n"
+  "paddsw %mm0,%mm1\n"
+  "paddsw %mm0,%mm2\n"
+  "psraw  $0x6,%mm1\n"
+  "psraw  $0x6,%mm2\n"
+  "packuswb %mm2,%mm1\n"
+  "movntq %mm1,0x0(%ebp)\n"
+  "lea    8(%ebp),%ebp\n"
+  "sub    $0x2,%ecx\n"
+  "ja     1b\n"
+  "popa\n"
+  "ret\n"
+);
+
 void FastConvertYUV444ToRGB32Row(const uint8* y_buf,
                                  const uint8* u_buf,
                                  const uint8* v_buf,
@@ -250,7 +420,11 @@ void FastConvertYToRGB32Row(const uint8* y_buf,
 static inline void YuvPixel(uint8 y,
                             uint8 u,
                             uint8 v,
-                            uint8* rgb_buf) {
+                            uint8* rgb_buf,
+                            int ashift,
+                            int rshift,
+                            int gshift,
+                            int bshift) {
 
   int b = _kCoefficientsRgbY[256+u][0];
   int g = _kCoefficientsRgbY[256+u][1];
@@ -272,10 +446,10 @@ static inline void YuvPixel(uint8 y,
   r >>= 6;
   a >>= 6;
 
-  *reinterpret_cast<uint32*>(rgb_buf) = (packuswb(b)) |
-                                        (packuswb(g) << 8) |
-                                        (packuswb(r) << 16) |
-                                        (packuswb(a) << 24);
+  *reinterpret_cast<uint32*>(rgb_buf) = (packuswb(b) << bshift) |
+                                        (packuswb(g) << gshift) |
+                                        (packuswb(r) << rshift) |
+                                        (packuswb(a) << ashift);
 }
 
 void FastConvertYUVToRGB32Row(const uint8* y_buf,
@@ -287,10 +461,46 @@ void FastConvertYUVToRGB32Row(const uint8* y_buf,
     uint8 u = u_buf[x >> 1];
     uint8 v = v_buf[x >> 1];
     uint8 y0 = y_buf[x];
-    YuvPixel(y0, u, v, rgb_buf);
+    YuvPixel(y0, u, v, rgb_buf, 24, 16, 8, 0);
     if ((x + 1) < width) {
       uint8 y1 = y_buf[x + 1];
-      YuvPixel(y1, u, v, rgb_buf + 4);
+      YuvPixel(y1, u, v, rgb_buf + 4, 24, 16, 8, 0);
+    }
+    rgb_buf += 8;  // Advance 2 pixels.
+  }
+}
+
+void FastConvertYUVToBGRARow(const uint8* y_buf,
+                             const uint8* u_buf,
+                             const uint8* v_buf,
+                             uint8* rgb_buf,
+                             int width) {
+  for (int x = 0; x < width; x += 2) {
+    uint8 u = u_buf[x >> 1];
+    uint8 v = v_buf[x >> 1];
+    uint8 y0 = y_buf[x];
+    YuvPixel(y0, u, v, rgb_buf, 0, 8, 16, 24);
+    if ((x + 1) < width) {
+      uint8 y1 = y_buf[x + 1];
+      YuvPixel(y1, u, v, rgb_buf + 4, 0, 8, 16, 24);
+    }
+    rgb_buf += 8;  // Advance 2 pixels.
+  }
+}
+
+void FastConvertYUVToABGRRow(const uint8* y_buf,
+                             const uint8* u_buf,
+                             const uint8* v_buf,
+                             uint8* rgb_buf,
+                             int width) {
+  for (int x = 0; x < width; x += 2) {
+    uint8 u = u_buf[x >> 1];
+    uint8 v = v_buf[x >> 1];
+    uint8 y0 = y_buf[x];
+    YuvPixel(y0, u, v, rgb_buf, 24, 0, 8, 16);
+    if ((x + 1) < width) {
+      uint8 y1 = y_buf[x + 1];
+      YuvPixel(y1, u, v, rgb_buf + 4, 24, 0, 8, 16);
     }
     rgb_buf += 8;  // Advance 2 pixels.
   }
diff --git a/source/row_table.cc b/source/row_table.cc
index 71ee07308..7ce4a7ebd 100644
--- a/source/row_table.cc
+++ b/source/row_table.cc
@@ -12,6 +12,203 @@
 
 extern "C" {
 
+#define MAKETABLE(NAME) \
+SIMD_ALIGNED(const int16 NAME[256 * 3][4]) = {\
+  RGBY(0x00), RGBY(0x01), RGBY(0x02), RGBY(0x03), \
+  RGBY(0x04), RGBY(0x05), RGBY(0x06), RGBY(0x07), \
+  RGBY(0x08), RGBY(0x09), RGBY(0x0A), RGBY(0x0B), \
+  RGBY(0x0C), RGBY(0x0D), RGBY(0x0E), RGBY(0x0F), \
+  RGBY(0x10), RGBY(0x11), RGBY(0x12), RGBY(0x13), \
+  RGBY(0x14), RGBY(0x15), RGBY(0x16), RGBY(0x17), \
+  RGBY(0x18), RGBY(0x19), RGBY(0x1A), RGBY(0x1B), \
+  RGBY(0x1C), RGBY(0x1D), RGBY(0x1E), RGBY(0x1F), \
+  RGBY(0x20), RGBY(0x21), RGBY(0x22), RGBY(0x23), \
+  RGBY(0x24), RGBY(0x25), RGBY(0x26), RGBY(0x27), \
+  RGBY(0x28), RGBY(0x29), RGBY(0x2A), RGBY(0x2B), \
+  RGBY(0x2C), RGBY(0x2D), RGBY(0x2E), RGBY(0x2F), \
+  RGBY(0x30), RGBY(0x31), RGBY(0x32), RGBY(0x33), \
+  RGBY(0x34), RGBY(0x35), RGBY(0x36), RGBY(0x37), \
+  RGBY(0x38), RGBY(0x39), RGBY(0x3A), RGBY(0x3B), \
+  RGBY(0x3C), RGBY(0x3D), RGBY(0x3E), RGBY(0x3F), \
+  RGBY(0x40), RGBY(0x41), RGBY(0x42), RGBY(0x43), \
+  RGBY(0x44), RGBY(0x45), RGBY(0x46), RGBY(0x47), \
+  RGBY(0x48), RGBY(0x49), RGBY(0x4A), RGBY(0x4B), \
+  RGBY(0x4C), RGBY(0x4D), RGBY(0x4E), RGBY(0x4F), \
+  RGBY(0x50), RGBY(0x51), RGBY(0x52), RGBY(0x53), \
+  RGBY(0x54), RGBY(0x55), RGBY(0x56), RGBY(0x57), \
+  RGBY(0x58), RGBY(0x59), RGBY(0x5A), RGBY(0x5B), \
+  RGBY(0x5C), RGBY(0x5D), RGBY(0x5E), RGBY(0x5F), \
+  RGBY(0x60), RGBY(0x61), RGBY(0x62), RGBY(0x63), \
+  RGBY(0x64), RGBY(0x65), RGBY(0x66), RGBY(0x67), \
+  RGBY(0x68), RGBY(0x69), RGBY(0x6A), RGBY(0x6B), \
+  RGBY(0x6C), RGBY(0x6D), RGBY(0x6E), RGBY(0x6F), \
+  RGBY(0x70), RGBY(0x71), RGBY(0x72), RGBY(0x73), \
+  RGBY(0x74), RGBY(0x75), RGBY(0x76), RGBY(0x77), \
+  RGBY(0x78), RGBY(0x79), RGBY(0x7A), RGBY(0x7B), \
+  RGBY(0x7C), RGBY(0x7D), RGBY(0x7E), RGBY(0x7F), \
+  RGBY(0x80), RGBY(0x81), RGBY(0x82), RGBY(0x83), \
+  RGBY(0x84), RGBY(0x85), RGBY(0x86), RGBY(0x87), \
+  RGBY(0x88), RGBY(0x89), RGBY(0x8A), RGBY(0x8B), \
+  RGBY(0x8C), RGBY(0x8D), RGBY(0x8E), RGBY(0x8F), \
+  RGBY(0x90), RGBY(0x91), RGBY(0x92), RGBY(0x93), \
+  RGBY(0x94), RGBY(0x95), RGBY(0x96), RGBY(0x97), \
+  RGBY(0x98), RGBY(0x99), RGBY(0x9A), RGBY(0x9B), \
+  RGBY(0x9C), RGBY(0x9D), RGBY(0x9E), RGBY(0x9F), \
+  RGBY(0xA0), RGBY(0xA1), RGBY(0xA2), RGBY(0xA3), \
+  RGBY(0xA4), RGBY(0xA5), RGBY(0xA6), RGBY(0xA7), \
+  RGBY(0xA8), RGBY(0xA9), RGBY(0xAA), RGBY(0xAB), \
+  RGBY(0xAC), RGBY(0xAD), RGBY(0xAE), RGBY(0xAF), \
+  RGBY(0xB0), RGBY(0xB1), RGBY(0xB2), RGBY(0xB3), \
+  RGBY(0xB4), RGBY(0xB5), RGBY(0xB6), RGBY(0xB7), \
+  RGBY(0xB8), RGBY(0xB9), RGBY(0xBA), RGBY(0xBB), \
+  RGBY(0xBC), RGBY(0xBD), RGBY(0xBE), RGBY(0xBF), \
+  RGBY(0xC0), RGBY(0xC1), RGBY(0xC2), RGBY(0xC3), \
+  RGBY(0xC4), RGBY(0xC5), RGBY(0xC6), RGBY(0xC7), \
+  RGBY(0xC8), RGBY(0xC9), RGBY(0xCA), RGBY(0xCB), \
+  RGBY(0xCC), RGBY(0xCD), RGBY(0xCE), RGBY(0xCF), \
+  RGBY(0xD0), RGBY(0xD1), RGBY(0xD2), RGBY(0xD3), \
+  RGBY(0xD4), RGBY(0xD5), RGBY(0xD6), RGBY(0xD7), \
+  RGBY(0xD8), RGBY(0xD9), RGBY(0xDA), RGBY(0xDB), \
+  RGBY(0xDC), RGBY(0xDD), RGBY(0xDE), RGBY(0xDF), \
+  RGBY(0xE0), RGBY(0xE1), RGBY(0xE2), RGBY(0xE3), \
+  RGBY(0xE4), RGBY(0xE5), RGBY(0xE6), RGBY(0xE7), \
+  RGBY(0xE8), RGBY(0xE9), RGBY(0xEA), RGBY(0xEB), \
+  RGBY(0xEC), RGBY(0xED), RGBY(0xEE), RGBY(0xEF), \
+  RGBY(0xF0), RGBY(0xF1), RGBY(0xF2), RGBY(0xF3), \
+  RGBY(0xF4), RGBY(0xF5), RGBY(0xF6), RGBY(0xF7), \
+  RGBY(0xF8), RGBY(0xF9), RGBY(0xFA), RGBY(0xFB), \
+  RGBY(0xFC), RGBY(0xFD), RGBY(0xFE), RGBY(0xFF), \
+  RGBU(0x00), RGBU(0x01), RGBU(0x02), RGBU(0x03), \
+  RGBU(0x04), RGBU(0x05), RGBU(0x06), RGBU(0x07), \
+  RGBU(0x08), RGBU(0x09), RGBU(0x0A), RGBU(0x0B), \
+  RGBU(0x0C), RGBU(0x0D), RGBU(0x0E), RGBU(0x0F), \
+  RGBU(0x10), RGBU(0x11), RGBU(0x12), RGBU(0x13), \
+  RGBU(0x14), RGBU(0x15), RGBU(0x16), RGBU(0x17), \
+  RGBU(0x18), RGBU(0x19), RGBU(0x1A), RGBU(0x1B), \
+  RGBU(0x1C), RGBU(0x1D), RGBU(0x1E), RGBU(0x1F), \
+  RGBU(0x20), RGBU(0x21), RGBU(0x22), RGBU(0x23), \
+  RGBU(0x24), RGBU(0x25), RGBU(0x26), RGBU(0x27), \
+  RGBU(0x28), RGBU(0x29), RGBU(0x2A), RGBU(0x2B), \
+  RGBU(0x2C), RGBU(0x2D), RGBU(0x2E), RGBU(0x2F), \
+  RGBU(0x30), RGBU(0x31), RGBU(0x32), RGBU(0x33), \
+  RGBU(0x34), RGBU(0x35), RGBU(0x36), RGBU(0x37), \
+  RGBU(0x38), RGBU(0x39), RGBU(0x3A), RGBU(0x3B), \
+  RGBU(0x3C), RGBU(0x3D), RGBU(0x3E), RGBU(0x3F), \
+  RGBU(0x40), RGBU(0x41), RGBU(0x42), RGBU(0x43), \
+  RGBU(0x44), RGBU(0x45), RGBU(0x46), RGBU(0x47), \
+  RGBU(0x48), RGBU(0x49), RGBU(0x4A), RGBU(0x4B), \
+  RGBU(0x4C), RGBU(0x4D), RGBU(0x4E), RGBU(0x4F), \
+  RGBU(0x50), RGBU(0x51), RGBU(0x52), RGBU(0x53), \
+  RGBU(0x54), RGBU(0x55), RGBU(0x56), RGBU(0x57), \
+  RGBU(0x58), RGBU(0x59), RGBU(0x5A), RGBU(0x5B), \
+  RGBU(0x5C), RGBU(0x5D), RGBU(0x5E), RGBU(0x5F), \
+  RGBU(0x60), RGBU(0x61), RGBU(0x62), RGBU(0x63), \
+  RGBU(0x64), RGBU(0x65), RGBU(0x66), RGBU(0x67), \
+  RGBU(0x68), RGBU(0x69), RGBU(0x6A), RGBU(0x6B), \
+  RGBU(0x6C), RGBU(0x6D), RGBU(0x6E), RGBU(0x6F), \
+  RGBU(0x70), RGBU(0x71), RGBU(0x72), RGBU(0x73), \
+  RGBU(0x74), RGBU(0x75), RGBU(0x76), RGBU(0x77), \
+  RGBU(0x78), RGBU(0x79), RGBU(0x7A), RGBU(0x7B), \
+  RGBU(0x7C), RGBU(0x7D), RGBU(0x7E), RGBU(0x7F), \
+  RGBU(0x80), RGBU(0x81), RGBU(0x82), RGBU(0x83), \
+  RGBU(0x84), RGBU(0x85), RGBU(0x86), RGBU(0x87), \
+  RGBU(0x88), RGBU(0x89), RGBU(0x8A), RGBU(0x8B), \
+  RGBU(0x8C), RGBU(0x8D), RGBU(0x8E), RGBU(0x8F), \
+  RGBU(0x90), RGBU(0x91), RGBU(0x92), RGBU(0x93), \
+  RGBU(0x94), RGBU(0x95), RGBU(0x96), RGBU(0x97), \
+  RGBU(0x98), RGBU(0x99), RGBU(0x9A), RGBU(0x9B), \
+  RGBU(0x9C), RGBU(0x9D), RGBU(0x9E), RGBU(0x9F), \
+  RGBU(0xA0), RGBU(0xA1), RGBU(0xA2), RGBU(0xA3), \
+  RGBU(0xA4), RGBU(0xA5), RGBU(0xA6), RGBU(0xA7), \
+  RGBU(0xA8), RGBU(0xA9), RGBU(0xAA), RGBU(0xAB), \
+  RGBU(0xAC), RGBU(0xAD), RGBU(0xAE), RGBU(0xAF), \
+  RGBU(0xB0), RGBU(0xB1), RGBU(0xB2), RGBU(0xB3), \
+  RGBU(0xB4), RGBU(0xB5), RGBU(0xB6), RGBU(0xB7), \
+  RGBU(0xB8), RGBU(0xB9), RGBU(0xBA), RGBU(0xBB), \
+  RGBU(0xBC), RGBU(0xBD), RGBU(0xBE), RGBU(0xBF), \
+  RGBU(0xC0), RGBU(0xC1), RGBU(0xC2), RGBU(0xC3), \
+  RGBU(0xC4), RGBU(0xC5), RGBU(0xC6), RGBU(0xC7), \
+  RGBU(0xC8), RGBU(0xC9), RGBU(0xCA), RGBU(0xCB), \
+  RGBU(0xCC), RGBU(0xCD), RGBU(0xCE), RGBU(0xCF), \
+  RGBU(0xD0), RGBU(0xD1), RGBU(0xD2), RGBU(0xD3), \
+  RGBU(0xD4), RGBU(0xD5), RGBU(0xD6), RGBU(0xD7), \
+  RGBU(0xD8), RGBU(0xD9), RGBU(0xDA), RGBU(0xDB), \
+  RGBU(0xDC), RGBU(0xDD), RGBU(0xDE), RGBU(0xDF), \
+  RGBU(0xE0), RGBU(0xE1), RGBU(0xE2), RGBU(0xE3), \
+  RGBU(0xE4), RGBU(0xE5), RGBU(0xE6), RGBU(0xE7), \
+  RGBU(0xE8), RGBU(0xE9), RGBU(0xEA), RGBU(0xEB), \
+  RGBU(0xEC), RGBU(0xED), RGBU(0xEE), RGBU(0xEF), \
+  RGBU(0xF0), RGBU(0xF1), RGBU(0xF2), RGBU(0xF3), \
+  RGBU(0xF4), RGBU(0xF5), RGBU(0xF6), RGBU(0xF7), \
+  RGBU(0xF8), RGBU(0xF9), RGBU(0xFA), RGBU(0xFB), \
+  RGBU(0xFC), RGBU(0xFD), RGBU(0xFE), RGBU(0xFF), \
+  RGBV(0x00), RGBV(0x01), RGBV(0x02), RGBV(0x03), \
+  RGBV(0x04), RGBV(0x05), RGBV(0x06), RGBV(0x07), \
+  RGBV(0x08), RGBV(0x09), RGBV(0x0A), RGBV(0x0B), \
+  RGBV(0x0C), RGBV(0x0D), RGBV(0x0E), RGBV(0x0F), \
+  RGBV(0x10), RGBV(0x11), RGBV(0x12), RGBV(0x13), \
+  RGBV(0x14), RGBV(0x15), RGBV(0x16), RGBV(0x17), \
+  RGBV(0x18), RGBV(0x19), RGBV(0x1A), RGBV(0x1B), \
+  RGBV(0x1C), RGBV(0x1D), RGBV(0x1E), RGBV(0x1F), \
+  RGBV(0x20), RGBV(0x21), RGBV(0x22), RGBV(0x23), \
+  RGBV(0x24), RGBV(0x25), RGBV(0x26), RGBV(0x27), \
+  RGBV(0x28), RGBV(0x29), RGBV(0x2A), RGBV(0x2B), \
+  RGBV(0x2C), RGBV(0x2D), RGBV(0x2E), RGBV(0x2F), \
+  RGBV(0x30), RGBV(0x31), RGBV(0x32), RGBV(0x33), \
+  RGBV(0x34), RGBV(0x35), RGBV(0x36), RGBV(0x37), \
+  RGBV(0x38), RGBV(0x39), RGBV(0x3A), RGBV(0x3B), \
+  RGBV(0x3C), RGBV(0x3D), RGBV(0x3E), RGBV(0x3F), \
+  RGBV(0x40), RGBV(0x41), RGBV(0x42), RGBV(0x43), \
+  RGBV(0x44), RGBV(0x45), RGBV(0x46), RGBV(0x47), \
+  RGBV(0x48), RGBV(0x49), RGBV(0x4A), RGBV(0x4B), \
+  RGBV(0x4C), RGBV(0x4D), RGBV(0x4E), RGBV(0x4F), \
+  RGBV(0x50), RGBV(0x51), RGBV(0x52), RGBV(0x53), \
+  RGBV(0x54), RGBV(0x55), RGBV(0x56), RGBV(0x57), \
+  RGBV(0x58), RGBV(0x59), RGBV(0x5A), RGBV(0x5B), \
+  RGBV(0x5C), RGBV(0x5D), RGBV(0x5E), RGBV(0x5F), \
+  RGBV(0x60), RGBV(0x61), RGBV(0x62), RGBV(0x63), \
+  RGBV(0x64), RGBV(0x65), RGBV(0x66), RGBV(0x67), \
+  RGBV(0x68), RGBV(0x69), RGBV(0x6A), RGBV(0x6B), \
+  RGBV(0x6C), RGBV(0x6D), RGBV(0x6E), RGBV(0x6F), \
+  RGBV(0x70), RGBV(0x71), RGBV(0x72), RGBV(0x73), \
+  RGBV(0x74), RGBV(0x75), RGBV(0x76), RGBV(0x77), \
+  RGBV(0x78), RGBV(0x79), RGBV(0x7A), RGBV(0x7B), \
+  RGBV(0x7C), RGBV(0x7D), RGBV(0x7E), RGBV(0x7F), \
+  RGBV(0x80), RGBV(0x81), RGBV(0x82), RGBV(0x83), \
+  RGBV(0x84), RGBV(0x85), RGBV(0x86), RGBV(0x87), \
+  RGBV(0x88), RGBV(0x89), RGBV(0x8A), RGBV(0x8B), \
+  RGBV(0x8C), RGBV(0x8D), RGBV(0x8E), RGBV(0x8F), \
+  RGBV(0x90), RGBV(0x91), RGBV(0x92), RGBV(0x93), \
+  RGBV(0x94), RGBV(0x95), RGBV(0x96), RGBV(0x97), \
+  RGBV(0x98), RGBV(0x99), RGBV(0x9A), RGBV(0x9B), \
+  RGBV(0x9C), RGBV(0x9D), RGBV(0x9E), RGBV(0x9F), \
+  RGBV(0xA0), RGBV(0xA1), RGBV(0xA2), RGBV(0xA3), \
+  RGBV(0xA4), RGBV(0xA5), RGBV(0xA6), RGBV(0xA7), \
+  RGBV(0xA8), RGBV(0xA9), RGBV(0xAA), RGBV(0xAB), \
+  RGBV(0xAC), RGBV(0xAD), RGBV(0xAE), RGBV(0xAF), \
+  RGBV(0xB0), RGBV(0xB1), RGBV(0xB2), RGBV(0xB3), \
+  RGBV(0xB4), RGBV(0xB5), RGBV(0xB6), RGBV(0xB7), \
+  RGBV(0xB8), RGBV(0xB9), RGBV(0xBA), RGBV(0xBB), \
+  RGBV(0xBC), RGBV(0xBD), RGBV(0xBE), RGBV(0xBF), \
+  RGBV(0xC0), RGBV(0xC1), RGBV(0xC2), RGBV(0xC3), \
+  RGBV(0xC4), RGBV(0xC5), RGBV(0xC6), RGBV(0xC7), \
+  RGBV(0xC8), RGBV(0xC9), RGBV(0xCA), RGBV(0xCB), \
+  RGBV(0xCC), RGBV(0xCD), RGBV(0xCE), RGBV(0xCF), \
+  RGBV(0xD0), RGBV(0xD1), RGBV(0xD2), RGBV(0xD3), \
+  RGBV(0xD4), RGBV(0xD5), RGBV(0xD6), RGBV(0xD7), \
+  RGBV(0xD8), RGBV(0xD9), RGBV(0xDA), RGBV(0xDB), \
+  RGBV(0xDC), RGBV(0xDD), RGBV(0xDE), RGBV(0xDF), \
+  RGBV(0xE0), RGBV(0xE1), RGBV(0xE2), RGBV(0xE3), \
+  RGBV(0xE4), RGBV(0xE5), RGBV(0xE6), RGBV(0xE7), \
+  RGBV(0xE8), RGBV(0xE9), RGBV(0xEA), RGBV(0xEB), \
+  RGBV(0xEC), RGBV(0xED), RGBV(0xEE), RGBV(0xEF), \
+  RGBV(0xF0), RGBV(0xF1), RGBV(0xF2), RGBV(0xF3), \
+  RGBV(0xF4), RGBV(0xF5), RGBV(0xF6), RGBV(0xF7), \
+  RGBV(0xF8), RGBV(0xF9), RGBV(0xFA), RGBV(0xFB), \
+  RGBV(0xFC), RGBV(0xFD), RGBV(0xFE), RGBV(0xFF), \
+};
+
+// ARGB table
 #define RGBY(i) { \
   static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
   static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
@@ -34,212 +231,74 @@ extern "C" {
 }
 
 #ifdef OSX
-SIMD_ALIGNED(const int16 kCoefficientsRgbY[256 * 3][4]) = {
+MAKETABLE(kCoefficientsRgbY)
 #else
-SIMD_ALIGNED(const int16 _kCoefficientsRgbY[256 * 3][4]) = {
+MAKETABLE(_kCoefficientsRgbY)
 #endif
 
-  // Luminance table.
-  RGBY(0x00), RGBY(0x01), RGBY(0x02), RGBY(0x03),
-  RGBY(0x04), RGBY(0x05), RGBY(0x06), RGBY(0x07),
-  RGBY(0x08), RGBY(0x09), RGBY(0x0A), RGBY(0x0B),
-  RGBY(0x0C), RGBY(0x0D), RGBY(0x0E), RGBY(0x0F),
-  RGBY(0x10), RGBY(0x11), RGBY(0x12), RGBY(0x13),
-  RGBY(0x14), RGBY(0x15), RGBY(0x16), RGBY(0x17),
-  RGBY(0x18), RGBY(0x19), RGBY(0x1A), RGBY(0x1B),
-  RGBY(0x1C), RGBY(0x1D), RGBY(0x1E), RGBY(0x1F),
-  RGBY(0x20), RGBY(0x21), RGBY(0x22), RGBY(0x23),
-  RGBY(0x24), RGBY(0x25), RGBY(0x26), RGBY(0x27),
-  RGBY(0x28), RGBY(0x29), RGBY(0x2A), RGBY(0x2B),
-  RGBY(0x2C), RGBY(0x2D), RGBY(0x2E), RGBY(0x2F),
-  RGBY(0x30), RGBY(0x31), RGBY(0x32), RGBY(0x33),
-  RGBY(0x34), RGBY(0x35), RGBY(0x36), RGBY(0x37),
-  RGBY(0x38), RGBY(0x39), RGBY(0x3A), RGBY(0x3B),
-  RGBY(0x3C), RGBY(0x3D), RGBY(0x3E), RGBY(0x3F),
-  RGBY(0x40), RGBY(0x41), RGBY(0x42), RGBY(0x43),
-  RGBY(0x44), RGBY(0x45), RGBY(0x46), RGBY(0x47),
-  RGBY(0x48), RGBY(0x49), RGBY(0x4A), RGBY(0x4B),
-  RGBY(0x4C), RGBY(0x4D), RGBY(0x4E), RGBY(0x4F),
-  RGBY(0x50), RGBY(0x51), RGBY(0x52), RGBY(0x53),
-  RGBY(0x54), RGBY(0x55), RGBY(0x56), RGBY(0x57),
-  RGBY(0x58), RGBY(0x59), RGBY(0x5A), RGBY(0x5B),
-  RGBY(0x5C), RGBY(0x5D), RGBY(0x5E), RGBY(0x5F),
-  RGBY(0x60), RGBY(0x61), RGBY(0x62), RGBY(0x63),
-  RGBY(0x64), RGBY(0x65), RGBY(0x66), RGBY(0x67),
-  RGBY(0x68), RGBY(0x69), RGBY(0x6A), RGBY(0x6B),
-  RGBY(0x6C), RGBY(0x6D), RGBY(0x6E), RGBY(0x6F),
-  RGBY(0x70), RGBY(0x71), RGBY(0x72), RGBY(0x73),
-  RGBY(0x74), RGBY(0x75), RGBY(0x76), RGBY(0x77),
-  RGBY(0x78), RGBY(0x79), RGBY(0x7A), RGBY(0x7B),
-  RGBY(0x7C), RGBY(0x7D), RGBY(0x7E), RGBY(0x7F),
-  RGBY(0x80), RGBY(0x81), RGBY(0x82), RGBY(0x83),
-  RGBY(0x84), RGBY(0x85), RGBY(0x86), RGBY(0x87),
-  RGBY(0x88), RGBY(0x89), RGBY(0x8A), RGBY(0x8B),
-  RGBY(0x8C), RGBY(0x8D), RGBY(0x8E), RGBY(0x8F),
-  RGBY(0x90), RGBY(0x91), RGBY(0x92), RGBY(0x93),
-  RGBY(0x94), RGBY(0x95), RGBY(0x96), RGBY(0x97),
-  RGBY(0x98), RGBY(0x99), RGBY(0x9A), RGBY(0x9B),
-  RGBY(0x9C), RGBY(0x9D), RGBY(0x9E), RGBY(0x9F),
-  RGBY(0xA0), RGBY(0xA1), RGBY(0xA2), RGBY(0xA3),
-  RGBY(0xA4), RGBY(0xA5), RGBY(0xA6), RGBY(0xA7),
-  RGBY(0xA8), RGBY(0xA9), RGBY(0xAA), RGBY(0xAB),
-  RGBY(0xAC), RGBY(0xAD), RGBY(0xAE), RGBY(0xAF),
-  RGBY(0xB0), RGBY(0xB1), RGBY(0xB2), RGBY(0xB3),
-  RGBY(0xB4), RGBY(0xB5), RGBY(0xB6), RGBY(0xB7),
-  RGBY(0xB8), RGBY(0xB9), RGBY(0xBA), RGBY(0xBB),
-  RGBY(0xBC), RGBY(0xBD), RGBY(0xBE), RGBY(0xBF),
-  RGBY(0xC0), RGBY(0xC1), RGBY(0xC2), RGBY(0xC3),
-  RGBY(0xC4), RGBY(0xC5), RGBY(0xC6), RGBY(0xC7),
-  RGBY(0xC8), RGBY(0xC9), RGBY(0xCA), RGBY(0xCB),
-  RGBY(0xCC), RGBY(0xCD), RGBY(0xCE), RGBY(0xCF),
-  RGBY(0xD0), RGBY(0xD1), RGBY(0xD2), RGBY(0xD3),
-  RGBY(0xD4), RGBY(0xD5), RGBY(0xD6), RGBY(0xD7),
-  RGBY(0xD8), RGBY(0xD9), RGBY(0xDA), RGBY(0xDB),
-  RGBY(0xDC), RGBY(0xDD), RGBY(0xDE), RGBY(0xDF),
-  RGBY(0xE0), RGBY(0xE1), RGBY(0xE2), RGBY(0xE3),
-  RGBY(0xE4), RGBY(0xE5), RGBY(0xE6), RGBY(0xE7),
-  RGBY(0xE8), RGBY(0xE9), RGBY(0xEA), RGBY(0xEB),
-  RGBY(0xEC), RGBY(0xED), RGBY(0xEE), RGBY(0xEF),
-  RGBY(0xF0), RGBY(0xF1), RGBY(0xF2), RGBY(0xF3),
-  RGBY(0xF4), RGBY(0xF5), RGBY(0xF6), RGBY(0xF7),
-  RGBY(0xF8), RGBY(0xF9), RGBY(0xFA), RGBY(0xFB),
-  RGBY(0xFC), RGBY(0xFD), RGBY(0xFE), RGBY(0xFF),
-
-  // Chroma U table.
-  RGBU(0x00), RGBU(0x01), RGBU(0x02), RGBU(0x03),
-  RGBU(0x04), RGBU(0x05), RGBU(0x06), RGBU(0x07),
-  RGBU(0x08), RGBU(0x09), RGBU(0x0A), RGBU(0x0B),
-  RGBU(0x0C), RGBU(0x0D), RGBU(0x0E), RGBU(0x0F),
-  RGBU(0x10), RGBU(0x11), RGBU(0x12), RGBU(0x13),
-  RGBU(0x14), RGBU(0x15), RGBU(0x16), RGBU(0x17),
-  RGBU(0x18), RGBU(0x19), RGBU(0x1A), RGBU(0x1B),
-  RGBU(0x1C), RGBU(0x1D), RGBU(0x1E), RGBU(0x1F),
-  RGBU(0x20), RGBU(0x21), RGBU(0x22), RGBU(0x23),
-  RGBU(0x24), RGBU(0x25), RGBU(0x26), RGBU(0x27),
-  RGBU(0x28), RGBU(0x29), RGBU(0x2A), RGBU(0x2B),
-  RGBU(0x2C), RGBU(0x2D), RGBU(0x2E), RGBU(0x2F),
-  RGBU(0x30), RGBU(0x31), RGBU(0x32), RGBU(0x33),
-  RGBU(0x34), RGBU(0x35), RGBU(0x36), RGBU(0x37),
-  RGBU(0x38), RGBU(0x39), RGBU(0x3A), RGBU(0x3B),
-  RGBU(0x3C), RGBU(0x3D), RGBU(0x3E), RGBU(0x3F),
-  RGBU(0x40), RGBU(0x41), RGBU(0x42), RGBU(0x43),
-  RGBU(0x44), RGBU(0x45), RGBU(0x46), RGBU(0x47),
-  RGBU(0x48), RGBU(0x49), RGBU(0x4A), RGBU(0x4B),
-  RGBU(0x4C), RGBU(0x4D), RGBU(0x4E), RGBU(0x4F),
-  RGBU(0x50), RGBU(0x51), RGBU(0x52), RGBU(0x53),
-  RGBU(0x54), RGBU(0x55), RGBU(0x56), RGBU(0x57),
-  RGBU(0x58), RGBU(0x59), RGBU(0x5A), RGBU(0x5B),
-  RGBU(0x5C), RGBU(0x5D), RGBU(0x5E), RGBU(0x5F),
-  RGBU(0x60), RGBU(0x61), RGBU(0x62), RGBU(0x63),
-  RGBU(0x64), RGBU(0x65), RGBU(0x66), RGBU(0x67),
-  RGBU(0x68), RGBU(0x69), RGBU(0x6A), RGBU(0x6B),
-  RGBU(0x6C), RGBU(0x6D), RGBU(0x6E), RGBU(0x6F),
-  RGBU(0x70), RGBU(0x71), RGBU(0x72), RGBU(0x73),
-  RGBU(0x74), RGBU(0x75), RGBU(0x76), RGBU(0x77),
-  RGBU(0x78), RGBU(0x79), RGBU(0x7A), RGBU(0x7B),
-  RGBU(0x7C), RGBU(0x7D), RGBU(0x7E), RGBU(0x7F),
-  RGBU(0x80), RGBU(0x81), RGBU(0x82), RGBU(0x83),
-  RGBU(0x84), RGBU(0x85), RGBU(0x86), RGBU(0x87),
-  RGBU(0x88), RGBU(0x89), RGBU(0x8A), RGBU(0x8B),
-  RGBU(0x8C), RGBU(0x8D), RGBU(0x8E), RGBU(0x8F),
-  RGBU(0x90), RGBU(0x91), RGBU(0x92), RGBU(0x93),
-  RGBU(0x94), RGBU(0x95), RGBU(0x96), RGBU(0x97),
-  RGBU(0x98), RGBU(0x99), RGBU(0x9A), RGBU(0x9B),
-  RGBU(0x9C), RGBU(0x9D), RGBU(0x9E), RGBU(0x9F),
-  RGBU(0xA0), RGBU(0xA1), RGBU(0xA2), RGBU(0xA3),
-  RGBU(0xA4), RGBU(0xA5), RGBU(0xA6), RGBU(0xA7),
-  RGBU(0xA8), RGBU(0xA9), RGBU(0xAA), RGBU(0xAB),
-  RGBU(0xAC), RGBU(0xAD), RGBU(0xAE), RGBU(0xAF),
-  RGBU(0xB0), RGBU(0xB1), RGBU(0xB2), RGBU(0xB3),
-  RGBU(0xB4), RGBU(0xB5), RGBU(0xB6), RGBU(0xB7),
-  RGBU(0xB8), RGBU(0xB9), RGBU(0xBA), RGBU(0xBB),
-  RGBU(0xBC), RGBU(0xBD), RGBU(0xBE), RGBU(0xBF),
-  RGBU(0xC0), RGBU(0xC1), RGBU(0xC2), RGBU(0xC3),
-  RGBU(0xC4), RGBU(0xC5), RGBU(0xC6), RGBU(0xC7),
-  RGBU(0xC8), RGBU(0xC9), RGBU(0xCA), RGBU(0xCB),
-  RGBU(0xCC), RGBU(0xCD), RGBU(0xCE), RGBU(0xCF),
-  RGBU(0xD0), RGBU(0xD1), RGBU(0xD2), RGBU(0xD3),
-  RGBU(0xD4), RGBU(0xD5), RGBU(0xD6), RGBU(0xD7),
-  RGBU(0xD8), RGBU(0xD9), RGBU(0xDA), RGBU(0xDB),
-  RGBU(0xDC), RGBU(0xDD), RGBU(0xDE), RGBU(0xDF),
-  RGBU(0xE0), RGBU(0xE1), RGBU(0xE2), RGBU(0xE3),
-  RGBU(0xE4), RGBU(0xE5), RGBU(0xE6), RGBU(0xE7),
-  RGBU(0xE8), RGBU(0xE9), RGBU(0xEA), RGBU(0xEB),
-  RGBU(0xEC), RGBU(0xED), RGBU(0xEE), RGBU(0xEF),
-  RGBU(0xF0), RGBU(0xF1), RGBU(0xF2), RGBU(0xF3),
-  RGBU(0xF4), RGBU(0xF5), RGBU(0xF6), RGBU(0xF7),
-  RGBU(0xF8), RGBU(0xF9), RGBU(0xFA), RGBU(0xFB),
-  RGBU(0xFC), RGBU(0xFD), RGBU(0xFE), RGBU(0xFF),
-
-  // Chroma V table.
-  RGBV(0x00), RGBV(0x01), RGBV(0x02), RGBV(0x03),
-  RGBV(0x04), RGBV(0x05), RGBV(0x06), RGBV(0x07),
-  RGBV(0x08), RGBV(0x09), RGBV(0x0A), RGBV(0x0B),
-  RGBV(0x0C), RGBV(0x0D), RGBV(0x0E), RGBV(0x0F),
-  RGBV(0x10), RGBV(0x11), RGBV(0x12), RGBV(0x13),
-  RGBV(0x14), RGBV(0x15), RGBV(0x16), RGBV(0x17),
-  RGBV(0x18), RGBV(0x19), RGBV(0x1A), RGBV(0x1B),
-  RGBV(0x1C), RGBV(0x1D), RGBV(0x1E), RGBV(0x1F),
-  RGBV(0x20), RGBV(0x21), RGBV(0x22), RGBV(0x23),
-  RGBV(0x24), RGBV(0x25), RGBV(0x26), RGBV(0x27),
-  RGBV(0x28), RGBV(0x29), RGBV(0x2A), RGBV(0x2B),
-  RGBV(0x2C), RGBV(0x2D), RGBV(0x2E), RGBV(0x2F),
-  RGBV(0x30), RGBV(0x31), RGBV(0x32), RGBV(0x33),
-  RGBV(0x34), RGBV(0x35), RGBV(0x36), RGBV(0x37),
-  RGBV(0x38), RGBV(0x39), RGBV(0x3A), RGBV(0x3B),
-  RGBV(0x3C), RGBV(0x3D), RGBV(0x3E), RGBV(0x3F),
-  RGBV(0x40), RGBV(0x41), RGBV(0x42), RGBV(0x43),
-  RGBV(0x44), RGBV(0x45), RGBV(0x46), RGBV(0x47),
-  RGBV(0x48), RGBV(0x49), RGBV(0x4A), RGBV(0x4B),
-  RGBV(0x4C), RGBV(0x4D), RGBV(0x4E), RGBV(0x4F),
-  RGBV(0x50), RGBV(0x51), RGBV(0x52), RGBV(0x53),
-  RGBV(0x54), RGBV(0x55), RGBV(0x56), RGBV(0x57),
-  RGBV(0x58), RGBV(0x59), RGBV(0x5A), RGBV(0x5B),
-  RGBV(0x5C), RGBV(0x5D), RGBV(0x5E), RGBV(0x5F),
-  RGBV(0x60), RGBV(0x61), RGBV(0x62), RGBV(0x63),
-  RGBV(0x64), RGBV(0x65), RGBV(0x66), RGBV(0x67),
-  RGBV(0x68), RGBV(0x69), RGBV(0x6A), RGBV(0x6B),
-  RGBV(0x6C), RGBV(0x6D), RGBV(0x6E), RGBV(0x6F),
-  RGBV(0x70), RGBV(0x71), RGBV(0x72), RGBV(0x73),
-  RGBV(0x74), RGBV(0x75), RGBV(0x76), RGBV(0x77),
-  RGBV(0x78), RGBV(0x79), RGBV(0x7A), RGBV(0x7B),
-  RGBV(0x7C), RGBV(0x7D), RGBV(0x7E), RGBV(0x7F),
-  RGBV(0x80), RGBV(0x81), RGBV(0x82), RGBV(0x83),
-  RGBV(0x84), RGBV(0x85), RGBV(0x86), RGBV(0x87),
-  RGBV(0x88), RGBV(0x89), RGBV(0x8A), RGBV(0x8B),
-  RGBV(0x8C), RGBV(0x8D), RGBV(0x8E), RGBV(0x8F),
-  RGBV(0x90), RGBV(0x91), RGBV(0x92), RGBV(0x93),
-  RGBV(0x94), RGBV(0x95), RGBV(0x96), RGBV(0x97),
-  RGBV(0x98), RGBV(0x99), RGBV(0x9A), RGBV(0x9B),
-  RGBV(0x9C), RGBV(0x9D), RGBV(0x9E), RGBV(0x9F),
-  RGBV(0xA0), RGBV(0xA1), RGBV(0xA2), RGBV(0xA3),
-  RGBV(0xA4), RGBV(0xA5), RGBV(0xA6), RGBV(0xA7),
-  RGBV(0xA8), RGBV(0xA9), RGBV(0xAA), RGBV(0xAB),
-  RGBV(0xAC), RGBV(0xAD), RGBV(0xAE), RGBV(0xAF),
-  RGBV(0xB0), RGBV(0xB1), RGBV(0xB2), RGBV(0xB3),
-  RGBV(0xB4), RGBV(0xB5), RGBV(0xB6), RGBV(0xB7),
-  RGBV(0xB8), RGBV(0xB9), RGBV(0xBA), RGBV(0xBB),
-  RGBV(0xBC), RGBV(0xBD), RGBV(0xBE), RGBV(0xBF),
-  RGBV(0xC0), RGBV(0xC1), RGBV(0xC2), RGBV(0xC3),
-  RGBV(0xC4), RGBV(0xC5), RGBV(0xC6), RGBV(0xC7),
-  RGBV(0xC8), RGBV(0xC9), RGBV(0xCA), RGBV(0xCB),
-  RGBV(0xCC), RGBV(0xCD), RGBV(0xCE), RGBV(0xCF),
-  RGBV(0xD0), RGBV(0xD1), RGBV(0xD2), RGBV(0xD3),
-  RGBV(0xD4), RGBV(0xD5), RGBV(0xD6), RGBV(0xD7),
-  RGBV(0xD8), RGBV(0xD9), RGBV(0xDA), RGBV(0xDB),
-  RGBV(0xDC), RGBV(0xDD), RGBV(0xDE), RGBV(0xDF),
-  RGBV(0xE0), RGBV(0xE1), RGBV(0xE2), RGBV(0xE3),
-  RGBV(0xE4), RGBV(0xE5), RGBV(0xE6), RGBV(0xE7),
-  RGBV(0xE8), RGBV(0xE9), RGBV(0xEA), RGBV(0xEB),
-  RGBV(0xEC), RGBV(0xED), RGBV(0xEE), RGBV(0xEF),
-  RGBV(0xF0), RGBV(0xF1), RGBV(0xF2), RGBV(0xF3),
-  RGBV(0xF4), RGBV(0xF5), RGBV(0xF6), RGBV(0xF7),
-  RGBV(0xF8), RGBV(0xF9), RGBV(0xFA), RGBV(0xFB),
-  RGBV(0xFC), RGBV(0xFD), RGBV(0xFE), RGBV(0xFF),
-};
-
 #undef RGBY
 #undef RGBU
 #undef RGBV
 
+// BGRA table
+#define RGBY(i) { \
+  static_cast<int16>(256 * 64 - 1), \
+  static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
+  static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
+  static_cast<int16>(1.164 * 64 * (i - 16) + 0.5) \
+}
+
+#define RGBU(i) { \
+  0, \
+  0, \
+  static_cast<int16>(-0.391 * 64 * (i - 128) + 0.5), \
+  static_cast<int16>(2.018 * 64 * (i - 128) + 0.5) \
+}
+
+#define RGBV(i) { \
+  0, \
+  static_cast<int16>(1.596 * 64 * (i - 128) + 0.5), \
+  static_cast<int16>(-0.813 * 64 * (i - 128) + 0.5), \
+  0 \
+}
+
+#ifdef OSX
+MAKETABLE(kCoefficientsBgraY)
+#else
+MAKETABLE(_kCoefficientsBgraY)
+#endif
+
+
+#undef RGBY
+#undef RGBU
+#undef RGBV
+
+// ABGR table
+#define RGBY(i) { \
+  static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
+  static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
+  static_cast<int16>(1.164 * 64 * (i - 16) + 0.5), \
+  static_cast<int16>(256 * 64 - 1) \
+}
+
+#define RGBU(i) { \
+  0, \
+  static_cast<int16>(-0.391 * 64 * (i - 128) + 0.5), \
+  static_cast<int16>(2.018 * 64 * (i - 128) + 0.5), \
+  0 \
+}
+
+#define RGBV(i) { \
+  static_cast<int16>(1.596 * 64 * (i - 128) + 0.5), \
+  static_cast<int16>(-0.813 * 64 * (i - 128) + 0.5), \
+  0, \
+  0 \
+}
+
+#ifdef OSX
+MAKETABLE(kCoefficientsAbgrY)
+#else
+MAKETABLE(_kCoefficientsAbgrY)
+#endif
+
 }  // extern "C"
diff --git a/source/row_win.cc b/source/row_win.cc
index e26cc9934..f127f4835 100644
--- a/source/row_win.cc
+++ b/source/row_win.cc
@@ -7,13 +7,8 @@
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
- 
- 
-#include "row.h"
 
-#define kCoefficientsRgbY _kCoefficientsRgbY + 0
-#define kCoefficientsRgbU _kCoefficientsRgbY + 2048
-#define kCoefficientsRgbV _kCoefficientsRgbY + 4096
+#include "row.h"
 
 extern "C" {
 
@@ -25,24 +20,106 @@ void FastConvertYUVToRGB32Row(const uint8* y_buf,
                               int width) {
   __asm {
     pushad
-    mov       edx, [esp + 32 + 4]   // Y
-    mov       edi, [esp + 32 + 8]   // U
-    mov       esi, [esp + 32 + 12]  // V
-    mov       ebp, [esp + 32 + 16]  // rgb
-    mov       ecx, [esp + 32 + 20]  // width
+    mov       edx, [esp + 32 + 4]
+    mov       edi, [esp + 32 + 8]
+    mov       esi, [esp + 32 + 12]
+    mov       ebp, [esp + 32 + 16]
+    mov       ecx, [esp + 32 + 20]
 
  convertloop :
     movzx     eax, byte ptr [edi]
     lea       edi, [edi + 1]
     movzx     ebx, byte ptr [esi]
     lea       esi, [esi + 1]
-    movq      mm0, [kCoefficientsRgbU + 8 * eax]
+    movq      mm0, [_kCoefficientsRgbY + 2048 + 8 * eax]
     movzx     eax, byte ptr [edx]
-    paddsw    mm0, [kCoefficientsRgbV + 8 * ebx]
+    paddsw    mm0, [_kCoefficientsRgbY + 4096 + 8 * ebx]
     movzx     ebx, byte ptr [edx + 1]
-    movq      mm1, [kCoefficientsRgbY + 8 * eax]
+    movq      mm1, [_kCoefficientsRgbY + 8 * eax]
     lea       edx, [edx + 2]
-    movq      mm2, [kCoefficientsRgbY + 8 * ebx]
+    movq      mm2, [_kCoefficientsRgbY + 8 * ebx]
+    paddsw    mm1, mm0
+    paddsw    mm2, mm0
+    psraw     mm1, 6
+    psraw     mm2, 6
+    packuswb  mm1, mm2
+    movntq    [ebp], mm1
+    lea       ebp, [ebp + 8]
+    sub       ecx, 2
+    ja        convertloop
+
+    popad
+    ret
+  }
+}
+
+__declspec(naked)
+void FastConvertYUVToBGRARow(const uint8* y_buf,
+                             const uint8* u_buf,
+                             const uint8* v_buf,
+                             uint8* rgb_buf,
+                             int width) {
+  __asm {
+    pushad
+    mov       edx, [esp + 32 + 4]
+    mov       edi, [esp + 32 + 8]
+    mov       esi, [esp + 32 + 12]
+    mov       ebp, [esp + 32 + 16]
+    mov       ecx, [esp + 32 + 20]
+
+ convertloop :
+    movzx     eax, byte ptr [edi]
+    lea       edi, [edi + 1]
+    movzx     ebx, byte ptr [esi]
+    lea       esi, [esi + 1]
+    movq      mm0, [_kCoefficientsBgraY + 2048 + 8 * eax]
+    movzx     eax, byte ptr [edx]
+    paddsw    mm0, [_kCoefficientsBgraY + 4096 + 8 * ebx]
+    movzx     ebx, byte ptr [edx + 1]
+    movq      mm1, [_kCoefficientsBgraY + 8 * eax]
+    lea       edx, [edx + 2]
+    movq      mm2, [_kCoefficientsBgraY + 8 * ebx]
+    paddsw    mm1, mm0
+    paddsw    mm2, mm0
+    psraw     mm1, 6
+    psraw     mm2, 6
+    packuswb  mm1, mm2
+    movntq    [ebp], mm1
+    lea       ebp, [ebp + 8]
+    sub       ecx, 2
+    ja        convertloop
+
+    popad
+    ret
+  }
+}
+
+__declspec(naked)
+void FastConvertYUVToABGRRow(const uint8* y_buf,
+                             const uint8* u_buf,
+                             const uint8* v_buf,
+                             uint8* rgb_buf,
+                             int width) {
+  __asm {
+    pushad
+    mov       edx, [esp + 32 + 4]
+    mov       edi, [esp + 32 + 8]
+    mov       esi, [esp + 32 + 12]
+    mov       ebp, [esp + 32 + 16]
+    mov       ecx, [esp + 32 + 20]
+
+ convertloop :
+    movzx     eax, byte ptr [edi]
+    lea       edi, [edi + 1]
+    movzx     ebx, byte ptr [esi]
+    lea       esi, [esi + 1]
+    movq      mm0, [_kCoefficientsAbgrY + 2048 + 8 * eax]
+    movzx     eax, byte ptr [edx]
+    paddsw    mm0, [_kCoefficientsAbgrY + 4096 + 8 * ebx]
+    movzx     ebx, byte ptr [edx + 1]
+    movq      mm1, [_kCoefficientsAbgrY + 8 * eax]
+    lea       edx, [edx + 2]
+    movq      mm2, [_kCoefficientsAbgrY + 8 * ebx]
     paddsw    mm1, mm0
     paddsw    mm2, mm0
     psraw     mm1, 6
@@ -77,11 +154,11 @@ void FastConvertYUV444ToRGB32Row(const uint8* y_buf,
     lea       edi, [edi + 1]
     movzx     ebx, byte ptr [esi]
     lea       esi, [esi + 1]
-    movq      mm0, [kCoefficientsRgbU + 8 * eax]
+    movq      mm0, [_kCoefficientsRgbY + 2048 + 8 * eax]
     movzx     eax, byte ptr [edx]
-    paddsw    mm0, [kCoefficientsRgbV + 8 * ebx]
+    paddsw    mm0, [_kCoefficientsRgbY + 4096 + 8 * ebx]
     lea       edx, [edx + 1]
-    paddsw    mm0, [kCoefficientsRgbY + 8 * eax]
+    paddsw    mm0, [_kCoefficientsRgbY + 8 * eax]
     psraw     mm0, 6
     packuswb  mm0, mm0
     movd      [ebp], mm0
@@ -106,10 +183,10 @@ void FastConvertYToRGB32Row(const uint8* y_buf,
 
  convertloop :
     movzx     ebx, byte ptr [eax]
-    movq      mm0, [kCoefficientsRgbY + 8 * ebx]
+    movq      mm0, [_kCoefficientsRgbY + 8 * ebx]
     psraw     mm0, 6
     movzx     ebx, byte ptr [eax + 1]
-    movq      mm1, [kCoefficientsRgbY + 8 * ebx]
+    movq      mm1, [_kCoefficientsRgbY + 8 * ebx]
     psraw     mm1, 6
     packuswb  mm0, mm1
     lea       eax, [eax + 2]
diff --git a/source/scale.cc b/source/scale.cc
index 2efcb8579..f056de50b 100644
--- a/source/scale.cc
+++ b/source/scale.cc
@@ -46,8 +46,12 @@ namespace libyuv {
 // NOT the optimized versions. Useful for debugging and
 // when comparing the quality of the resulting YUV planes
 // as produced by the optimized and non-optimized versions.
-bool YuvScaler::use_reference_impl_ = false;
 
+bool use_reference_impl_ = false;
+
+void SetUseReferenceImpl(bool use) {
+  use_reference_impl_ = use;
+}
 
 /**
  * NEON downscalers with interpolation.
@@ -2790,13 +2794,13 @@ static void ScalePlane(const uint8 *in, int32 istride,
  * suitable for handling the desired resolutions.
  *
  */
-bool YuvScaler::Scale(const uint8 *inY, const uint8 *inU, const uint8 *inV,
-                      int32 istrideY, int32 istrideU, int32 istrideV,
-                      int32 iwidth, int32 iheight,
-                      uint8 *outY, uint8 *outU, uint8 *outV,
-                      int32 ostrideY, int32 ostrideU, int32 ostrideV,
-                      int32 owidth, int32 oheight,
-                      bool interpolate) {
+bool Scale(const uint8 *inY, const uint8 *inU, const uint8 *inV,
+           int32 istrideY, int32 istrideU, int32 istrideV,
+           int32 iwidth, int32 iheight,
+           uint8 *outY, uint8 *outU, uint8 *outV,
+           int32 ostrideY, int32 ostrideU, int32 ostrideV,
+           int32 owidth, int32 oheight,
+           bool interpolate) {
   if (!inY || !inU || !inV || iwidth <= 0 || iheight <= 0 ||
       !outY || !outU || !outV || owidth <= 0 || oheight <= 0) {
     return false;
@@ -2818,9 +2822,9 @@ bool YuvScaler::Scale(const uint8 *inY, const uint8 *inU, const uint8 *inV,
   return true;
 }
 
-bool YuvScaler::Scale(const uint8 *in, int32 iwidth, int32 iheight,
-                      uint8 *out, int32 owidth, int32 oheight, int32 ooffset,
-                      bool interpolate) {
+bool Scale(const uint8 *in, int32 iwidth, int32 iheight,
+           uint8 *out, int32 owidth, int32 oheight, int32 ooffset,
+           bool interpolate) {
   if (!in || iwidth <= 0 || iheight <= 0 ||
       !out || owidth <= 0 || oheight <= 0 || ooffset < 0 ||
       ooffset >= oheight) {