Add MSA optimized I422ToRGB565Row_MSA, I422ToARGB4444Row_MSA and I422ToARGB1555Row_MSA functions

R=fbarchard@google.com
BUG=libyuv:634

Performance Gain (vs C vectorized)
I422ToRGB565Row_MSA             : ~1.5x
I422ToRGB565Row_Any_MSA         : ~1.5x
I422ToARGB4444Row_MSA           : ~1.4x
I422ToARGB4444Row_Any_MSA       : ~1.4x
I422ToARGB1555Row_MSA           : ~1.4x
I422ToARGB1555Row_Any_MSA       : ~1.4x

Performance Gain (vs C non-vectorized)
I422ToRGB565Row_MSA             : ~6.8x
I422ToRGB565Row_Any_MSA         : ~6.8x
I422ToARGB4444Row_MSA           : ~6.6x
I422ToARGB4444Row_Any_MSA       : ~6.6x
I422ToARGB1555Row_MSA           : ~6.6x
I422ToARGB1555Row_Any_MSA       : ~6.6x

Review URL: https://codereview.chromium.org/2445343007 .

include/libyuv/row.h

@@ -678,6 +678,24 @@ void I422ToRGB24Row_MSA(const uint8* src_y,
                         uint8* dst_rgb24,
                         const struct YuvConstants* yuvconstants,
                         int width);
+void I422ToRGB565Row_MSA(const uint8* src_y,
+                         const uint8* src_u,
+                         const uint8* src_v,
+                         uint8* dst_rgb565,
+                         const struct YuvConstants* yuvconstants,
+                         int width);
+void I422ToARGB4444Row_MSA(const uint8* src_y,
+                           const uint8* src_u,
+                           const uint8* src_v,
+                           uint8* dst_argb4444,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
+void I422ToARGB1555Row_MSA(const uint8* src_y,
+                           const uint8* src_u,
+                           const uint8* src_v,
+                           uint8* dst_argb1555,
+                           const struct YuvConstants* yuvconstants,
+                           int width);
 
 void ARGBToYRow_AVX2(const uint8* src_argb, uint8* dst_y, int width);
 void ARGBToYRow_Any_AVX2(const uint8* src_argb, uint8* dst_y, int width);
@@ -1681,6 +1699,24 @@ void I422ToRGB24Row_Any_MSA(const uint8* src_y,
                             uint8* dst_rgb24,
                             const struct YuvConstants* yuvconstants,
                             int width);
+void I422ToRGB565Row_Any_MSA(const uint8* src_y,
+                             const uint8* src_u,
+                             const uint8* src_v,
+                             uint8* dst_rgb565,
+                             const struct YuvConstants* yuvconstants,
+                             int width);
+void I422ToARGB4444Row_Any_MSA(const uint8* src_y,
+                               const uint8* src_u,
+                               const uint8* src_v,
+                               uint8* dst_argb4444,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
+void I422ToARGB1555Row_Any_MSA(const uint8* src_y,
+                               const uint8* src_u,
+                               const uint8* src_v,
+                               uint8* dst_argb1555,
+                               const struct YuvConstants* yuvconstants,
+                               int width);
 
 void YUY2ToYRow_AVX2(const uint8* src_yuy2, uint8* dst_y, int width);
 void YUY2ToUVRow_AVX2(const uint8* src_yuy2, int stride_yuy2,

source/convert_from.cc

@@ -657,6 +657,14 @@ int I420ToARGB1555(const uint8* src_y, int src_stride_y,
     }
   }
 #endif
+#if defined(HAS_I422TOARGB1555ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGB1555Row = I422ToARGB1555Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB1555Row = I422ToARGB1555Row_MSA;
+    }
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     I422ToARGB1555Row(src_y, src_u, src_v, dst_argb1555, &kYuvI601Constants,
@@ -720,6 +728,14 @@ int I420ToARGB4444(const uint8* src_y, int src_stride_y,
     }
   }
 #endif
+#if defined(HAS_I422TOARGB4444ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToARGB4444Row = I422ToARGB4444Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToARGB4444Row = I422ToARGB4444Row_MSA;
+    }
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     I422ToARGB4444Row(src_y, src_u, src_v, dst_argb4444, &kYuvI601Constants,
@@ -782,6 +798,14 @@ int I420ToRGB565(const uint8* src_y, int src_stride_y,
     }
   }
 #endif
+#if defined(HAS_I422TORGB565ROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    I422ToRGB565Row = I422ToRGB565Row_Any_MSA;
+    if (IS_ALIGNED(width, 8)) {
+      I422ToRGB565Row = I422ToRGB565Row_MSA;
+    }
+  }
+#endif
 
   for (y = 0; y < height; ++y) {
     I422ToRGB565Row(src_y, src_u, src_v, dst_rgb565, &kYuvI601Constants, width);

source/row_any.cc

@@ -172,6 +172,9 @@ ANY31C(I422ToRGB565Row_Any_NEON, I422ToRGB565Row_NEON, 1, 0, 2, 7)
 ANY31C(I422ToARGBRow_Any_MSA, I422ToARGBRow_MSA, 1, 0, 4, 7)
 ANY31C(I422ToRGBARow_Any_MSA, I422ToRGBARow_MSA, 1, 0, 4, 7)
 ANY31C(I422ToRGB24Row_Any_MSA, I422ToRGB24Row_MSA, 1, 0, 3, 15)
+ANY31C(I422ToARGB4444Row_Any_MSA, I422ToARGB4444Row_MSA, 1, 0, 2, 7)
+ANY31C(I422ToARGB1555Row_Any_MSA, I422ToARGB1555Row_MSA, 1, 0, 2, 7)
+ANY31C(I422ToRGB565Row_Any_MSA, I422ToRGB565Row_MSA, 1, 0, 2, 7)
 #endif
 #undef ANY31C