Add MSA optimized ARGB4444ToI420 and ARGB4444ToARGB functions

source/convert.cc

Index: source/convert.cc
diff --git a/source/convert.cc b/source/convert.cc
index ba07b3959c71d8e7b46abbd43d08f4aebf14db1b..f2eafe216d45acc36ef0b11b164defcbf8729d52 100644
--- a/source/convert.cc
+++ b/source/convert.cc
@@ -1229,7 +1229,7 @@ int ARGB4444ToI420(const uint8* src_argb4444, int src_stride_argb4444,
                    uint8* dst_v, int dst_stride_v,
                    int width, int height) {
   int y;
-#if defined(HAS_ARGB4444TOYROW_NEON)
+#if (defined(HAS_ARGB4444TOYROW_NEON) || defined(HAS_ARGB4444TOYROW_MSA))

[fbarchard1, 2016/10/14 21:35:16]

I'd have suggested taking the intel approach, which write a single function for
this function to convert ARGB4444 to ARGB.  A bit of calling overhead but much
less code.
On Neon the vector load instructions made it easy to write direct loaders for
many formats, but on Intel and Mips a lot of code is needed to read different
RGB formats.

[manojkumar.bhosale, 2016/10/19 11:56:27]

Done.

   void (*ARGB4444ToUVRow)(const uint8* src_argb4444, int src_stride_argb4444,
       uint8* dst_u, uint8* dst_v, int width) = ARGB4444ToUVRow_C;
   void (*ARGB4444ToYRow)(const uint8* src_argb4444, uint8* dst_y, int width) =
@@ -1265,6 +1265,17 @@ int ARGB4444ToI420(const uint8* src_argb4444, int src_stride_argb4444,
       }
     }
   }
+#elif defined(HAS_ARGB4444TOYROW_MSA)
+  if (TestCpuFlag(kCpuHasMSA)) {
+    ARGB4444ToUVRow = ARGB4444ToUVRow_Any_MSA;
+    ARGB4444ToYRow = ARGB4444ToYRow_Any_MSA;
+    if (IS_ALIGNED(width, 16)) {
+      ARGB4444ToYRow = ARGB4444ToYRow_MSA;
+      if (IS_ALIGNED(width, 32)) {
+        ARGB4444ToUVRow = ARGB4444ToUVRow_MSA;
+      }
+    }
+  }
 // Other platforms do intermediate conversion from ARGB4444 to ARGB.
 #else
 #if defined(HAS_ARGB4444TOARGBROW_SSE2)
@@ -1310,7 +1321,7 @@ int ARGB4444ToI420(const uint8* src_argb4444, int src_stride_argb4444,
 #endif
 
     for (y = 0; y < height - 1; y += 2) {
-#if defined(HAS_ARGB4444TOYROW_NEON)
+#if (defined(HAS_ARGB4444TOYROW_NEON) || defined(HAS_ARGB4444TOYROW_MSA))
       ARGB4444ToUVRow(src_argb4444, src_stride_argb4444, dst_u, dst_v, width);
       ARGB4444ToYRow(src_argb4444, dst_y, width);
       ARGB4444ToYRow(src_argb4444 + src_stride_argb4444, dst_y + dst_stride_y,
@@ -1329,7 +1340,7 @@ int ARGB4444ToI420(const uint8* src_argb4444, int src_stride_argb4444,
       dst_v += dst_stride_v;
     }
     if (height & 1) {
-#if defined(HAS_ARGB4444TOYROW_NEON)
+#if (defined(HAS_ARGB4444TOYROW_NEON) || defined(HAS_ARGB4444TOYROW_MSA))
       ARGB4444ToUVRow(src_argb4444, 0, dst_u, dst_v, width);
       ARGB4444ToYRow(src_argb4444, dst_y, width);
 #else
@@ -1338,7 +1349,7 @@ int ARGB4444ToI420(const uint8* src_argb4444, int src_stride_argb4444,
       ARGBToYRow(row, dst_y, width);
 #endif
     }
-#if !defined(HAS_ARGB4444TOYROW_NEON)
+#if !(defined(HAS_ARGB4444TOYROW_NEON) || defined(HAS_ARGB4444TOYROW_MSA))
     free_aligned_buffer_64(row);
   }
 #endif