[Remoting Android] Cursor & Cursor Feedback for OpenGL Renderer

remoting/base/util.cc

Index: remoting/base/util.cc
diff --git a/remoting/base/util.cc b/remoting/base/util.cc
index c72f3a8105fd0bd0a0a64041490f2dda23098b20..f6adcb4b9be0b71973c510843badb79c4cb36ebf 100644
--- a/remoting/base/util.cc
+++ b/remoting/base/util.cc
@@ -72,6 +72,15 @@ void CopyRGB32Rect(const uint8_t* source_buffer,
   }
 }
 
+void CopyAndSwapRB(const uint8_t* source, uint8_t* dest, int length) {

[Sergey Ulanov, 2016/07/19 00:42:48]

Please use RGBAToARGB() from libyuv. It's optimized with NEON, so it should be
significantly faster. It doesn't matter for cursor, but someone will want to use
this function for something bigger than mouse cursor.

[Yuwei, 2016/07/19 04:11:52]

Wouldn't it be BGRAToRGBA()? Looks like there is no such function but
ABGRToARGB() seems to be doing the right thing... I guess that's in little
endian?

We are doing row-by-row memcpy in GlDesktop when we need to manually pack the
image. Is there any similar optimization for that?

We will need to list libyuv as the dependency of the renderer. Not sure whether
that's a good idea...

[Yuwei, 2016/07/19 20:34:23]

Done. Using libyuv::ABGRToARGB

[Sergey Ulanov, 2016/07/20 18:06:11]

memcpy() itself should be already optimized. 
Client already depends on libyuv, so this is not a problem

+  for (int i = 0; i < length / kBytesPerPixelRGB32; i++) {
+    dest[i * kBytesPerPixelRGB32] = source[i * kBytesPerPixelRGB32 + 2];
+    dest[i * kBytesPerPixelRGB32 + 1] = source[i * kBytesPerPixelRGB32 + 1];
+    dest[i * kBytesPerPixelRGB32 + 2] = source[i * kBytesPerPixelRGB32 + 0];
+    dest[i * kBytesPerPixelRGB32 + 3] = source[i * kBytesPerPixelRGB32 + 3];
+  }
+}
+
 std::string ReplaceLfByCrLf(const std::string& in) {
   std::string out;
   out.resize(2 * in.size());