Coordinated Graphics: Decode an image directly into a GraphicsSurface.

[Dongseong Hwang]

We need to decode an image directly into a GraphicsSurface when the image is directly composited image that RenderLayerBacking give CoordinatedGraphicsLayer via setContentsToImage.
If we make it, we can save temporary memory and remove a copy overhead.

[Noam Rosenthal]

Created attachment 173925 [details]
Partially working patch

This shows how it can work, but I'm not sure yet if it's the way to go.

[Dongseong Hwang]

I think you show the direction where we should go.

I think it would be better if we can change only ImageDecoder without changing BitmapImage and ImageSource.
I think we can do it as following chromium's approach. 
for instance, define as follow in NativeImagePtr.h

#elif PLATFORM(QT)
#if USE(GRAPHICS_SURFACE)
typedef GraphicsSurface* NativeImagePtr;
#else
typedef QPixmap* NativeImagePtr;
#endif

and every image decoders access NativeImagePtr's pixel.

btw, I think the most tricky blocker is caching mechanism of MemoryCache.
MemoryCache needs to remove decoded data of CachedImage. But if MemoryCache removes directed-composited image, UI Process encounters null pointer exception. we need to make the plan for this.