- GPU: Unify the GPUEngineBase::_PixelEffect*() functions.
- GPU: For native-sized captures, always read from native VRAM -- never from custom VRAM. Should make captures a little faster when running a custom framebuffer and/or using RGB888 color format.
- GPU: Fix bug where sprite blending would no longer work. (Regression from commit 624dd57.)
- GPU: Another significant performance improvement to GPU compositing in many games.
- Now composites pixels using explicit functions for simple copy,
masked copy, and masked effect.
- On SSE2 systems, pixels composited using a simple copy no longer
require the destination pixels to be loaded first, since all the pixels
are guaranteed to be overwritten anyways.
- Try and move the window test as far up the pipeline as possible so
that pixel rendering can bail as soon as possible if the window test
- Clean up and further standardize the code for compositing BG layers,
OBJ layers, and 3D layers.