I've been working recently on improving the video performance on a Samsung 2413 processor.
The platform I'm working on does not have hardware video acceleration or a hardware cursor.
Did you know that if you disable the software cursor in the driver and remove the overlap detection for emulated line and blit you can get a two fold speed increase at least with drawing polygons.
The hardware supports DMA and I've started to look at using this to accelerate line fills and various blits. Unfortunately the hardware does not support block DMA which cuts down how it could help with majority of blit operations. Another snag with this approach is the overhead it takes to setup the DMA controller and cope with odd end of line conditions.
