C64 emulation is still choppy even on 1ghz turbo mode. I think this is down to the code needing more optimisation for the specific arm type used on the pi.
The problem right now with almost every piece of software in the Pi is libSDL. If you take any emulator or native game and use gdb with it, you'll see A LOT of time in erch running cycle (let's say from vsync to vsync) is spent in the SDL software blitting functions. In an X86 machine, libSDL provides asm code for mmx blitting, etc.
There's also the scaling problem: any game/emulator on SDL being stretched from low-res modes is eating almost every remaining CPU cycle per frame in software scaling.
That's lame.
I implemented a dispmanx backend for SDL (
https://github.com/vanfanel/SDL12-kms-dispmanx) wich solves the scaling problem, but the blitting problem remains. GLES could be used somehow to do it, but I have yet to understand how
