just wondering how much this has been optimized for the OS X and the G5 (and other ppc chips to I guess) and if it user apple's accelerate framework at all. would using it help? or is what you guys have done faster than other methods/wouldn't be affected by accelerate? Anyway, great app.
FFFF does not use Apple's Accelerate Framework.
Since Acclereate is a generic vector-math abstraction layer over native SSE*/AltiVec, I guess it should be slower (FFFF uses native assembly code).
Anyway it would be nice to test AF to get an idea of its performance.
P.S. Note that FFFF code-paths for the G5 (both FPU and AltiVec) are not yet optimized.