On Tue, 17 Nov 1998, Maurice Marks wrote:
> My results so far are 2X realtime (twice as fast as realtime)
> on a 600MHz 21164 system and 3X realtime on a 600MHz 21264.
really?
it's about 2.75X on a P166 (freebsd).
> The critical loop on Alpha at the moment is in quantize().
Same for everyone I fear. If you want to see how much better quantize is,
compile the original ISO source and do a profile.
> I moved a multiply (by 1/step) out of the loop by premultiplying
> the constant boundary tests by stepsize. This gave me a few % improvement.
oops. I thought i'd done that.
later
mike
(my last exam is in 7 hours!!!)
