Subject: Re: Math Library
Date: Sat, 28 May 1994 18:25:34 +1000 (EST)
From: MiKE <cheng@sun.mech.uq.oz.au>

On Fri, 27 May 1994, MR. R.L.COLLIER wrote:
> can one of you guys be more specific on where in the amninet this library is a
s I would very much like to get it!!
 aminet: util/libs (I think)

For those that care, I ran the benchmarks for the libraries on my a1200
(4Mb Fast,'881 co-pro, vanilla '020)
I'll just list the total times (in seconds) , as it does about 15 different
functions 
The old versions of my libraris were 37.1 as supplied with WB3.0
The new libraries are v40.3

					before		after
mathdoubtranstest			208.38s		184.87s	
(tests mathsieeedoubtrans.library)     

mathsingtrantest			142.08		120.49

mathtranstest				591.97		143.82


The speedups in the mathdoubtranstest and the mathsingtest were almost
toatlly due to speed up in the 'pow' function. (which I asume is powers).
					before		after
	ie mathdoubtrans     DPPow	44.65		24.02
	   mathsingtrans     SPPow	27.49		5.61

Would anyone with knowledge of how these functions relate to ray-tracing
algorithms, please tell me whether I can expect any decrease in rendering
times with these improvements.  I suspect that only the doubtrans library
is important and that the Pow function is not heavily used with ray
tracing;  so overall I don't expect better Imagine render times.

(I only have the INT version from AF, so I can't test render times)

MiKE
   cheng@sun.mech.uq.oz.au  The Uni of Queensland. Brisbane. Australia   
            "I don't stand on ceremony; he never stood on me." 




                           -=> RETURN TO CONTENTS!<=-