[plt-scheme] V301.5 Speed Up

From: Stephen Bloch (sbloch at adelphi.edu)
Date: Fri Feb 10 14:52:26 EST 2006

At 9:17 AM -0800 2/10/06, Noel Welsh wrote:
>   - unexpectedly, the variance of my measurements was
>crazy!  When I made benchmarks (just loops adding up
>numbers) long enough to measure the time reliable I got
>results like this:
>...
>p 1.0
>s1 mean: 6.799999999999996 var: 22.204081632653068
>s2 mean: 13492.2 var: 18376.693877551028

Somebody pointed out that time-apply sometimes returns a negative 
number, which would obviously screw things up.  That sounds like a 
bug in time-apply.  However, there's a more general problem: when 
I've run benchmarks like this (in various languages, on various 
compilers and OSes), I've frequently found that the FIRST execution 
of the function to be tested takes much longer than subsequent ones, 
possibly due to virtual-memory and/or caching issues.  So you might 
simply exclude the first run from the statistics and see whether the 
variance improves.

-- 
							Stephen Bloch
							Math/CS Dept
 
Adelphi University
 
sbloch at adelphi.edu


Posted on the users mailing list.