[Shootout-list] fannkuch (timer resolution; HZ=1000?)
Jon Harrop
jon@ffconsultancy.com
Fri, 20 May 2005 18:47:01 +0100
On Friday 20 May 2005 18:26, Greg Buchholz wrote:
> --- Christophe TROESTLER <del-con@tiscali.be> wrote:
> > The point is that the variation is so big w.r.t. the running time
> > (25%) that it shows the program should run longer to avoid random
> > effects to take such an importance in the result.
>
> Quite a few people have been discussing the 0.01 second timing
> resolution that the shootout uses. This sounds suspiciously like it
> might be limited by Linux's timer resolution...
Two bigger problems are that most modern language will add much bigger random
noise and a bias. OCaml running times often vary by 0.1s and I've heard (from
Microsoft) that C# does poorly in the shootout because it has a big startup
time.
I thought it would be nice to do those Skaller plots using only tests which
ran for >0.1s or even >1s but there is virtually no such data on the shootout
(i.e. the plots are all blank).
--
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
Objective CAML for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists