[Shootout-list] small change (that i can do) that would benefit 3 people
Jon Harrop
jon@ffconsultancy.com
Sat, 7 May 2005 19:55:22 +0100
On Saturday 07 May 2005 18:02, Will M. Farr wrote:
> What about tests which run in order n^2 time? Or order n log(n), or
> tests which don't run long enough to be in the asymptotic range and
> therefore run order a*n + b*log(n) + c*n^2 or something? Then the
> time/n will not be a good measure of the speed because it will not be a
> constant across different n. I haven't looked, but surely not all the
> tests run linearly with n.
Asymptotically many of the tests may be O(n) but the time taken will rarely be
linear in n.
I was thinking along similar lines recently though. Why not have at least some
tests which try to compute as much as possible in a given time (e.g. 3s)? You
may lose the ability to do a linear measurement but you still get a ranking
and you can have more language compete as the very slow ones just don't get a
lot done in 3s.
--
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
Objective CAML for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists