[MLton-devel] Porting from Moscow ML

Matthew Fluet fluet@CS.Cornell.EDU
Fri, 27 Sep 2002 12:46:41 -0400 (EDT)

> Re benchmarks, things like count-graphs and smith-normal-form can be tuned
> to take as long or short as desired.  This isn't a matter of chaning a repeat
> count, but the problem that count-graphs is doing is way exponential so you
> just set the size correctly.  Similarly smith-normal-form is dealing with
> a matrix reduction and we can make the matrix as large or small as we want.

O.k.  Then I guess I would simply suggest (and will do so in the not too
distant future), that we go through all the benchmarks on a modern machine
with the latest MLton and ensure that everything is running between the 1
and 2 minute mark, and adjust "problem sizes" rather than iterations on
the ones for which that makes sense.

This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
MLton-devel mailing list