SML numerical benchmark and MLton

Stephen Weeks sweeks@wasabi.epr.com
Sat, 17 Jul 1999 00:30:58 -0700 (PDT)


I was wondering if it would be ok with you if I included some portion
of your SML benchmark (http://www.arrakis.es/~worm/smlapl.tgz) in the
next round of performance numbers that I compute comparing MLton to
SML/NJ (http://www.neci.nj.nec.com/PLS/MLton/performance.html).  I
just ran your benchmark and see that MLton is almost 40 times faster
than SML/NJ for the +* test.  This is far and away the best (from
MLton's point of view anyway) ratio that we have seen for a benchmark,
excepting those using IntInf.

I'd be interested in any other SML benchmarks that you have and if you
did any comparisons of MLton to C and how MLton fared on those.

Thanks.

I'd also like to mention the new release of MLton is available (a bit
later than promised) which should fix the RedHat 6.0 problems that
you pointed out.