Re: saner shootout programs

On Sun, May 11, 2008 at 1:40 PM, Don Stewart <dons <at> galois.com> wrote:

> n-body requires updating a global array of double values to be

I think the array and any side-effects on it can and should be local
to the simulation procedure.

> competitive performance-wise, though we haven't really nailed this
> benchmark yet. The current entry should be considered an older approach
> to raw performance -- typically we can get good (or better) results in
> using the ST monad.

As I understand, a function can use ST, but still be pure. If so,
that's the version I'd like to time (assuming it's also type-safe).
The only pure functional solution on that page that I found claims to
have a huge leak.