[computer-go] Language War!

[Mark Boon]

A rewrite can often be a good idea, as long as you can afford the
resources. After a while you end up in a situation you want to rewrite
before your previous rewrite is complete. You always learn as you go.
The important thing is to identify the right modules and a good way to
communicate between them. Then you can start rewriting modules, which
is much cheaper.

I don't know what specific performance problems you're having in Java
that you think will be resolved in C++.There are a few areas where C++
obviously performs much better, but only few. And fundamentally your
program can look very similar. Still, coming from Java I predict
moving to C++ will prove to be very frustrating. But it's your call.

I'm a performance freak as bad as almost anyone else, but I doubt I'll
ever be going back to C++.

Mark


On 7/29/06, Peter Drake <drake at lclark.edu> wrote:
> I know where this thread is going, so I thought I'd get it out of the
> way and put it in the Subject: line.
>
> Here's the situation. Orego is now a relatively strong program,
> rapidly climbing the 9x9 ladder. Though knowledge-free, it can
> sometimes beat GNU Go. Our current algorithm uses a Monte Carlo
> approach similar to Coulom's "Crazy Stone".
>
> While we have a bunch of improvements to consider, the algorithm is
> so computationally intensive that the inefficiencies of Java are
> hurting us in both time and space. We're considering switching to C+
> +. None of us has extensive C++ experience, although we've all seen
> some C and one of us has done a lot of Objective C. (I've ordered a
> copy of Weiss' "C++ for Java Programmers".)
>
> Should we start from scratch or use parts of another open source
> program, such as GNU Go?
>
> Peter Drake
> Assistant Professor of Computer Science
> Lewis & Clark College
> http://www.lclark.edu/~drake/
>
>
> _______________________________________________
> computer-go mailing list
> computer-go at computer-go.org
> http://www.computer-go.org/mailman/listinfo/computer-go/
>