Re: [gnugo-devel] SlugGo v.s. Many Faces

[David G Doshay]

On Aug 25, 2004, at 4:16 AM, max-d wrote:

> ----- Original Message -----
> From: "David G Doshay" <address@hidden>
> To: "GNU Go development" <address@hidden>
> Sent: Wednesday, August 25, 2004 1:30 AM
> Subject: Re: [gnugo-devel] SlugGo v.s. Many Faces
>
>
> > On Aug 24, 2004, at 4:05 PM, Marco Scheurer wrote:
> >
> > > You should contact max_d, who often post here results of tournament 
> > > he
> > > plays between several engines. Apparently Go++ is now in the pack. I
> > > think these tournaments are automated, probably using screen
> > > scrapping. Maybe a GTP interface could be wrapped around his tools?
> >
> > We thought about screen scraping and decided to wait. If anybody is
> > willing to share tools I'd be thrilled. We now have a set of C# 
> > modules
> > for GTP ethernet communication with MFG. We use mono for running the 
> > C#
> > on the Mac cluster.
>
> First , congrats for your work!

Thanks, it has been FUN.

> 2 , if your SlugGo program can run under Windows ,it should work 
> against
> Go++ using autogo .

No, SlugGo runs on a cluster of OS X machines, although it will run 
fine on
any cluster that can use LAM-MPI ... a whole raft of unix and linux 
type boxes.
Our infrastructure is a layer on top of LAM-MPI. LAM-MPI is not Windows
compatible.

While it can run on a single cpu machine it would be painfully slow for 
any
kind of game play. We have run single cpu for debugging, but only for 
very
small boards and short lookahead.

> Right now for unlimited matches, autogo need a tool named AutoIt , but 
> for a
> great project like yours i can rewrite its Go++ interface  to make it
> autoworking.

I would love to be able to automate games against Go++. If you have or 
are
willing to help write the tools to make that happen then I am ready to 
move
forward.

Perhaps we should take the rest of this thread off of gnugo-devl, unless
others are still interested.

Thanks much,
David