Re: [computer-go] Computer Go tournament at EGF

[Erik van der Werf <E.vanderWerf@xxxxxxxxxxxxx>]

Arend Bayer wrote:

> On Fri, 11 Feb 2005, Erik van der Werf wrote:
>
>  
> > David G Doshay wrote:
> >
> >    
> > > SlugGo plays to a specified search depth. Because this is a serial
> > > search I can think of no way to speed this up beyond what we have
> > > already done with hash tables, other than to make search depth
> > > a time dependent variable.
> > >      
> > You could try to implement something like iterative deepening.
> >    
> I don't understand what you mean by iterative deepening in the context
> of SlugGo's search. Remember it only branches at ply=1, and the number
> of branches is smaller than the number of available CPU's in their
> cluster. So the search time is mostly
> 	depth*(time per move of the slowest GNU Go instance)
>  
Well, I wrote "something like" because I don't know exactly what would 
be the best solution for SluGo. The thing I had in mind is that you 
could start with a small depth and gradually increase the depth until 
time runs out. In addition it might be useful to gradually increase the 
level of GNU Go (assuming it is faster at lower levels). Since you have 
more processors available you might even be able to do several 
'iterations' in parallel.

E.


_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/