[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

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

On Aug 25, 2004, at 4:56 AM, Don Dailey wrote:




The primary reason  the code is so  very slow is that we  have found a
lookahead depth  of 16 works very well,  and it thus takes  a bit more
than 16 times as long as GNU Go to settle on its move.



Can you  explain this statement?  You  make it sound  that your search
takes linear time.  Would a 32 ply search take only twice as long?




At one limit the search time is linear in the plys because there is no 
branching

after the first set of candidate moves are selected. This is how we 
were running

in the reported games against Many Faces.



At the other limit it is still linear in the ply depth as long as there 
are cpus

available to supply the desired branching rate. We can do that too. We

have 72 cpus in our pool and intelligent use of them should result in 
stronger

play.



It is interesting that 16 moves of once-branched play is worth something

like 3 stones. We will continue to gather statistics to find out.



The infrastructure we are developing is designed to make it easy to 
program

simulations which make decisions at run time resulting in more cpus 
being

put to work. Go is our testbed application.



Thanks again to the GNU Go folks for supplying us with a code we barely 
had

to touch to do this.





Cheers,

David



_______________________________________________

computer-go mailing list

computer-go@xxxxxxxxxxxxxxxxx

http://www.computer-go.org/mailman/listinfo/computer-go/