[Computer-go] effectiveness of transposition tables for go

Petr Baudis pasky at ucw.cz
Tue May 11 02:09:13 PDT 2010


On Mon, May 10, 2010 at 04:38:31PM -1000, Mark Boon wrote:
> On Mon, May 10, 2010 at 3:30 PM, Petr Baudis <pasky at ucw.cz> wrote:
> > It seems memory issues start to pop up when you try to support very long
> > thinking times (minutes per move) or massive influx of simulations
> > (cluster computation). Pachi was originally extremely wasteful about
> > memory (malloc()ating each single node separately!) and it never really
> > mattered (unless you wanted to play very slow on 19x19) until Jean-loup
> > had to fix that in order to handle many-core machines sensibly - tens
> > of cores simulating in parallel will grow the tree very quickly. ;-)
> 
> OK. Then I'm curious, was the solution something along the lines what
> Don described?

I admit I got a bit lost in who described what. ;-)

Our solution is to garbage-collect the tree in that case, pruning
useless branches. And we actually don't even implement doing that
throughout single move reading, we just stop the search when we hit
the memory and while the move is played, the tree is pruned on the
background.

-- 
		Petr "Pasky, but Jean-loup wrote all that" Baudis
When I feel like exercising, I just lie down until the feeling
goes away.  -- xed_over



More information about the Computer-go mailing list