Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: MTD(f) (was Re: New(?) search idea.)

Author: Rickard Björklund

Date: 08:00:09 01/26/98

Go up one level in this thread


On January 26, 1998 at 09:28:05, Robert Hyatt wrote:

>this has already been done.  Baudet tried this approach at least 20
>years
>ago.  It was found to limit the speedup to around 5X, which is far worse
>than splitting the tree itself into smaller slices that can be searched
>in parallel.  In fact, if your initial "guess" of the score is dead on,
>searching different "windows" in parallel would get you *no* speedup at
>all...
>

I'm only considering of using parallel window split for the first few
plies in the iterative deepening, before I switch to YBW. This is
because
of the communication overhead I have (using a couple of Sun Ultras over
a slow network).

I think it's possible for MTD(f) to benefit more from the parallel
window
split, than normal alphabeta, since the results are used to chose new
bounds. I'm no quite sure of how the hash table would be designed to
interact
properly with parallel windowsplit using MTD(f).

Rickard



This page took 0.02 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.