Author: Robert Hyatt
Date: 10:59:57 05/03/04
Go up one level in this thread
On May 03, 2004 at 13:51:18, Uri Blass wrote: >On May 03, 2004 at 11:37:44, Robert Hyatt wrote: > >>On May 03, 2004 at 11:13:11, Vincent Diepeveen wrote: >> >>>On May 03, 2004 at 10:53:19, Robert Hyatt wrote: >>> >>>>On May 03, 2004 at 10:19:46, Sune Fischer wrote: >>>> >>>>> >>>>>>I don't see any at 8. I don't personally have access to a 16-way box yet so I >>>>>>can't say anything there. But there is nothing that really makes null-move hurt >>>>>>parallel search... >>>>> >>>>>Couldn't it be, that nullmove hurts scalability in much the same way alpha-beta >>>>>"hurts" parallel search compared to minimax? >>>> >>>>I don't see how. It might make _some_ positions more unstable, and unstable >>>>positions hurt parallel search. But it also makes other positions more stable. >>> >>>What a nonsense. "i don't see how." >>> >>>it is very clearly proven. For everyone doing parallel research it is *trivial* >>>that more unstable trees are harder to split. Note that in old issues of ICGA it >>>is already mentionned for example by Jonathan Schaeffer. >> >>So? Do you claim that null-move makes the search more "unstable"? > >I understand nothing about parallel search so I will say nothing about it >but I know that null move is one of the factors that can cause wrong fail high >when after research you get fail low. > >If this is not unstability of the search then I do not know what do you mean by >unstability. > >Uri In this context it is a measure of move ordering. The better the move ordering the better the parallel search performs. One part of my dissertation dealt with searching _perfectly_ ordered trees, and the results show that it is possible to get a perfectly linear speedup on such trees. Another part of my dissertation showed the (trivial) case where worst-ordering can also be searched with linear speedup (this is pure minimax basically). It is the actual game-trees we see in chess that are a problem as they are not worst or best ordering, and that causes problems. So in this context, we are talking about how null-move affects basic move ordering. It does affect it somewhat, but it doesn't affect it nearly as much as Vincent would like to claim to make his parallel search look better. Your example is but one way this shows up. Clearly when a move fails high, and then comes back and fails low, move ordering on the fail-low search was bad because it was based on the fail-high tree previously searched...
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.