Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: best performance from rehashing schemes?

Author: Gerd Isenberg

Date: 06:16:28 12/13/01

Go up one level in this thread


I made good experience with a replacement scheme that probes up to N (eight
works fine) consecutive entries. The entry with the lowest Priority is replaced.
Entry-Priority is based on Timestamp, Draft and (Bound-)Flags (PV, EXACT, LOWER,
UPPER). I use the PV-Flag, to indicate that entries are Part of a Principle
Variation of any Root successor. This entries may rather important - therefore
this PV-Flag has higher priority than EXACT scores. After deepening (domove,
search, undomove) any root move i tag all entries of the backupped PV or
Refuting-Line as PV, so these entries became more resistant.

Replacing of already stored nodes is also done by a similar priority scheme,
with the option of merging some properties, like the PV-Flag and a possible
second best move.

I do not store or probe this transposition table during quiescence search except
the initial search depth of the root is less than four, or i am in the left most
variant of any root move.

Gerd



This page took 0.01 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.