Author: Ulrich Tuerke
Date: 02:34:44 02/25/99
Go up one level in this thread
On February 24, 1999 at 15:04:45, Will Singleton wrote: >On February 24, 1999 at 04:48:50, Ulrich Tuerke wrote: > >> >>On February 24, 1999 at 04:36:47, Ulrich Tuerke wrote: >> >>I use the 2 positions below to test correct work of my hash algos. You need >>correctly working hash to solve this. May be they are useful for you. >> >>8/k7/3p4/p2P1p2/P2P1P2/8/8/K7 w - - 0 1 id "HASH-01";c0 "Only Kb1! wins"; >>8/1p5k/1P1p4/3p4/3Pp2p/2K1P2p/7P/8 w - - 0 1 id "HASH-02";c0 "Kb2! draws"; > > >Thanks for posting these, they showed a problem with my tables. When I run >these from the initial positions, I get the solutions almost instantly, as it >goes quickly up to the maxply setting. But the search bogs down on subsequent >moves, and takes much longer. So I went ahead and cleared the tables after each >move, and now each move is like the first, almost instant! > >So I must have a problem storing current positions over existing table entries. >Or does it sound lile something else? Sounds like you are right. Comet stores the move number of the search into the hash table entry. This way entries from an older search can be recognized. These will be used in hash table probes but they will be generally overwritten by Store-operations of the current search. So older entries will quite naturally "age out" and not block the table. As far as I can say, this works well. So you don't have to clear the table between moves. Uli > >Will
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.