Author: Brian Richardson
Date: 15:34:24 05/08/02
Go up one level in this thread
On May 07, 2002 at 22:36:40, Will Singleton wrote: >On May 07, 2002 at 22:04:05, Brian Richardson wrote: > >>On May 07, 2002 at 13:34:34, Robert Hyatt wrote: >>>Nothing wrong with starting simple. Always store entries in the normal search, >>>don't in the q-search. It will be easy to add it to the q-search later if you >>>think you want to try it... >>> >>>A single always-replace table will not be anywhere near as good as the two-table >>>approach most of us use (one is always replace, the other is replace with deeper >>>draft only). >> >>I have tested this several times with Tinker. In nearly all cases a single >>table (replacing when id is different, or based on depth when the same) does >>better than the 2 table approach. Some of the tests in ICCA are based on much >>older systems with smaller main memory sizes. I think with modern memory sizes >>and hash tables much greater than 1 million entries, things change. Tinker >>usually runs with 32M entries. >> >>In any case, you may want to try both 1 table and 2 in your program, since each >>search is somewhat different. For example, hashing q-search entries is also >>faster for Tinker, but apparently not for Crafty. > >Have you experimented with your single table approach on Fine70? Vs the dual >table approach? On that position, and others, 2 tables seems better. > >Will Yes, either way works with Fine70...only the single hash table is about 5% faster, at least for Tinker. Brian
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.