Author: Dann Corbit
Date: 10:54:23 05/26/05
Go up one level in this thread
On May 26, 2005 at 13:43:31, Dieter Buerssner wrote: >On May 25, 2005 at 17:07:49, Dann Corbit wrote: > >>Sometimes, I do not want the engine to clear the hash. >> >>For instance, I might be analyzing a game move by move in the form of EPD >>records. My analysis will be almost twice as good for the same time interval if >>I do not clear the hash > >I agree Dann. But in this discussion we referred to (unrelated) test positions >only, where clearing HTs should never hurt. In some cases, it could have (small) >advantages - namely to have reproducable node numbers in the output (for many >non SMP engines at least). Very frequently, I take a game, and feed it to pgn2epd or some such utility to create EPD records. Then I analyze these with a chess engine. That is the case I was referring to. So, to clear the hash is good if the position is not in the table. To cleart the hash is bad if the position is in the table. So, I think the engine should decide. In fact, if I get a command to clear the hash, I ignore it. If I see that a new position to analyze is in the hash table already, then I will not clear it. But if the position does not live in the hash table, I will clear it. Requests to clear the hash table are the sort of thing that naughty interfaces might use to degrade performance.
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.