Author: Dann Corbit
Date: 13:02:47 01/12/05
Go up one level in this thread
On January 12, 2005 at 14:55:12, Norm Pollock wrote: >Can someone explain (1) how does the engine go about probing the egtb? You just set up a board in a struct that the Nalimov EGTB likes and call the probing interface. Deiter showed a nice example in the Winboard forum a few days ago. >(2) what does the egtb cache do? There are two functions for EGTB memory: 1. When you load a compressed Nalimov tablebase file, it is decompressed from disk into RAM. Also, if you probe the same file over and over, it will try to hold it in memory so you don't have to hit the disk again. >As I understand it, with the 3-4 man egtb files and an egtb cache of 32M, all >the egtb files can be stored and diskswapping of them is eliminated. For the 5 >man egtb files, storing them all in ram is not feasible at this time (7G). What is feasible now might surprise some... One gig of Ram for a fast computer is about $250: http://www.memorygiant.com/index.cfm?fuseaction=search.memorySearch&manufacturer_id=87&model_id=38000%26model%3DTransport+FX27+%28B4520T4SR%29 So you can get 7 gigs for about $1750. So with $2500 in Ram, you could comfortably memory map the 5 man files and have Ram left over for a nice hash table, etc. The biggest problem with large Ram systems is that most motherboards do not support more than 4 gigs per processor now, even though the 64 bit CPUs can address terrabytes.
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.