Author: Robert Hyatt
Date: 21:04:34 10/23/01
Go up one level in this thread
On October 23, 2001 at 19:21:30, Ratko V Tomic wrote: >> You have to read in some "blocksize" anyway. Our blocksize >> is 8K bytes. But that is one I/O to get the byte I want. >> I don't have to read in 99 _more_ blocks to track down to >> the mate-in-N score... > >You need to read and decompress thousands of distance-to-mate values >to extract the one you're looking for? If so, it is not quite as >quick as one lookup, like an array lookup via index, per leaf node, >as you were describing earlier. A leaf node could cost you 10-20 ms, >which is plenty of time for a best-move method move generator to >run through some rule based sequence, with occasional, if any, >correction from a much smaller exception table covering a >class of endgame of interest. Yes, but it _isn't_ "plenty of time" to read in 100+ blocks to find that the best move leads to a mate in 50+... That is my concern... your approach will _still_ have to decompress on the fly, just like Eugene's does today. So that is a wash. But we do exactly one I/O operation to get that same mate in N score.
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.