Author: Dann Corbit
Date: 14:58:10 11/21/03
Go up one level in this thread
On November 21, 2003 at 17:49:53, K. Burcham wrote: [snip] >for some reason it seems that Johan thinks that CM strength is strong even >though we have said that the program writes over the top of the 32 or 64 megs >that some will set for CM. > >It seems I am missing something here in having a better understanding of this. If you have 64 MB allocated for hash, and a hash slot uses 32 bytes (for example) then that hash table would hold two million precalculated entries. Most of the time, that will be plenty. It is the old entries that will get over-written most of the time. When I look into the hash table, if there has been a recent transposition, I will find it. The only way that you will run into problems is if your hash table is so small that recent entries are frequently over-written. From a visual standpoint, imagine this: If I move the knight first, and then he captures with a knight and I retake with a bishop then he captures with a pawn - we will calculate that and store it in the table. Then, if I take with the bishop, he captures with a knight I retake with a knight and he captures with a pawn. That position is the same now. So when we look in our table, we will find the calculations already performed. After we have stuffed a few million of these recent transpositions into a table, we won't miss very often. And writing over top is probably fine, since a new calculation is likely to be needed again soon. And if I should happen to write over a fresh calculation, it is rare enough that it won't matter much. I'll just have to recalculate that position from scratch. Not a great loss, really.
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.