Author: José Carlos
Date: 06:44:41 12/30/01
Go up one level in this thread
On December 30, 2001 at 08:14:14, Gian-Carlo Pascutto wrote: >Hi all, > >I'm wondering about the following: > >before searching, right now, I always clear my >depth-priority hashtable (I have 2 tables, depth-prior >and always-replace). However, that takes some time, >which gets significant with huge hashtables, and I >will be getting rid of it via aging. > >My problem is, is it 'ok' to have old hash entries >lying around in the depth priority table? > >When doing this, and the move of the opponent is what >was in my PV, the search immediately will zoom up to >the depth it had reached. This is nice, but I am wondering >about the implications for iterative deepening. It seems >to me that this could actually be slower because it now >has to start a very deep search with nearly no move >ordering info. Actually it doesn't _allways_ work like that. Most of the time, eventhough you have the pv move with depth==12 in the hash table, you don't have information enough to quickly discard the rest of the moves, so the program does some search before getting to ply 12. If the move is clearly the best, then you'll get to ply 12 very quick, but if the position is complex, the program will spend some time trying to fail high for other moves it doesn't have information enough. I've seen this many times because this is the way I ponder. When I receive the opponent move, I undo my guessed move, make the opponent's and start thinking. And about aging hash entries, I do something 'original'. I save the total material with the hash entry. As long as the material in there is smaller or equal than the root position, I think the hashed position is interesting, because it is still possible to reach it. When it is bigger, it is immediately replaced. José C. >Has anyone tested this? What are you doing? > >-- >GCP
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.