Author: Don Dailey
Date: 11:37:34 01/19/98
Go up one level in this thread
Stuart, I have had no success with rehashing. I save and store positions well into the quies search unlike most of the program here though so it might change things. Once your table fills up you will have to overwrite something and will be doing some work to find the best candidate. From my discussions with others, most do not rehash either. I would still give it a try though. One idea is set associativity. I'm pretty sure Cray Blitz used this but not positive. Bob should have some good ideas for you. One idea I want to research is the concept of "key positions." Certain hash entries may be overwhelmingly important compared to others and perhaps should NEVER be overwritten. Like key positions in the Ruben Fine 70 ending. I don't know if it's possible to identify these but it could be a minor breakthrough if we could. Nahhh... - Don On January 19, 1998 at 14:01:30, Stuart Cracraft wrote: >I plan to implement simple linear rehashing. > >Right now, my hash table is not very good and doesn't give >good numbers. I'd like to fix that first but haven't found >the fix. My numbers are typically under 10% (e.g. total >probes successful / total probes tried) for the regular >transposition table (pawn hash table is always very high and >seems okay.) > >Anyway, I'm hoping that this <=10% number for the main transposition >table will take a jump up after rehashing. > >If anyone has implemented rehashing and tested after implementing >it, I'd be interested to know what you saw. And I'll report back too. > >--Stuart
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.