Author: Robert Hyatt
Date: 19:31:22 03/12/03
Go up one level in this thread
On March 12, 2003 at 15:32:26, Tom Kerrigan wrote: >On March 12, 2003 at 13:44:46, Bernardo Wesler wrote: > >>Although I am not an expert in computing, I know that the larger amount of hash >>tables you set, the performance of the processor goes down. > >Not really. As long as you have hash tables that are much bigger than your >processor's cache (and who doesn't?) you are doing a main memory access per >node, and that access has fixed latency, regardless of the table size. Note that this is not quite complete. The MMU does two additional memory accesses when the requested virtual page number is not currently in the TLB. But this is probably true for nearly any hash entry you hit on, since 256 TLB entries only addresses 1M of memory... > >In other words, your NPS should be the same if you switch from, e.g., 32MB >tables to 64MB. If there is a variation, it's probably because the search tree >is somewhat different because of the extra information the hash table is >providing. It seems equally likely that you would search _more_ NPS due to >larger hash tables. > >-Tom
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.