Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash hit ratios

Author: Rick Bischoff

Date: 07:51:51 09/30/04

Go up one level in this thread


>Your numbers seem low. I can get 16-25% exact+lower+upper hash cutoffs in the
>opening position.

Here are the results from a 100s search from the opening... They do indeed seem
low, in particular I am only getting ~50% saturation...

***** 100552 ms elapsed 2736128 nodes 27211.1 nps.
109476/262144 saturation.
638188 probes %12.6201 hits,
%0.0394868 exact
%2.56351 lower,
%1.11127 upper
188227 stores
36616 collisions
36939 overwrites (same key, same age, higher depth)

Now, I made the change to store depth=0 stuff too (e.g., qsearch):

***** 100227 ms elapsed 2768896 nodes 27626.2 nps.
259769/262144 saturation.
625716 probes %35.9582 hits,
%0.045388 exact,
%2.37967 lower,
%1.11137 upper.
1754748 stores,
1117774 collisions
363316 overwrites.

Now, I increased the hash table size by 16x:

***** 100146 ms elapsed 2670592 nodes 26667 nps.
1064999/4194304 saturation.
621147 probes %52.1934 hits
%0.0381552 exact
%2.5073 lower
%1.12389 upper
1706027 stores
176856 collisions
443522 overwrites

Something smells fishy... I have no idea what!!



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.