Author: Robert Hyatt
Date: 06:52:34 01/19/00
Go up one level in this thread
On January 19, 2000 at 01:30:58, Jeremiah Penery wrote: >On January 18, 2000 at 23:53:45, Dann Corbit wrote: > >>On January 18, 2000 at 23:36:20, Tom Kerrigan wrote: >> >>>On January 18, 2000 at 23:17:01, Dann Corbit wrote: >>> >>>>Conversely, we might allocate a *single* hash entry. Either method ought to >>>>dumb the program down just a touch. >>>>;-) >>> >>>Having a single hash entry will not make the program much worse. >>I am surprised to hear that. I have seen curves drawn which show a massive rise >>in ELO for the first few megabytes, which then tapers off to what appears to be >>a logarithmic increase. > >Here is a post Bob made a while back in RGCC, comparing the time taken and the >number of nodes searched for a certain position to a fixed depth, varying the >hash sizes each time. I'm not sure of what version of Crafty it was - whatever >was current on Feb. 24 of 1999. >----------- > >Komputer Korner <korner1@netcom.ca> wrote: >: Yes but it doesn't explain the drastic performance hit if the table is >: not quite "large" enough to fill the last complete iteration. > >You will have to define 'drastic'. Here is some output for crafty, searching to >a fixed depth in a middlegame position. I varied the hash from very small to >very large and recorded the time it took to reach the same depth (only one >processor used to avoid the problem of SMP non-determinism). search speed is >about 200K nodes per sec on this PII/400/xeon machine (one processor). > > hash table sizes >time bytes entries nodes searched >------------------------------------------- >13:13 96K 6K 157,085,451 >12:03 192K 12K 142,633,162 >10:31 384K 24K 123,762,238 > 9:28 768K 49K 110,838,220 > 8:38 1.5M 98K 100,802,339 > 7:48 3M 196K 90,979,000 > 7:22 6M 392K 85,975,960 > 6:53 12M 800K 80,347,212 > 6:32 24M 1.5M 76,465,119 > 6:22 48M 3.0M 74,738,532 > 6:13 96M 6.0M 73,253,374 > 6:05 192M 12.0M 71,581,397 > 6:03 384M 24.0M 71,156,722 > >Here is the position I searched, to a depth of 11 plies: > >[D]3rr1k1/1p2b1pp/p1b1p3/8/PnBBq3/2Q2N2/1PP3PP/R4R1K b - - 0 1 >--------------- > >Jeremiah You're good. _really_ good. :)
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.