Author: Albrecht Heeffer
Date: 23:24:07 01/27/99
Go up one level in this thread
On January 27, 1999 at 22:18:50, Robert Hyatt wrote: > > >I am running bruce's epd suite for bionic-to-move positions on my quad >xeon box... using version 15.20... but I just realized how we ought to >do this... much faster, much more accurate, much less work: > >simply take the bionic executable, _and_ the wcrafty-15.20 executable from >my ftp machine, and run them on this suite on the same machine (one cpu >only preferably) to the same time limit, and compare the results. That >takes out _all_ variability... it ought to be run to 1 min, 2 min and then >3 min, just to allow for search speed variability for whatever they did. >But the 1 min each run first might really answer the question about how close >they are. > >My 15.20 run is matching a lot of moves, most of them in fact, so far. I am >running at 1 min per move, which at 1600mhz is about like 1.5 mins or so on >their hardware which I believe was about 1100mhz total... > >I'll post results probably in the morning. I am using the epdpfga command to >produce one normal epd output record for each input line... using 'st=60' >to set the time before doing the epd command, and I am running with hash=96M >and hashp=20M, not knowing exactly what they used... > >more later... The log files published on the website give the exact settings for each game. In the Open Dutch Tournament we used: hash=96M hasp=10m You can also compare nps statistics to match the same conditions. We reached on average 385.000 nodes and we are about 10% slower than Crafty 15.20 on the same hardware. I'm curious about your results. Albrecht Heeffer
This page took 0.02 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.