Author: Robert Hyatt
Date: 11:42:46 05/06/04
Go up one level in this thread
I found a total of 4 runs on the 24 Cray Blitz positions with mt=4. I have copied all 4 to the ftp machine address previously. I did not look at the last 2 except to be sure they were run under exactly the same conditions (AMD changed from 1.8ghz to 2.2ghz at some point, but all of these were on the 2.2ghz processors, all were run back to back with nobody on the machine whatsoever. There are now 4 mt=4 logs, 1 mt=2 log and 1 mt=0 log. I should have some more mt=2 logs. I am looking around now. If I find them I will put them over there as well. Perhaps we can finally define the error for my linear fit formula as well as decide whether 1.7 and 3.1 are bad estimates for 2 and 4 processors respectively, although my judgement says Vincent will never shut up about it. That leads me to believe that he is not doing that well, otherwise he would just be happy to post better results than mine, which he hasn't done since he had to stop reporting that all positions produced super-linear speedups with his program... I'll be interested to see what your numbers look like... And if you don't object, I'll include them and you as a second-author when I write the Crafty parallel search paper, since the numbers will be useful...
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.