[Computer-go] Experiment?

Petr Baudis pasky at ucw.cz
Wed Mar 10 16:17:56 PST 2010

On Wed, Mar 10, 2010 at 12:29:47AM -0800, David Fotland wrote:
> I'm curious if other programs scale strength better with more cores.  

These are some very old data - 9x9 no komi, lossrates (less is better):
20k                     400g  0.188 +-0.02
5k t4 treepar virtloss  400g  0.175 +-0.019
5k t4 treepar           400g  0.183 +-0.019
5k t4 rootpar           400g  0.294 +-0.023

I thought I have much more, but it doesn't seem so. :( I do have free
testing capacity right now, unfortunately the autotest framework cannot
schedule well multi-threaded instances on nodes.

				Petr "Pasky" Baudis
http://pasky.or.cz/ | "Ars longa, vita brevis." -- Hippocrates

More information about the Computer-go mailing list