Author: Pallav Nawani
Date: 23:15:32 06/11/05
Go up one level in this thread
On June 11, 2005 at 08:45:19, Ricardo Gibert wrote: >I came across the following article: >http://www.newscientist.com/article.ns?id=dn7508 > >and wondered if this could find application in a chess program. I know that the >idea of using the GPU to enhance a chess program has been discussed before and >presumably there is little payoff with this. The truth is that this probably has >not been well investigated. For those interested, it looks like this new >programming language can make the issue much easier to investigate. > GPUs typically don't have branching instructions, which is not good for computer chess. Memory access from GPUs is _very_very_slow_, and this is by design. This will greatly slow down chess engines. GPUs are good for rendering type of jobs (Which they are already used for) where you give them instructions, dump a lot of data, and wait for the results. Pallav
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.