Author: Vincent Diepeveen
Date: 10:02:00 10/30/01
Go up one level in this thread
On October 30, 2001 at 03:24:07, Dan Andersson wrote: >A substantial discussion mentioning APHID was held a while ago. My two cents >where that there are newer and better work division algorithms available now. >APHID works, no doubt. But it represents an old centralised view of parallel >work control. > >MvH Dan Andersson Regrettably you failed to show an algorithm that could get a decent speedup over a cluster! Just saying: "it's possible" is not very smart. I say: for DIEP it's real tough to do it as i either need major bandwidth or a very small latency. So speedups at n nodes where n is a big number, is not going to impress when compared to a 16 processor shared memory machine, for example. I see no way how to get a decent speedup at a cluster without slowing down the program first a huge number of times (because of hashtable latency). Best regards, Vincent
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.