Author: Robert Hyatt
Date: 07:31:54 02/27/04
Go up one level in this thread
On February 26, 2004 at 17:04:23, Charles Roberson wrote: >On February 26, 2004 at 12:06:10, Anthony Cozzie wrote: > >>On February 26, 2004 at 11:18:07, Charles Roberson wrote: >> >>> > >>> The second algorithm reminds me of Bob's paper on DTS or any other work >>> stealing approach. >> >>I don't think DTS really suggested a split strategy (other than split at ALL >>nodes if possible). Bob's paper is more, "how can we design a parallel >>structure so that we can split anywhere in the tree". Once you have a working >>DTS implementation, you can split however you want . . . >> >>anthony > > Work stealing algorithms typically try to reduce the overhead of finding a > processor that needs help by assigning the free processor to the one the has > the most work left to do. I see Bob's work as help for work stealing > algorithms. The problem is, in alpha/beta this is hard to determine (most work left to do). You might be about to terminate the search due to a fail-high, for example, but there is not much that would let you know that. It's a very difficult problem, and it is why the YBW-approach is what most everyone uses. IE DTS has YBW-like ideas, although most of what DTS does has to do with the mechanics of doing the parallel search, rather than determining where to further sub-divide the work...
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.