Author: Robert Hyatt
Date: 21:01:24 10/04/03
Go up one level in this thread
On October 04, 2003 at 12:48:21, Dan Andersson wrote: > I saying that to be able to probe N times without any latency hits each probe >must be in a local TT memory. If you have a shared TT and mutiple memory banks >you might get lucky and only incur the cache snoop penalty. > I am a bit tired and might have misunderstood the topic. But that's my take on >it. > >MvH Dan Andersson One point is that it is _possible_ (but not probable) that even if all hash probes are to remote memory addresses, none conflict and everything is evenly distributed so there are no router hot-spots. But not very likely.
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.