Author: Robert Hyatt
Date: 11:13:10 09/03/03
Go up one level in this thread
On September 03, 2003 at 14:08:13, Gian-Carlo Pascutto wrote: >On September 03, 2003 at 14:06:50, Robert Hyatt wrote: > >>On September 03, 2003 at 13:35:36, Gian-Carlo Pascutto wrote: >> >>>On September 03, 2003 at 13:21:27, Robert Hyatt wrote: >>> >>>>I just set all the "helper" process "stop" flags to 1 and >>> >>>...and by doing so force a remote memory access. >>> >>>-- >>>GCP >> >>At most 6,000 times in a 1 minute search (from data posted elsewhere in >>this thread.) How much time does it take to do 6000 stores? worst case >>of 6000 microseconds? 6 milliseconds? Out of one minute X N cpus. Or >>6ms out of N minutes? I think that can be totally discounted without >>introducing any significant error at all. > >Does this increase with more cpus? What kind of slope? > >-- >GCP I don't have data beyond 4. But for 1, there are no splits. ( :) ). For 2 and 4, it appears to be near linear, but 2 and 4 are not enough data points. I have some old Cray Blitz data that goes to 8 and 16 (I can post some here) that suggests that this same "near-linear" number holds there. However, CB and Crafty don't work the same so comparing them is misleading.
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.