Author: Robert Hyatt
Date: 20:54:14 09/02/02
Go up one level in this thread
On September 02, 2002 at 23:48:58, Robert Hyatt wrote: Now, whether I have done something to slightly hurt SMP performance or not, I don't really know. But the SMP version has been out forever, and _many_ people have run tests and posted the results, and the 1.7 number for duals and 3.0-3.1 numbers for quads are certainly common, as it was in the log I sent you from my quad 700. I like the fact that 2.8 is so much less than 3.1 that it totally refutes my estimate formula. Ask someone with some experience in math and nondeterministic parallel search why your statement is nonsense... In ending this, I _will_ compile a non-SMP version and an SMP version, using the same optimizations and profile-directed feedback, and post the speeds, just to show some real data. Another thing. "proof". You think that you "prove" something when you simply state it is true. That doesn't hold up in _my_ world... While you have that dictionary out, look up "prove" or "proof" as well and conversations here will go much better. SMP vs non-SMP data follows in a few minutes....
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.