Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why is SMP not standard in chessprograms?

Author: Dann Corbit

Date: 10:40:46 09/26/01

Go up one level in this thread


On September 26, 2001 at 10:17:06, Robert Hyatt wrote:
[snip]
>I don't know where this rumor comes from.  My SMP version is so close to my
>non-SMP version in speed, I can't tell the difference.  The only reason I have
>a non-SMP version is that it doesn't make sense to run two threads on a single
>cpu.  And beginners would make that mistake too often to think about.
>
>In Crafty, the "SMP overhead" on a non-SMP machine is one memory load and
>one conditional branch.  This load/branch tests to see if there is an idle
>processor.  The program currently executes maybe 8,000 instructions per node.
>That turns it into 8,002.   That is no penalty to speak of.
>
>Other issues like libraries simply don't count as they aren't used inside the
>search...

There is a much larger difference with the Windows compilers for SMP/non-SMP.
Could be there is some other difference in compiler settings that accounts for
it, however.




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.