Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: SMP variability

Author: Robert Hyatt

Date: 17:04:51 09/05/02

Go up one level in this thread


On September 05, 2002 at 00:47:17, martin fierz wrote:

>On September 05, 2002 at 00:03:37, Robert Hyatt wrote:
>
>>Since you seem to like this kind of data, here is another run.  Kopec 4.
>>It is at the _other_ end of the bad behavior spectrum...
>>
>>log.001:              time=2:15  cpu=99%  mat=0  n=45269425
>>log.002:              time=1:07  cpu=199%  mat=0  n=45786559
>>log.003:              time=1:06  cpu=199%  mat=0  n=45177069
>>log.004:              time=1:05  cpu=199%  mat=0  n=45911977
>>log.005:              time=1:08  cpu=199%  mat=0  n=46238262
>>
>>Again, first line is 1 cpu, rest are 2.  Hate to let Vincent see that as it
>>is "right close" to 2.0 of course.  :)
>>
>>As you can see, some vary wildly, some are rock-solid.  Most lie somewhere
>>in between those two extremes...
>>
>>More data coming...
>
>thanks, this is really interesting. i never realized that SMP searches could be
>so variable in some positions and very stable in others.
>actually, vincent might like this, because it's >2 speedup :-)
>
>aloha
>  martin


Don't give up on me yet.  I have run the 24 DTS positions Vincent loves
with one cpu.  I am running them 4 times with 2 cpus.  I'll post the raw
times tonight and the speedups so we can see exactly what happens.  With
2 cpus, the variability is usually lower than with 3 or 4, of course, so
I will continue the test with 3 and 4 later...

The last couple of 2-cpu tests are running.  They take about 1 hour each
at 3 minutes per position, 24 positions...



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.