Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Neverending story with incomplete tablebases

Author: Robert Hyatt

Date: 09:02:37 08/19/03

Go up one level in this thread


On August 19, 2003 at 05:39:17, Sune Fischer wrote:

>On August 19, 2003 at 05:05:52, Gerd Isenberg wrote:
>
>Hi Gerd,
>
>>>Hmm, I don't really understand what that means.
>>
>>Using MMX/XMM does not contradict save threads.
>>Os is responcable to save/restore all registers during context switch.
>
>thanks, that is good news.
>
>>>Is it possible to set the affinity of each thread, so that no context switching
>>>occurs? That way they wouldn't mess up eachothers registers, would they?
>>>
>>
>>Context switching occurs in milli second range.
>>
>>Gerd
>
>That doesn't sound ultra fast to me, but if it doesn't happen too often I guess
>there is no problem.
>
>-S.


A good O/S does this correctly.  IE on Linux, threads don't "bounce" around
very much, although there are plenty of interrupts that have to be handled by
some processor or another.

It's not a big deal.  And since linux maps the kernel into every process at
the same address, the map stuff doesn't have to be diddled with, and since
X86 caches real addresses, caches don't have to be flushed frequently.  The
overall effect is that there is very little overhead to deal with in the normal
case of N cpus, M threads, M <= N.  For M > N, things can become problematic.
But in chess nobody would want to do that.



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.