Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 14 Feb 1999 20:47:50 -0800 (PST)
From:      Matthew Jacob <mjacob@feral.com>
To:        Matthew Dillon <dillon@apollo.backplane.com>
Cc:        Jaye Mathisen <mrcpu@internetcds.com>, hackers@FreeBSD.ORG
Subject:   Re: Processor affinity?
Message-ID:  <Pine.LNX.4.04.9902142047020.29164-100000@feral-gw>
In-Reply-To: <199902150410.UAA12131@apollo.backplane.com>

next in thread | previous in thread | raw e-mail | index | archive | help

>     as trying to keep a process on a specific cpu.  Whatever algorithm is
>     chosen must deal with the situation under a greater process load.  Often,
>     as on IRIX 6.1 boxes, affinity could make things worse rather then better
>     by unbalancing the cpu's.  Processor affinity makes sense when you have
>     a lot of processors ( you can schedule a process to a group of cpu's to
>     maintain reasonable balancing across the system), but doesn't make much 
>     sense if you only have 2-4.
> 
>     Note that processor affinity scheduling is different from hard-assigning
>     a process to a processor.  Even so, there are very few circumstances where
>     even hard-assigning will do a better job then letting the scheduler do it.
> 

Doesn't it also really depend upon the cache architecture?


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-hackers" in the body of the message



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.LNX.4.04.9902142047020.29164-100000>