[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <200802142130.06155.ak@suse.de>
Date: Thu, 14 Feb 2008 21:30:05 +0100
From: Andi Kleen <ak@...e.de>
To: Mike Travis <travis@....com>
Cc: Christoph Lameter <clameter@....com>, Paul Jackson <pj@....com>,
David Rientjes <rientjes@...gle.com>, Lee.Schermerhorn@...com,
mel@....ul.ie, linux-kernel@...r.kernel.org,
akpm@...ux-foundation.org
Subject: Re: [RFC] bitmap relative operator for mempolicy extensions
On Thursday 14 February 2008 21:25:59 Mike Travis wrote:
> Christoph Lameter wrote:
> > On Thu, 14 Feb 2008, Andi Kleen wrote:
> >
> >> You're saying the kernel should use these relative masks internally?
> >
> > There is just some thoughts about this. Did not have time to look into the
> > details. Mike?
>
> There are a few places where the entire cpumask is not needed. For
> example, in the area of core siblings on a node. There's a limit
> to how many cores/threads can be on a node and the full 4k cpumask
> is not needed. How this pertains to this new functionality I'm
> not sure yet.
That would require that the BIOS enumerates the CPUs in a way that
the cores of a socket are continuous. While that's usually true
I don't think there's a guarantee. In theory they could be all scattered.
Ok I theory Linux could remap later but that seems hardly worth
the trouble.
I would rather just use arrays of integers in this case with a reasonable fixed
upper limit (e.g. 16 or 32 -- if there are ever >32 thread x86 CPUs presumably they will
require an updated cpufreq driver too...)
-Andi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists