[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20071030071919.GB17074@elte.hu>
Date: Tue, 30 Oct 2007 08:19:19 +0100
From: Ingo Molnar <mingo@...e.hu>
To: Ian Pratt <Ian.Pratt@...cam.ac.uk>
Cc: Jeremy Fitzhardinge <jeremy@...p.org>,
Zachary Amsden <zach@...are.com>,
Glauber de Oliveira Costa <gcosta@...hat.com>,
linux-kernel@...r.kernel.org, tglx@...utronix.de,
rusty@...tcorp.com.au, Avi Kivity <avi@...ranet.com>,
kvm-devel@...ts.sourceforge.net,
Glauber de Oliveira Costa <glauber@....localdomain>,
Dan Hecht <dhecht@...are.com>,
Garrett Smith <garrett@...are.com>
Subject: Re: [PATCH] raise tsc clocksource rating
* Ian Pratt <Ian.Pratt@...cam.ac.uk> wrote:
> > > Sigh, I don't really want to have this fight again.
> >
> > i dont remember us having discussed this before, ever. If there's any
> > "fight" about monotonicity and SMP then it would be a pretty onesided
> > affair, with you being beaten up seriously ;-)
>
> Actually, it is possible, even for NUMA systems with CPUs running off
> completely different oscillators, and in the presence of CPU frequency
> changes, power management, and even in the presence of thermal
> throttling (though the latter introduces temporary inaccuracies it
> doesn't affect monotonicity or rate).
>
> Take a look at the Xen code to see how each physical CPU is
> independently calibrated on an ongoing basis, how movement of VCPUs
> between physical CPUs is tracked, and how shared variables are used to
> ensure montonicity if a guest requires it.
I think that's wishful thinking. Check out:
http://people.redhat.com/mingo/time-warp-test/time-warp-test.c
change TEST_TSC to 0, run it on an SMP guest (on a reasonably fast
machine) and let me know whether you can make SMP guests not come up
with monotonicity violations in the CLOCK tests.
Ingo
-
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