Re: [PATCH V6 0/2] tracing, perf: cpu hotplug trace events

[Amit Kucheria]

On Fri, Mar 4, 2011 at 12:42 AM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> On Wed, 2011-03-02 at 23:07 +0100, Thomas Renninger wrote:
>
>> I doubt cpu offlining is the proper instrument to save power.
>> You want to prevent the CPU of being used by ripping it out from scheduler decisions and
>> make sure it doesn't get interrupts by offlining. But the (latency) price is high.
>
> I could imagine that a server could use this for power savings to take
> down all but 1 CPU on off hours. When it knows its not going to get much
> action but still needs to remain online. Then just before peak times
> begin, online the other CPUs.
>
> But anything more dynamic than that, I can't see it really worth it. As
> the latency to bring the other CPU online, may miss a peak when it was
> needed.
>

ARM SoCs require both cores to be idle to hit the really low power
retention/off states. Hoping for both cores to go idle at the same
instant causes several low power opportunities to be lost. Hence the
experiments with hotplug to improve the idle characteristics of the
system. On our wiki page[1], you can see some results under "Idle
improvement" section.

But sched_mc does seem like a more appropriate way to help nudge all
the workload onto a single core.

Regards,
Amit


[1] https://wiki.linaro.org/WorkingGroups/PowerManagement/Doc/Hotplug
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/