[PATCH v2 00/15] Make SMP timers standalone

[Shilimkar, Santosh]

On Thu, Jan 5, 2012 at 11:33 AM, Marc Zyngier <marc.zyngier at arm.com> wrote:
> On 04/01/12 21:47, Russell King - ARM Linux wrote:
>> On Wed, Jan 04, 2012 at 05:39:16PM +0000, Marc Zyngier wrote:
>>> On 22/12/11 19:32, Russell King - ARM Linux wrote:
>>>> You're also aware I assume that local timers are different from the global
>>>> timer itself, and require to have additional callbacks for broadcasting
>>>> the global timer tick?
>>>
>>> I was under the impression that we either have the global timer together
>>> with broadcasting or the local timers. Completely removing the broadcast
>>> callback doesn't seem to generate any ill effect as long as the local
>>> timers are used instead of the global timer (which of course requires
>>> broadcast in the SMP configuration).
>>
>> I believe even if you have local timers, there are situations where the
>> these will be disabled and the kernel will switch to broadcasting from a
>> global timer tick.  I think such a scenario would be like that encountered
>> with OMAP, where suspending a CPU stops its TWD - meaning that the TWD
>> can't be used to wake the CPU from one of the deeper idle states.
>
> Yes, being able to wake up is definitely a good reason to keep a global
> timer around, not to mention other artifacts (TWD calibration being one,
> though on its way out thanks to LinusW patch series).
>
Catching up bit late on this tread.
I just want to echo the fact that we always need the wakeup capable
clock-event till the local timers themselves can wakup the CPU
from low power states.

Regards
Santosh