I'm planning on finalising the schedule for the power management
minisummit at the Linux Plumbers Conference by the end of this week. If
anyone has topics that they'd like to discuss (or be discussed, if you
can't make it) then please let me know or register at
http://www.linuxplumbersconf.org/2010/ocw/events/proposals/ and file a
proposal there. We're likely to be concentrating on userspace/kernel
interaction rather than pure kernel topics, so if you've got anything to
say about how we can handle application requirements without
compromising on our power savings then we'd love to hear from you.
--
Matthew Garrett | [email protected]
Matthew Garrett wrote:
> I'm planning on finalising the schedule for the power management
> minisummit at the Linux Plumbers Conference by the end of this week. If
> anyone has topics that they'd like to discuss (or be discussed, if you
> can't make it) then please let me know or register at
> http://www.linuxplumbersconf.org/2010/ocw/events/proposals/ and file a
That link did not really work for me, but I think I found the right
place after a few clicks.
> proposal there. We're likely to be concentrating on userspace/kernel
> interaction rather than pure kernel topics, so if you've got anything to
> say about how we can handle application requirements without
> compromising on our power savings then we'd love to hear from you.
Just FYI, I proposed a discussion on the issues that create such a large
power gap between a CPUIdle-only system and a system with suspend (with
the assumption that the lowest C-state can be hit by both).
I don't really want to present a formal lecture on it, but it would be
nice to discuss with the group what can be done to shrink that gap a
little in kernel space (badly behaving timers in platform-independent
code) and in userspace (power-aware timers, coalescing timers, etc).
This is not intended to be another suspend-blocker discussion, but more
of a discussion on what it takes to get CPUIdle to be more effective.
Regards,
Mike
On Tue, Aug 03, 2010 at 03:30:56PM -0500, Mike Turquette wrote:
> That link did not really work for me, but I think I found the right
> place after a few clicks.
Oops! The link should have been
http://www.linuxplumbersconf.org/2010/ocw/events/LPC2010MC/proposals .
> Just FYI, I proposed a discussion on the issues that create such a large
> power gap between a CPUIdle-only system and a system with suspend (with
> the assumption that the lowest C-state can be hit by both).
>
> I don't really want to present a formal lecture on it, but it would be
> nice to discuss with the group what can be done to shrink that gap a
> little in kernel space (badly behaving timers in platform-independent
> code) and in userspace (power-aware timers, coalescing timers, etc).
That sounds great. The idea is to be fairly discussion-oriented, perhaps
with a short presentation to cover the outline of the issue, so this
kind of topic is ideal.
Thanks,
--
Matthew Garrett | [email protected]