Idle-suspend a little too intrusive to user experience?

Ben Schwartz bmschwar at fas.harvard.edu
Tue May 11 23:00:50 EDT 2010


On Tue, 2010-05-11 at 22:31 -0400, Martin Langhoff wrote:
> On Tue, May 11, 2010 at 8:59 PM, John Gilmore <gnu at toad.com> wrote:
> >> > just fix the kernel so the suspend
> >> > ends when the next process wants to run.
> >>
> >> Have a look at powertop -- you'll never suspend, there are several
> >> hundred wakeups per second.
> >
> > Did we give up on fixing these?
> 
> No, and as Ben mentions, it gets better gradually.

So for the record, I am actually making the much stronger claim that _we
are already there_.  Thanks to work done over the last few years, Sugar
is free of polling and ready for principled kernel-based aggressive
suspend.  Only a few other components generate wakeups in recent builds,
amounting to less than one wakeup a second.  I predict they will be easy
to fix, and are now polling only because there's no incentive to do
otherwise.  Thanks to much work by Richard Smith, we also have reliable
wakeup timers from the EC.  

All that remains, then, is to write the cpuidle driver that integrates
correctly with EC and DCON.  Compared to the software OLPC has
accomplished so far, this is really a small job.  It's simply never been
assigned, and no one has volunteered.

--Ben




More information about the Devel mailing list