Re: [RFC PATCH 2/3] sched: add yield_to function

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Fri, 2010-12-03 at 16:09 +0100, Mike Galbraith wrote:
> On Fri, 2010-12-03 at 09:48 -0500, Rik van Riel wrote:
> > On 12/03/2010 09:45 AM, Mike Galbraith wrote:
> > 
> > > I'll have to go back and re-read that.  Off the top of my head, I see no
> > > way it could matter which container the numbers live in as long as they
> > > keep advancing, and stay in the same runqueue.  (hm, task weights would
> > > have to be the same too or scaled. dangerous business, tinkering with
> > > vruntimes)
> > 
> > They're not necessarily in the same runqueue, the
> > VCPU that is given time might be on another CPU
> > than the one that was spinning on a lock.
> 
> I don't think pumping vruntime cross cfs_rq would be safe, for the
> reason noted (et al).  No competition means vruntime is meaningless.
> Donating just advances a clock that nobody's looking at.

Yeah, cross-cpu you have to model it like exchanging lag. That's a
slightly more complicated trick (esp. since we still don't have a proper
measure for lag) but it should be doable.