On Wed, 2016-06-01 at 08:01 +0800, Yuyang Du wrote:
> On Wed, Jun 01, 2016 at 07:07:13AM +0200, Mike Galbraith wrote:
> > On Tue, 2016-05-31 at 09:31 +0800, Yuyang Du wrote:
> > > On Tue, May 31, 2016 at 11:21:46AM +0200, Peter Zijlstra wrote:
> > > > On Tue, May 31, 2016 at 09:11:37AM +0800, Yuyang Du wrote:
> > > > > The SD_BALANCE_WAKE is irrelevant in the contexts of these two
> > > > > removals,
> > > > > and in addition SD_BALANCE_WAKE is not and should not be set in any
> > > > > sched_domain flags so far.
> > > >
> > > > This Changelog doesn't make any sense...
> > >
> > > How? SD_BALANCE_WAKE is not in any sched_domain flags (sd->flags), even if
> > > it is, it is not used anywhere, no?
> >
> > If the user chooses to set SD_BALANCE_WAKE in sd->flags, it is in fact
> > used. It's just not turned on by default due to full balance on every
> > wakeup being far too painful to do by default.
>
> Yup. Up to this point, we don't have any disagreement. And I don't think we
> have any disagreement conceptually. What the next patch really does is:
>
> (1) we don't remove SD_BALANCE_WAKE as an important sched_domain flag, on
> the contrary, we strengthen it.
>
> (2) the semantic of SD_BALANCE_WAKE is currently represented by
> SD_WAKE_AFFINE,
> we actually remove this representation.
Nope, those two have different meanings. We pass SD_BALANCE_WAKE to
identify a ttwu() wakeup, just as we pass SD_BALANCE_FORK to say we're
waking a child. SD_WAKE_AFFINE means exactly what it says, but is only
applicable to ttwu() wakeups.
> (3) regarding the semantic of SD_WAKE_AFFINE, it is really not about selecting
> waker CPU or about the fast path. Conceptually, it is just saying the
> waker
> CPU is a valid and important candidate if SD_BALANCE_WAKE, which is just
> so
> obvious, so I don't think it deserves to be a separate sched_domain flag.
SD_WAKE_AFFINE being a separate domain flag, the user can turn it
on/off... separately :)
> (4) the outcome is, if SD_BALANCE_WAKE, we definitely will/should try waker
> CPU,
> and if !SD_BALANCE_WAKE, we don't try waker CPU. So nothing functional is
> changed.
If wake_wide() says we do not want an affine wakeup, but you apply
SD_WAKE_AFFINE meaning to SD_BALANCE_WAKE and turn it on in ->flags,
we'll give the user a free sample of full balance cost, no?
-Mike
