On Thu, 2007-09-13 at 14:14 +0200, Roman Zippel wrote:
quoted text
> Hi,
>=20
> On Thu, 13 Sep 2007, Peter Zijlstra wrote:
>=20
> > >  There's a good reason=20
> > > I put that much effort into maintaining a good, but still cheap avera=
ge,=20
quoted text
> > > it's needed for a good task placement.
> >=20
> > While I agree that having this average is nice, your particular
> > implementation has the problem that it quickly overflows u64 at which
> > point it becomes a huge problem (a CPU hog could basically lock up your
> > box when that happens).
>=20
> If you look at the math, you'll see that I took the overflow into account=
,=20
quoted text
> I even expected it. If you see this effect in my implementation, it would=
=20
quoted text
> be a bug.

Ah, ok, I shall look to your patches in more detail, it was not obvious
from the formulae you posted.

quoted text
> > >  There is of course more than one=20
> > > way to implement this, so you'll have good chances to simply reimplem=
ent=20
quoted text
> > > it somewhat differently, but I'd be surprised if it would be somethin=
g=20
quoted text
> > > completely different.
> >=20
> > Currently we have 2 approximations in place:
> >=20
> >   (leftmost + rightmost) / 2
> >=20
> > and
> >=20
> >   leftmost + period/2   (where period should match the span of the tree=
)
quoted text
> >=20
> > neither are perfect but they seem to work quite well.
>=20
> You need more than two busy loops.=20

I'm missing context here, are you referring to the nice level error or
the avg approximation?

quoted text
> There's a reason I implemented a simple simulator first, so I could=20
> actually study the scheduling behaviour of different load situations. Tha=
t=20
quoted text
> doesn't protect from all surprises of course, but it gives me the=20
> necessary confidence the scheduler will work reasonably even in weird=20
> situations.

Right, I've build user-space simulators too, handy little things to play
with :-)

quoted text
> From these tests I already know that your approximations only work with=20
> rather simple loads.

I've not yet seen it go spectacularly wrong, although admittedly a
highly concurrent kbuild is the most complex task I let loose on it.

Could you perhaps be more specific on the circumstances it breaks down
and what the negative impact is?