Re: newidle balancing in NUMA domain?

[Ingo Molnar]

* Nick Piggin <npiggin@xxxxxxx> wrote:

> > (I hope i explained my point clearly enough.)
> > 
> > No argument that it could be done cleaner - the duality right now of 
> > both having the fuzzy stats and the rate limiting should be decided 
> > one way or another.
> 
> Well, I would say please keep domain balancing behaviour at least 
> somewhat close to how it was with O(1) scheduler at least until CFS is 
> more sorted out. There is no need to knee jerk because BFS is better 
> at something.

Well, see the change (commit 0ec9fab3d) attached below - even in 
hindsight it was well argued by Mike with quite a few numbers: +68% in 
x264 performance.

If you think it's a bad change (combined with newidle-rate-limit) then 
please show the numbers that counter it and preferably extend 'perf 
bench sched' with the testcases that you care about most.

IIRC (and Peter mentioned this too) Mike got into the whole kbuild angle 
due to comparing mainline scheduler to BFS - but that's really a 
positive thing IMO, not some "knee-jerk reaction".

	Ingo

----------------------->