Re: load balancing regression since commit 367456c7

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

On Fri, 2012-04-20 at 10:13 -0700, Tim Chen wrote:
> On Fri, 2012-04-20 at 18:53 +0200, Peter Zijlstra wrote:
> > On Fri, 2012-04-20 at 09:40 -0700, Tim Chen wrote:
> > > We also do not see regression for hackbench on WSM-EP, but only 
> > > on machines with Sandy-Bridge EP (2 socket, 8 cores/socket, HT enabled).  
> > 
> > Argh, Suresh, any idea what could be different and relevant to this
> > issue between WSM and SNB -EP ?
> > 
> > Tim, can you see the problem on the desktop SNB part? That's the only
> > SNB I have available.
> > 
> > > We are not running hackbench in cgroup for this test.
> > > 
> > > The Sandy Bridge EP machines installed has FC16, so I think it uses 
> > > systemd. The machine on our WSM EP has FC15, which also has systemd.
> > 
> > Ah, but your .config has:
> > 
> > CONFIG_CGROUP_SCHED=y
> > CONFIG_SCHED_AUTOGROUP=y
> > 
> > and systemd, when cpu-cgroup is available, will automagically use it.
> > 
> > Could you disable those two CONFIG knobs and see if it persists?
> > 
> 
> Turning those off did recover the regression.

OK, I'll go stare at the cgroup part then.. Thanks!