Re: [rfc][patch] select_idle_sibling() inducing bouncing on westmere

[Mike Galbraith <efault@gmx.de>]

On Fri, 2012-05-25 at 08:14 +0200, Mike Galbraith wrote: 
> On Thu, 2012-05-24 at 15:20 +0200, Peter Zijlstra wrote: 
> > On Thu, 2012-05-24 at 15:17 +0200, Peter Zijlstra wrote:
> > > For power7 with 4 smt it would end up as 1230 I guess.
> > 
> > One more thing, in light of that, can't you simplify the core stuff to
> > be a simple shift as well? map them 123450, does it really matter to
> > pair them off like 103254 ?
> 
> In my head it does.  Buddies need to be glued together so we can use
> them as a team to convert overlap and kick butt on ramp up.

Ew. 3.4 went broke for Q6600, and performance went... far far away.

[    0.200057] CPU0 attaching sched-domain:
[    0.204016]  domain 0: span 0-3 level MC
[    0.208015]   groups: 0 1 2 3
[    0.210970] CPU1 attaching sched-domain:
[    0.212014]  domain 0: span 0-3 level MC
[    0.216016]   groups: 1 2 3 0
[    0.220016] CPU2 attaching sched-domain:
[    0.224015]  domain 0: span 0-3 level MC
[    0.228016]   groups: 2 3 0 1
[    0.232015] CPU3 attaching sched-domain:
[    0.236016]  domain 0: span 0-3 level MC
[    0.240017]   groups: 3 0 1 2

11.791806 usecs/loop -- avg 11.534552 173.4 KHz

Cause: sometimes during boot, hw Siamese twins are 0-1 2-3, and
sometimes, as in this boot, 0-3 1-2, so busted groups above does the
worst thing possible.  When twins are 0-1 2-3, busted groups doesn't
matter with $subject patch applied, it glues Siamese twins back
together.  During this boot, it created Siamese aliens from hell.

-Mike