Re: Scheduler regression from caffcdd8d27ba78730d5540396ce72ad022aff2c

[Peter Zijlstra]

On Fri, Jul 18, 2014 at 08:01:26AM -0500, Bruno Wolff III wrote:
> build_sched_domain: cpu: 0 level: SMT cpu_map: 0-3 tl->mask: 0,2
> [    0.254433] build_sched_domain: cpu: 0 level: MC cpu_map: 0-3 tl->mask: 0
> [    0.254516] build_sched_domain: cpu: 0 level: DIE cpu_map: 0-3 tl->mask: 0-3
> [    0.254600] build_sched_domain: cpu: 1 level: SMT cpu_map: 0-3 tl->mask: 1,3
> [    0.254683] build_sched_domain: cpu: 1 level: MC cpu_map: 0-3 tl->mask: 1
> [    0.254766] build_sched_domain: cpu: 1 level: DIE cpu_map: 0-3 tl->mask: 0-3
> [    0.254850] build_sched_domain: cpu: 2 level: SMT cpu_map: 0-3 tl->mask: 0,2
> [    0.254932] build_sched_domain: cpu: 2 level: MC cpu_map: 0-3 tl->mask: 2
> [    0.255005] build_sched_domain: cpu: 2 level: DIE cpu_map: 0-3 tl->mask: 0-3
> [    0.255091] build_sched_domain: cpu: 3 level: SMT cpu_map: 0-3 tl->mask: 1,3
> [    0.255176] build_sched_domain: cpu: 3 level: MC cpu_map: 0-3 tl->mask: 3
> [    0.255260] build_sched_domain: cpu: 3 level: DIE cpu_map: 0-3 tl->mask: 0-3

*blink*...

That's, shall we say, unexpected. Let me ponder that a bit. HPA any clue
why a machine might report such a weird topology? AFAIK threads _always_
share cache.  So how can cpu_coregroup_mask be a subset (instead of a
superset) of topology_thread_cpumask?

Let me go stare at the x86 topology mask setup code.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/