Commit 6956dc56 authored Jul 20, 2012 by Alex Shi Committed by Ingo Molnar Jul 26, 2012

sched/numa: Add SD_PERFER_SIBLING to CPU domain



Commit 8e7fbcbc ("sched: Remove stale power aware scheduling remnants
and dysfunctional knobs") removed SD_PERFER_SIBLING from the CPU domain.

On NUMA machines this causes that load_balance() doesn't perfer LCPU in
 same physical CPU package.

It causes some actual performance regressions on our NUMA machines from
Core2 to NHM and SNB.

Adding this domain flag again recovers the performance drop.

This change doesn't have any bad impact on any of my benchmarks:
 specjbb, kbuild, fio, hackbench .. etc, on all my machines.

Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1342765190-21540-1-git-send-email-alex.shi@intel.com


Signed-off-by: Ingo Molnar <mingo@kernel.org>

parent 8323f26c

include/linux/topology.h

+1 −0

Original line number	Diff line number	Diff line
		@@ -164,6 +164,7 @@ int arch_update_cpu_topology(void);
		\| 0*SD_SHARE_CPUPOWER \
		\| 0*SD_SHARE_PKG_RESOURCES \
		\| 0*SD_SERIALIZE \
		\| 1*SD_PREFER_SIBLING \
		, \
		.last_balance = jiffies, \
		.balance_interval = 1, \

Admin message