From mboxrd@z Thu Jan 1 00:00:00 1970 From: Sergei Shtylyov Date: Wed, 06 Jun 2012 10:53:08 +0000 Subject: Re: [RFC PATCH] sched/numa: do load balance between remote nodes Message-Id: <4FCF3694.206@mvista.com> List-Id: References: <1338965571-9812-1-git-send-email-alex.shi@intel.com> In-Reply-To: <1338965571-9812-1-git-send-email-alex.shi@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: Alex Shi Cc: a.p.zijlstra@chello.nl, anton@samba.org, benh@kernel.crashing.org, cmetcalf@tilera.com, dhowells@redhat.com, davem@davemloft.net, fenghua.yu@intel.com, hpa@zytor.com, ink@jurassic.park.msu.ru, linux-alpha@vger.kernel.org, linux-ia64@vger.kernel.org, linux-kernel@vger.kernel.org, linux-mips@linux-mips.org, linuxppc-dev@lists.ozlabs.org, linux-sh@vger.kernel.org, mattst88@gmail.com, paulus@samba.org, lethal@linux-sh.org, ralf@linux-mips.org, rth@twiddle.net, sparclinux@vger.kernel.org, tony.luck@intel.com, x86@kernel.org, sivanich@sgi.com, greg.pearson@hp.com, kamezawa.hiroyu@jp.fujitsu.com, bob.picco@oracle.com, chris.mason@oracle.com, torvalds@linux-foundation.org, akpm@linux-foundation.org, mingo@kernel.org, pjt@google.com, tglx@linutronix.de, seto.hidetoshi@jp.fujitsu.com, ak@linux.intel.com, arjan.van.de.ven@intel.com Hello. On 06-06-2012 10:52, Alex Shi wrote: > commit cb83b629b Please also specify that commit's summary in parens. > remove the NODE sched domain and check if the node > distance in SLIT table is farther than REMOTE_DISTANCE, if so, it will > lose the load balance chance at exec/fork/wake_affine points. > But actually, even the node distance is farther than REMOTE_DISTANCE, > Modern CPUs also has QPI like connections, that make memory access is "Is" not needed here. > not too slow between nodes. So above losing on NUMA machine make a > huge performance regression on benchmark: hackbench, tbench, netperf > and oltp etc. > This patch will recover the scheduler behavior to old mode on all my > Intel platforms: NHM EP/EX, WSM EP, SNB EP/EP4S, and so remove the > perfromance regressions. (all of them just has 2 kinds distance, 10 21) > Signed-off-by: Alex Shi WBR, Sergei