From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1754480Ab3GOI3E (ORCPT ); Mon, 15 Jul 2013 04:29:04 -0400 Received: from merlin.infradead.org ([205.233.59.134]:33741 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754394Ab3GOI3C (ORCPT ); Mon, 15 Jul 2013 04:29:02 -0400 Date: Mon, 15 Jul 2013 10:28:53 +0200 From: Peter Zijlstra To: Vladimir Davydov Cc: Ingo Molnar , linux-kernel@vger.kernel.org, devel@openvz.org, pjt@google.com Subject: Re: [PATCH RFC] sched: move h_load calculation to task_h_load Message-ID: <20130715082853.GM17211@twins.programming.kicks-ass.net> References: <1373705235-22523-1-git-send-email-vdavydov@parallels.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1373705235-22523-1-git-send-email-vdavydov@parallels.com> User-Agent: Mutt/1.5.21 (2012-12-30) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Sat, Jul 13, 2013 at 12:47:15PM +0400, Vladimir Davydov wrote: > --- > kernel/sched/fair.c | 56 ++++++++++++++++++++++---------------------------- > kernel/sched/sched.h | 7 +++---- > 2 files changed, 28 insertions(+), 35 deletions(-) > > diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c > index f77f9c5..de90690 100644 > --- a/kernel/sched/fair.c > +++ b/kernel/sched/fair.c > @@ -4171,47 +4171,46 @@ static void update_blocked_averages(int cpu) > } > > /* > + * Compute the hierarchical load factor for cfs_rq and all its ascendants. > * This needs to be done in a top-down fashion because the load of a child > * group is a fraction of its parents load. > */ > +static void update_cfs_rq_h_load(struct cfs_rq *cfs_rq) > { > + struct rq *rq = rq_of(cfs_rq); > + struct sched_entity *se = cfs_rq->tg->se[cpu_of(rq)]; > unsigned long load; > > + cfs_rq->h_load_next = NULL; > + for_each_sched_entity(se) { > + cfs_rq = cfs_rq_of(se); > + cfs_rq->h_load_next = se; > + if (cfs_rq->last_h_load_update == rq->clock) > + break; > } > > + if (!se) { > + cfs_rq->h_load = rq->avg.load_avg_contrib; > + cfs_rq->last_h_load_update = rq->clock; > + } > > + while ((se = cfs_rq->h_load_next) != NULL) { > + load = cfs_rq->h_load; > + load = div64_ul(load * se->avg.load_avg_contrib, > + cfs_rq->runnable_load_avg + 1); > + cfs_rq = group_cfs_rq(se); > + cfs_rq->h_load = load; > + cfs_rq->last_h_load_update = rq->clock; > + } > } > > static unsigned long task_h_load(struct task_struct *p) > { > struct cfs_rq *cfs_rq = task_cfs_rq(p); > > + if (cfs_rq->last_h_load_update != rq_of(cfs_rq)->clock) > + update_cfs_rq_h_load(cfs_rq); > + > return div64_ul(p->se.avg.load_avg_contrib * cfs_rq->h_load, > cfs_rq->runnable_load_avg + 1); > } OK, fair enough. It does somewhat rely on us getting the single rq->clock update thing right, but that should be ok. But yeah, when you have stupid many cgroups we quickly need less h_load instances than there are cgroups.