From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755363Ab1ABH6Q (ORCPT ); Sun, 2 Jan 2011 02:58:16 -0500 Received: from mail.windriver.com ([147.11.1.11]:35283 "EHLO mail.windriver.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753560Ab1ABHXF (ORCPT ); Sun, 2 Jan 2011 02:23:05 -0500 From: Paul Gortmaker To: stable@kernel.org, linux-kernel@vger.kernel.org Cc: stable-review@kernel.org, Peter Zijlstra , Ingo Molnar , Paul Gortmaker Subject: [34-longterm 094/260] sched: Pre-compute cpumask_weight(sched_domain_span(sd)) Date: Sun, 2 Jan 2011 02:16:30 -0500 Message-Id: <1293952756-15010-95-git-send-email-paul.gortmaker@windriver.com> X-Mailer: git-send-email 1.7.3.3 In-Reply-To: <1293952756-15010-1-git-send-email-paul.gortmaker@windriver.com> References: <1293952756-15010-1-git-send-email-paul.gortmaker@windriver.com> X-OriginalArrivalTime: 02 Jan 2011 07:21:53.0227 (UTC) FILETIME=[BABBE1B0:01CBAA4D] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org From: Peter Zijlstra commit 669c55e9f99b90e46eaa0f98a67ec53d46dc969a upstream. Dave reported that his large SPARC machines spend lots of time in hweight64(), try and optimize some of those needless cpumask_weight() invocations (esp. with the large offstack cpumasks these are very expensive indeed). Reported-by: David Miller Signed-off-by: Peter Zijlstra LKML-Reference: Signed-off-by: Ingo Molnar Signed-off-by: Paul Gortmaker --- include/linux/sched.h | 1 + kernel/sched.c | 3 +++ kernel/sched_fair.c | 12 +++++------- 3 files changed, 9 insertions(+), 7 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index 25f9669..89e0fee 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -954,6 +954,7 @@ struct sched_domain { char *name; #endif + unsigned int span_weight; /* * Span of all CPUs in this domain. * diff --git a/kernel/sched.c b/kernel/sched.c index 074c4d8..b741de5 100644 --- a/kernel/sched.c +++ b/kernel/sched.c @@ -6296,6 +6296,9 @@ cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu) struct rq *rq = cpu_rq(cpu); struct sched_domain *tmp; + for (tmp = sd; tmp; tmp = tmp->parent) + tmp->span_weight = cpumask_weight(sched_domain_span(tmp)); + /* Remove the sched domains which do not contribute to scheduling. */ for (tmp = sd; tmp; ) { struct sched_domain *parent = tmp->parent; diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c index ae4d842..a29df86 100644 --- a/kernel/sched_fair.c +++ b/kernel/sched_fair.c @@ -1542,9 +1542,7 @@ select_task_rq_fair(struct rq *rq, struct task_struct *p, int sd_flag, int wake_ * Pick the largest domain to update shares over */ tmp = sd; - if (affine_sd && (!tmp || - cpumask_weight(sched_domain_span(affine_sd)) > - cpumask_weight(sched_domain_span(sd)))) + if (affine_sd && (!tmp || affine_sd->span_weight > sd->span_weight)) tmp = affine_sd; if (tmp) { @@ -1588,10 +1586,10 @@ select_task_rq_fair(struct rq *rq, struct task_struct *p, int sd_flag, int wake_ /* Now try balancing at a lower domain level of new_cpu */ cpu = new_cpu; - weight = cpumask_weight(sched_domain_span(sd)); + weight = sd->span_weight; sd = NULL; for_each_domain(cpu, tmp) { - if (weight <= cpumask_weight(sched_domain_span(tmp))) + if (weight <= tmp->span_weight) break; if (tmp->flags & sd_flag) sd = tmp; @@ -2323,7 +2321,7 @@ unsigned long __weak arch_scale_freq_power(struct sched_domain *sd, int cpu) unsigned long default_scale_smt_power(struct sched_domain *sd, int cpu) { - unsigned long weight = cpumask_weight(sched_domain_span(sd)); + unsigned long weight = sd->span_weight; unsigned long smt_gain = sd->smt_gain; smt_gain /= weight; @@ -2356,7 +2354,7 @@ unsigned long scale_rt_power(int cpu) static void update_cpu_power(struct sched_domain *sd, int cpu) { - unsigned long weight = cpumask_weight(sched_domain_span(sd)); + unsigned long weight = sd->span_weight; unsigned long power = SCHED_LOAD_SCALE; struct sched_group *sdg = sd->groups; -- 1.7.3.3