[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20180425051509.aohopadqw7q5urbd@vireshk-i7>
Date: Wed, 25 Apr 2018 10:45:09 +0530
From: Viresh Kumar <viresh.kumar@...aro.org>
To: Peter Zijlstra <peterz@...radead.org>
Cc: Valentin Schneider <valentin.schneider@....com>,
Ingo Molnar <mingo@...hat.com>,
Vincent Guittot <vincent.guittot@...aro.org>,
Daniel Lezcano <daniel.lezcano@...aro.org>,
linux-kernel@...r.kernel.org,
Quentin Perret <quentin.perret@....com>,
c@...ez.programming.kicks-ass.net
Subject: Re: [PATCH] sched/fair: Rearrange select_task_rq_fair() to optimize
it
On 24-04-18, 14:35, Peter Zijlstra wrote:
> In any case, if there not going to be conflicts here, this all looks
> good.
Thanks Peter.
I also had another patch and wasn't sure if that would be the right
thing to do. The main purpose of this is to avoid calling
sync_entity_load_avg() unnecessarily.
+++ b/kernel/sched/fair.c
@@ -6196,9 +6196,6 @@ static inline int find_idlest_cpu(struct sched_domain *sd, struct task_struct *p
{
int new_cpu = cpu;
- if (!cpumask_intersects(sched_domain_span(sd), &p->cpus_allowed))
- return prev_cpu;
-
while (sd) {
struct sched_group *group;
struct sched_domain *tmp;
@@ -6652,15 +6649,19 @@ select_task_rq_fair(struct task_struct *p, int prev_cpu, int sd_flag, int wake_f
if (unlikely(sd)) {
/* Slow path */
- /*
- * We're going to need the task's util for capacity_spare_wake
- * in find_idlest_group. Sync it up to prev_cpu's
- * last_update_time.
- */
- if (!(sd_flag & SD_BALANCE_FORK))
- sync_entity_load_avg(&p->se);
+ if (!cpumask_intersects(sched_domain_span(sd), &p->cpus_allowed)) {
+ new_cpu = prev_cpu;
+ } else {
+ /*
+ * We're going to need the task's util for
+ * capacity_spare_wake in find_idlest_group. Sync it up
+ * to prev_cpu's last_update_time.
+ */
+ if (!(sd_flag & SD_BALANCE_FORK))
+ sync_entity_load_avg(&p->se);
- new_cpu = find_idlest_cpu(sd, p, cpu, prev_cpu, sd_flag);
+ new_cpu = find_idlest_cpu(sd, p, cpu, prev_cpu, sd_flag);
+ }
} else if (sd_flag & SD_BALANCE_WAKE) { /* XXX always ? */
/* Fast path */
--
viresh
Powered by blists - more mailing lists