From: Peter Zijlstra <peterz@infradead.org>
To: Vincent Guittot <vincent.guittot@linaro.org>
Cc: Mel Gorman <mgorman@techsingularity.net>,
"Li, Aubrey" <aubrey.li@linux.intel.com>,
Ingo Molnar <mingo@redhat.com>,
Juri Lelli <juri.lelli@redhat.com>,
Valentin Schneider <valentin.schneider@arm.com>,
Qais Yousef <qais.yousef@arm.com>,
Dietmar Eggemann <dietmar.eggemann@arm.com>,
Steven Rostedt <rostedt@goodmis.org>,
Ben Segall <bsegall@google.com>,
Tim Chen <tim.c.chen@linux.intel.com>,
linux-kernel <linux-kernel@vger.kernel.org>,
Mel Gorman <mgorman@suse.de>, Jiang Biao <benbjiang@gmail.com>
Subject: Re: [RFC PATCH v7] sched/fair: select idle cpu from idle cpumask for task wakeup
Date: Mon, 14 Dec 2020 10:31:22 +0100 [thread overview]
Message-ID: <20201214093122.GX3040@hirez.programming.kicks-ass.net> (raw)
In-Reply-To: <CAKfTPtDBX+scBZiYtDSkXYn7SKDoGYWJiMpCiUvdW1XFz-Fb-Q@mail.gmail.com>
On Mon, Dec 14, 2020 at 09:11:29AM +0100, Vincent Guittot wrote:
> On Fri, 11 Dec 2020 at 23:50, Mel Gorman <mgorman@techsingularity.net> wrote:
> > I originally did something like that on purpose too but Vincent called
> > it out so it is worth mentioning now to avoid surprises. That said, I'm
> > at the point where anything SIS-related simply needs excessive exposure
> > because no matter what it does, someone is unhappy with it.
>
> Yeah, I don't like extending the idle core search loop for something
> that is not related to the idle core search. This adds more work out
> of control of the sis_prop. So I'm still against hiding some idle cpu
> search in idle core search.
The idea, of course, is to do less. The current code is pretty crap in
that it will do a whole bunch of things multiple times.
Also, a possible follow up, would be something like the below (and
remove all the sds->has_idle_cores crud), which brings core scanning
under SIS_PROP.
But it all needs lots of benchmarking :/
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -6095,6 +6095,9 @@ static inline bool test_idle_cores(int c
static inline int __select_idle_core(int core, struct cpumask *cpus, int *idle_cpu)
{
+ if (idle_cpu && (available_idle_cpu(core) || sched_idle_cpu(cpu))
+ *idle_cpu = core;
+
return -1;
}
@@ -6109,7 +6112,6 @@ static int select_idle_cpu(struct task_s
{
struct cpumask *cpus = this_cpu_cpumask_var_ptr(select_idle_mask);
int this = smp_processor_id();
- bool smt = test_idle_cores(this, false);
int i, cpu, idle_cpu = -1, nr = INT_MAX;
struct sched_domain *this_sd;
u64 time;
@@ -6120,7 +6122,7 @@ static int select_idle_cpu(struct task_s
cpumask_and(cpus, sched_domain_span(sd), p->cpus_ptr);
- if (sched_feat(SIS_PROP) && !smt) {
+ if (sched_feat(SIS_PROP)) {
u64 avg_cost, avg_idle, span_avg;
/*
@@ -6140,26 +6142,17 @@ static int select_idle_cpu(struct task_s
}
for_each_cpu_wrap(cpu, cpus, target) {
- if (smt) {
- i = __select_idle_core(cpu, cpus, &idle_cpu);
- if ((unsigned)i < nr_cpumask_bits)
- return i;
-
- } else {
- if (!--nr)
- return -1;
-
- if (available_idle_cpu(cpu) || sched_idle_cpu(cpu)) {
- idle_cpu = cpu;
- break;
- }
+ if (!--nr)
+ break;
+
+ i = __select_idle_core(cpu, cpus, &idle_cpu);
+ if ((unsigned)i < nr_cpumask_bits) {
+ idle_cpu = i;
+ break;
}
}
- if (smt)
- set_idle_cores(this, false);
-
- if (sched_feat(SIS_PROP) && !smt) {
+ if (sched_feat(SIS_PROP)) {
time = cpu_clock(this) - time;
update_avg(&this_sd->avg_scan_cost, time);
}
next prev parent reply other threads:[~2020-12-14 9:32 UTC|newest]
Thread overview: 22+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-12-09 6:24 Aubrey Li
2020-12-09 8:15 ` Vincent Guittot
2020-12-09 10:58 ` Li, Aubrey
2020-12-09 13:09 ` Vincent Guittot
2020-12-09 14:53 ` Li, Aubrey
2020-12-09 14:36 ` Mel Gorman
2020-12-10 8:23 ` Li, Aubrey
2020-12-10 11:34 ` Mel Gorman
2020-12-10 12:21 ` Li, Aubrey
2020-12-10 12:58 ` Mel Gorman
2020-12-11 17:44 ` Peter Zijlstra
2020-12-11 20:43 ` Mel Gorman
2020-12-11 22:19 ` Peter Zijlstra
2020-12-11 22:50 ` Mel Gorman
2020-12-14 8:11 ` Vincent Guittot
2020-12-14 9:31 ` Peter Zijlstra [this message]
2020-12-14 12:36 ` Mel Gorman
2020-12-14 15:01 ` Peter Zijlstra
2020-12-14 9:32 ` Peter Zijlstra
2020-12-14 9:18 ` Vincent Guittot
2020-12-14 12:42 ` Mel Gorman
2020-12-14 7:53 ` Li, Aubrey
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20201214093122.GX3040@hirez.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=aubrey.li@linux.intel.com \
--cc=benbjiang@gmail.com \
--cc=bsegall@google.com \
--cc=dietmar.eggemann@arm.com \
--cc=juri.lelli@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mgorman@suse.de \
--cc=mgorman@techsingularity.net \
--cc=mingo@redhat.com \
--cc=qais.yousef@arm.com \
--cc=rostedt@goodmis.org \
--cc=tim.c.chen@linux.intel.com \
--cc=valentin.schneider@arm.com \
--cc=vincent.guittot@linaro.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome