[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <ZaZTlcFZaQefnf1v@gmail.com>
Date: Tue, 16 Jan 2024 10:59:49 +0100
From: Ingo Molnar <mingo@...nel.org>
To: Vincent Guittot <vincent.guittot@...aro.org>
Cc: mingo@...hat.com, peterz@...radead.org, juri.lelli@...hat.com,
dietmar.eggemann@....com, rostedt@...dmis.org, bsegall@...gle.com,
mgorman@...e.de, bristot@...hat.com, vschneid@...hat.com,
wkarny@...il.com, torvalds@...ux-foundation.org,
qyousef@...alina.io, tglx@...utronix.de, rafael@...nel.org,
viresh.kumar@...aro.org, linux-kernel@...r.kernel.org,
linux-pm@...r.kernel.org
Subject: Re: [PATCH] sched/fair: Fix frequency selection for non invariant
case
* Vincent Guittot <vincent.guittot@...aro.org> wrote:
> When frequency invariance is not enabled, get_capacity_ref_freq(policy)
> returns the current frequency and the performance margin applied by
> map_util_perf(), enabled the utilization to go above the maximum compute
> capacity and to select a higher frequency than the current one.
>
> The performance margin is now applied earlier in the path to take into
> account some utilization clampings and we can't get an utilization higher
> than the maximum compute capacity.
>
> We must use a frequency above the current frequency to get a chance to
> select a higher OPP when the current one becomes fully used. Apply
> the same margin and returns a frequency 25% higher than the current one in
> order to switch to the next OPP before we fully use the cpu at the current
> one.
>
> Reported-by: Linus Torvalds <torvalds@...ux-foundation.org>
> Closes: https://lore.kernel.org/lkml/CAHk-=wgWcYX2oXKtgvNN2LLDXP7kXkbo-xTfumEjmPbjSer2RQ@mail.gmail.com/
> Reported-by: Wyes Karny <wkarny@...il.com>
> Closes: https://lore.kernel.org/lkml/20240114091240.xzdvqk75ifgfj5yx@wyes-pc/
> Fixes: 9c0b4bb7f630 ("sched/cpufreq: Rework schedutil governor performance estimation")
> Signed-off-by: Vincent Guittot <vincent.guittot@...aro.org>
> Tested-by: Wyes Karny <wkarny@...il.com>
> ---
> kernel/sched/cpufreq_schedutil.c | 6 +++++-
> 1 file changed, 5 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
> index 95c3c097083e..d12e95d30e2e 100644
> --- a/kernel/sched/cpufreq_schedutil.c
> +++ b/kernel/sched/cpufreq_schedutil.c
> @@ -133,7 +133,11 @@ unsigned long get_capacity_ref_freq(struct cpufreq_policy *policy)
> if (arch_scale_freq_invariant())
> return policy->cpuinfo.max_freq;
>
> - return policy->cur;
> + /*
> + * Apply a 25% margin so that we select a higher frequency than
> + * the current one before the CPU is full busy
> + */
> + return policy->cur + (policy->cur >> 2);
> }
I've updated the changelog to better express what was broken and how we
fixed it. Ack?
Ingo
==========================>
From: Vincent Guittot <vincent.guittot@...aro.org>
Date: Sun, 14 Jan 2024 19:36:00 +0100
Subject: [PATCH] sched/fair: Fix frequency selection for non-invariant case
Linus reported a ~50% performance regression on single-threaded
workloads on his AMD Ryzen system, and bisected it to:
9c0b4bb7f630 ("sched/cpufreq: Rework schedutil governor performance estimation")
When frequency invariance is not enabled, get_capacity_ref_freq(policy)
is supposed to return the current frequency and the performance margin
applied by map_util_perf(), enabling the utilization to go above the
maximum compute capacity and to select a higher frequency than the current one.
After the changes in 9c0b4bb7f630, the performance margin was applied
earlier in the path to take into account utilization clampings and
we couldn't get a utilization higher than the maximum compute capacity,
and the CPU remained 'stuck' at lower frequencies.
To fix this, we must use a frequency above the current frequency to
get a chance to select a higher OPP when the current one becomes fully used.
Apply the same margin and return a frequency 25% higher than the current
one in order to switch to the next OPP before we fully use the CPU
at the current one.
[ mingo: Clarified the changelog. ]
Fixes: 9c0b4bb7f630 ("sched/cpufreq: Rework schedutil governor performance estimation")
Reported-by: Linus Torvalds <torvalds@...ux-foundation.org>
Bisected-by: Linus Torvalds <torvalds@...ux-foundation.org>
Reported-by: Wyes Karny <wkarny@...il.com>
Signed-off-by: Vincent Guittot <vincent.guittot@...aro.org>
Signed-off-by: Ingo Molnar <mingo@...nel.org>
Tested-by: Wyes Karny <wkarny@...il.com>
Link: https://lore.kernel.org/r/20240114183600.135316-1-vincent.guittot@linaro.org
---
kernel/sched/cpufreq_schedutil.c | 6 +++++-
1 file changed, 5 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/cpufreq_schedutil.c b/kernel/sched/cpufreq_schedutil.c
index 95c3c097083e..eece6244f9d2 100644
--- a/kernel/sched/cpufreq_schedutil.c
+++ b/kernel/sched/cpufreq_schedutil.c
@@ -133,7 +133,11 @@ unsigned long get_capacity_ref_freq(struct cpufreq_policy *policy)
if (arch_scale_freq_invariant())
return policy->cpuinfo.max_freq;
- return policy->cur;
+ /*
+ * Apply a 25% margin so that we select a higher frequency than
+ * the current one before the CPU is fully busy:
+ */
+ return policy->cur + (policy->cur >> 2);
}
/**
Powered by blists - more mailing lists