[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <000601d59c8f$9fa191c0$dee4b540$@net>
Date: Sat, 16 Nov 2019 07:07:49 -0800
From: "Doug Smythies" <dsmythies@...us.net>
To: "'Vincent Guittot'" <vincent.guittot@...aro.org>,
"'Peter Zijlstra'" <peterz@...radead.org>
Cc: "'linux-kernel'" <linux-kernel@...r.kernel.org>,
"'Ingo Molnar'" <mingo@...hat.com>,
"'Dietmar Eggemann'" <dietmar.eggemann@....com>,
"'Juri Lelli'" <juri.lelli@...hat.com>,
"'Steven Rostedt'" <rostedt@...dmis.org>,
"'Mel Gorman'" <mgorman@...e.de>,
"'open list:THERMAL'" <linux-pm@...r.kernel.org>,
"'Linus Torvalds'" <torvalds@...ux-foundation.org>,
"'Thomas Gleixner'" <tglx@...utronix.de>,
"'Sargun Dhillon'" <sargun@...gun.me>,
"'Tejun Heo'" <tj@...nel.org>, "'Xie XiuQi'" <xiexiuqi@...wei.com>,
<xiezhipeng1@...wei.com>,
"'Srinivas Pandruvada'" <srinivas.pandruvada@...ux.intel.com>
Subject: RE: [PATCH v4] sched/freq: move call to cpufreq_update_util
Hi,
I tested both Vincent's V4 and Peter's, I called it V1.
8000 seconds intel_pstate_tracer on a very idle system.
Everything was fine.
Vincent-V4:
Maximum time between calls to the intel_pstate driver,
per CPU: 4.00813 seconds: GOOD/PASS
Total entries/exits to/from driver: 90,730:
Consistent with previous tests / expectations.
228,600 aperf clocks / second / cpu.
Consistent with previous tests / expectations.
Peter-V1:
Maximum time between calls to the intel_pstate driver,
per CPU: 4.00407 seconds: GOOD/PASS
Total entries/exits to/from driver: 96,440:
Consistent with previous tests / expectations.
241,310 aperf clocks / second / cpu.
Consistent with previous tests / expectations.
Baseline reference:
Maximum time between calls to the intel_pstate driver,
per CPU: 225.03 seconds: BAD/FAIL
Number of durations over an arbitrary threshold of
10 seconds: 379.
Total entries/exits to/from driver: 75,969:
Consistent with previous tests / expectations,
when the issue is present.
226,963 aperf clocks / second / cpu.
Consistent with previous tests / expectations.
I did not do the load no-load energy test.
It is not possible for it to be a problem
if the long duration issue is solved.
I'll do the test, if someone wants the proof.
Kernels were 5.4-rc7 + linux next.
Idle governor was TEO, but this isn't actually relevant.
Tested-by: Doug Smythies <dsmythies@...us.net>
... Doug
Powered by blists - more mailing lists