From: Morten Rasmussen <morten.rasmussen@arm.com> To: Peter Zijlstra <peterz@infradead.org> Cc: Vincent Guittot <vincent.guittot@linaro.org>, "mingo@kernel.org" <mingo@kernel.org>, "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>, "preeti@linux.vnet.ibm.com" <preeti@linux.vnet.ibm.com>, "linux@arm.linux.org.uk" <linux@arm.linux.org.uk>, "linux-arm-kernel@lists.infradead.org" <linux-arm-kernel@lists.infradead.org>, "riel@redhat.com" <riel@redhat.com>, "efault@gmx.de" <efault@gmx.de>, "nicolas.pitre@linaro.org" <nicolas.pitre@linaro.org>, "linaro-kernel@lists.linaro.org" <linaro-kernel@lists.linaro.org>, "daniel.lezcano@linaro.org" <daniel.lezcano@linaro.org>, Dietmar Eggemann <Dietmar.Eggemann@arm.com> Subject: Re: [PATCH v5 10/12] sched: get CPU's utilization statistic Date: Mon, 15 Sep 2014 20:28:50 +0100 [thread overview] Message-ID: <20140915192850.GB28083@e103687> (raw) In-Reply-To: <20140911123412.GY3190@worktop.ger.corp.intel.com> On Thu, Sep 11, 2014 at 01:34:12PM +0100, Peter Zijlstra wrote: > > @@ -4514,6 +4519,17 @@ static int select_idle_sibling(struct task_struct *p, int target) > > return target; > > } > > > > +static int get_cpu_utilization(int cpu) > > +{ > > + unsigned long usage = cpu_rq(cpu)->cfs.usage_load_avg; > > + unsigned long capacity = capacity_of(cpu); > > + > > + if (usage >= SCHED_LOAD_SCALE) > > + return capacity + 1; > > + > > + return (usage * capacity) >> SCHED_LOAD_SHIFT; > > +} > > So if I understood patch 9 correct, your changelog is iffy. > usage_load_avg should never get > 1 (of whatever unit), no matter how > many tasks are on the rq. You can only maximally run all the time. > > Therefore I can only interpret the if (usage >= SCHED_LOAD_SCALE) as > numerical error handling, nothing more. That is not entirely true unless you also classify transient usage spikes due to task migrations as numerical errors as well. Since each task sched_entity is carrying around 350ms worth of execution history with it between different cpus and cpu utilization is based on the sum of task entity usage_avg_contrib on the runqueue you may get cfs.usage_load_avg > 1 temporarily after task migrations. It will eventually converge to 1. The same goes for new tasks which are initialized to have a usage_avg_contrib of 1 and may be queued on cpu with tasks already running. In that case cfs.usage_load_avg is temporarily unbounded. > Also I'm not entirely sure I like the usage, utilization names/metrics. > I would suggest to reverse them. Call the pure running number > 'utilization' and this scaled with capacity 'usage' or so. I can agree with calling running for utilization, but I'm not convienced about capacity. What does it exactly cover here? I'm confused and jetlagged. Morten
WARNING: multiple messages have this Message-ID (diff)
From: morten.rasmussen@arm.com (Morten Rasmussen) To: linux-arm-kernel@lists.infradead.org Subject: [PATCH v5 10/12] sched: get CPU's utilization statistic Date: Mon, 15 Sep 2014 20:28:50 +0100 [thread overview] Message-ID: <20140915192850.GB28083@e103687> (raw) In-Reply-To: <20140911123412.GY3190@worktop.ger.corp.intel.com> On Thu, Sep 11, 2014 at 01:34:12PM +0100, Peter Zijlstra wrote: > > @@ -4514,6 +4519,17 @@ static int select_idle_sibling(struct task_struct *p, int target) > > return target; > > } > > > > +static int get_cpu_utilization(int cpu) > > +{ > > + unsigned long usage = cpu_rq(cpu)->cfs.usage_load_avg; > > + unsigned long capacity = capacity_of(cpu); > > + > > + if (usage >= SCHED_LOAD_SCALE) > > + return capacity + 1; > > + > > + return (usage * capacity) >> SCHED_LOAD_SHIFT; > > +} > > So if I understood patch 9 correct, your changelog is iffy. > usage_load_avg should never get > 1 (of whatever unit), no matter how > many tasks are on the rq. You can only maximally run all the time. > > Therefore I can only interpret the if (usage >= SCHED_LOAD_SCALE) as > numerical error handling, nothing more. That is not entirely true unless you also classify transient usage spikes due to task migrations as numerical errors as well. Since each task sched_entity is carrying around 350ms worth of execution history with it between different cpus and cpu utilization is based on the sum of task entity usage_avg_contrib on the runqueue you may get cfs.usage_load_avg > 1 temporarily after task migrations. It will eventually converge to 1. The same goes for new tasks which are initialized to have a usage_avg_contrib of 1 and may be queued on cpu with tasks already running. In that case cfs.usage_load_avg is temporarily unbounded. > Also I'm not entirely sure I like the usage, utilization names/metrics. > I would suggest to reverse them. Call the pure running number > 'utilization' and this scaled with capacity 'usage' or so. I can agree with calling running for utilization, but I'm not convienced about capacity. What does it exactly cover here? I'm confused and jetlagged. Morten
next prev parent reply other threads:[~2014-09-15 19:29 UTC|newest] Thread overview: 164+ messages / expand[flat|nested] mbox.gz Atom feed top 2014-08-26 11:06 [PATCH v5 00/12] sched: consolidation of cpu_capacity Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 01/12] sched: fix imbalance flag reset Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-19 11:47 ` [tip:sched/core] sched: Fix " tip-bot for Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 02/12] sched: remove a wake_affine condition Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-19 11:47 ` [tip:sched/core] sched: Remove a wake_affine() condition tip-bot for Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 03/12] sched: fix avg_load computation Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-08-30 12:00 ` Preeti U Murthy 2014-08-30 12:00 ` Preeti U Murthy 2014-09-03 11:09 ` Vincent Guittot 2014-09-03 11:09 ` Vincent Guittot 2014-09-03 23:43 ` Tim Chen 2014-09-03 23:43 ` Tim Chen 2014-09-04 7:17 ` Vincent Guittot 2014-09-04 7:17 ` Vincent Guittot 2014-09-04 16:26 ` Tim Chen 2014-09-04 16:26 ` Tim Chen 2014-09-05 11:10 ` Preeti U Murthy 2014-09-05 11:10 ` Preeti U Murthy 2014-09-19 11:47 ` [tip:sched/core] sched: Fix " tip-bot for Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 04/12] sched: Allow all archs to set the capacity_orig Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-08-27 13:12 ` Kamalesh Babulal 2014-08-27 13:12 ` Kamalesh Babulal 2014-08-30 17:07 ` Preeti U Murthy 2014-08-30 17:07 ` Preeti U Murthy 2014-09-01 8:05 ` Vincent Guittot 2014-09-01 8:05 ` Vincent Guittot 2014-09-03 8:41 ` Preeti U Murthy 2014-09-03 8:41 ` Preeti U Murthy 2014-09-10 13:50 ` Peter Zijlstra 2014-09-10 13:50 ` Peter Zijlstra 2014-09-10 14:22 ` Vincent Guittot 2014-09-10 14:22 ` Vincent Guittot 2014-09-11 10:36 ` Preeti U Murthy 2014-09-11 10:36 ` Preeti U Murthy 2014-09-19 11:47 ` [tip:sched/core] sched: Allow all architectures to set ' capacity_orig' tip-bot for Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 05/12] ARM: topology: use new cpu_capacity interface Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-11 18:52 ` Nicolas Pitre 2014-09-11 18:52 ` Nicolas Pitre 2014-09-19 11:48 ` [tip:sched/core] ARM: topology: Use the " tip-bot for Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 06/12] sched: add per rq cpu_capacity_orig Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-08-27 13:32 ` Kamalesh Babulal 2014-08-27 13:32 ` Kamalesh Babulal 2014-08-28 7:34 ` Vincent Guittot 2014-08-28 7:34 ` Vincent Guittot 2014-09-10 13:53 ` Peter Zijlstra 2014-09-10 13:53 ` Peter Zijlstra 2014-09-10 14:19 ` Vincent Guittot 2014-09-10 14:19 ` Vincent Guittot 2014-09-11 19:02 ` Nicolas Pitre 2014-09-11 19:02 ` Nicolas Pitre 2014-09-15 21:22 ` Vincent Guittot 2014-09-15 21:22 ` Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 07/12] sched: test the cpu's capacity in wake affine Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-10 14:19 ` Peter Zijlstra 2014-09-10 14:19 ` Peter Zijlstra 2014-09-19 11:48 ` [tip:sched/core] sched: Test the CPU's capacity in wake_affine() tip-bot for Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 08/12] sched: move cfs task on a CPU with higher capacity Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-08-30 17:50 ` Preeti U Murthy 2014-08-30 17:50 ` Preeti U Murthy 2014-09-01 8:45 ` Vincent Guittot 2014-09-01 8:45 ` Vincent Guittot 2014-09-03 9:11 ` Preeti U Murthy 2014-09-03 9:11 ` Preeti U Murthy 2014-09-03 11:44 ` Vincent Guittot 2014-09-03 11:44 ` Vincent Guittot 2014-09-03 12:26 ` Preeti U Murthy 2014-09-03 12:26 ` Preeti U Murthy 2014-09-03 12:49 ` Vincent Guittot 2014-09-03 12:49 ` Vincent Guittot 2014-09-11 9:27 ` Peter Zijlstra 2014-09-11 9:27 ` Peter Zijlstra 2014-09-05 12:06 ` Preeti U Murthy 2014-09-05 12:06 ` Preeti U Murthy 2014-09-05 12:24 ` Vincent Guittot 2014-09-05 12:24 ` Vincent Guittot 2014-09-11 10:07 ` Peter Zijlstra 2014-09-11 10:07 ` Peter Zijlstra 2014-09-11 11:20 ` Vincent Guittot 2014-09-11 11:20 ` Vincent Guittot 2014-09-11 10:13 ` Peter Zijlstra 2014-09-11 10:13 ` Peter Zijlstra 2014-09-11 12:14 ` Vincent Guittot 2014-09-11 12:14 ` Vincent Guittot 2014-09-11 11:54 ` Peter Zijlstra 2014-09-11 11:54 ` Peter Zijlstra 2014-08-26 11:06 ` [PATCH v5 09/12] sched: add usage_load_avg Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-04 7:34 ` [PATCH v5 09/11] " Vincent Guittot 2014-09-04 7:34 ` Vincent Guittot 2014-09-11 11:17 ` Peter Zijlstra 2014-09-11 11:17 ` Peter Zijlstra 2014-09-11 11:17 ` [PATCH v5 09/12] " Peter Zijlstra 2014-09-11 11:17 ` Peter Zijlstra 2014-09-11 12:18 ` Vincent Guittot 2014-09-11 12:18 ` Vincent Guittot 2014-09-11 12:20 ` Vincent Guittot 2014-09-11 12:20 ` Vincent Guittot 2014-09-15 19:15 ` Morten Rasmussen 2014-09-15 19:15 ` Morten Rasmussen 2014-09-15 22:33 ` Vincent Guittot 2014-09-15 22:33 ` Vincent Guittot 2014-08-26 11:06 ` [PATCH v5 10/12] sched: get CPU's utilization statistic Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-11 12:34 ` Peter Zijlstra 2014-09-11 12:34 ` Peter Zijlstra 2014-09-11 13:07 ` Vincent Guittot 2014-09-11 13:07 ` Vincent Guittot 2014-09-11 14:04 ` Peter Zijlstra 2014-09-11 14:04 ` Peter Zijlstra 2014-09-11 19:17 ` Nicolas Pitre 2014-09-11 19:17 ` Nicolas Pitre 2014-09-12 7:41 ` Vincent Guittot 2014-09-12 7:41 ` Vincent Guittot 2014-09-15 19:45 ` Morten Rasmussen 2014-09-15 19:45 ` Morten Rasmussen 2014-09-16 22:43 ` Vincent Guittot 2014-09-16 22:43 ` Vincent Guittot 2014-09-15 19:28 ` Morten Rasmussen [this message] 2014-09-15 19:28 ` Morten Rasmussen 2014-08-26 11:06 ` [PATCH v5 11/12] sched: replace capacity_factor by utilization Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot 2014-09-11 15:39 ` Peter Zijlstra 2014-09-11 15:39 ` Peter Zijlstra 2014-09-11 16:15 ` Peter Zijlstra 2014-09-11 16:15 ` Peter Zijlstra 2014-09-11 17:26 ` Vincent Guittot 2014-09-11 17:26 ` Vincent Guittot 2014-09-14 19:41 ` Peter Zijlstra 2014-09-14 19:41 ` Peter Zijlstra 2014-09-14 19:51 ` Peter Zijlstra 2014-09-14 19:51 ` Peter Zijlstra 2014-09-15 11:42 ` Peter Zijlstra 2014-09-15 11:42 ` Peter Zijlstra 2014-09-15 19:07 ` Nicolas Pitre 2014-09-15 19:07 ` Nicolas Pitre 2014-09-15 20:01 ` Peter Zijlstra 2014-09-15 20:01 ` Peter Zijlstra 2014-09-17 18:45 ` Morten Rasmussen 2014-09-17 18:45 ` Morten Rasmussen 2014-09-17 18:58 ` Morten Rasmussen 2014-09-17 18:58 ` Morten Rasmussen 2014-09-17 23:03 ` Peter Zijlstra 2014-09-17 23:03 ` Peter Zijlstra 2014-09-15 22:14 ` Vincent Guittot 2014-09-15 22:14 ` Vincent Guittot 2014-09-15 22:18 ` Vincent Guittot 2014-09-15 22:18 ` Vincent Guittot 2014-09-17 22:25 ` Peter Zijlstra 2014-09-17 22:25 ` Peter Zijlstra 2014-09-18 1:32 ` Vincent Guittot 2014-09-18 1:32 ` Vincent Guittot 2014-09-16 17:00 ` Dietmar Eggemann 2014-09-16 17:00 ` Dietmar Eggemann 2014-08-26 11:06 ` [PATCH v5 12/12] sched: add SD_PREFER_SIBLING for SMT level Vincent Guittot 2014-08-26 11:06 ` Vincent Guittot
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=20140915192850.GB28083@e103687 \ --to=morten.rasmussen@arm.com \ --cc=Dietmar.Eggemann@arm.com \ --cc=daniel.lezcano@linaro.org \ --cc=efault@gmx.de \ --cc=linaro-kernel@lists.linaro.org \ --cc=linux-arm-kernel@lists.infradead.org \ --cc=linux-kernel@vger.kernel.org \ --cc=linux@arm.linux.org.uk \ --cc=mingo@kernel.org \ --cc=nicolas.pitre@linaro.org \ --cc=peterz@infradead.org \ --cc=preeti@linux.vnet.ibm.com \ --cc=riel@redhat.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: linkBe sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.