From: Andi Kleen <andi@firstfloor.org>
To: "Liang, Kan" <kan.liang@intel.com>
Cc: Andi Kleen <andi@firstfloor.org>,
"peterz@infradead.org" <peterz@infradead.org>,
"mingo@redhat.com" <mingo@redhat.com>,
"acme@kernel.org" <acme@kernel.org>,
"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
"alexander.shishkin@linux.intel.com"
<alexander.shishkin@linux.intel.com>,
"tglx@linutronix.de" <tglx@linutronix.de>,
"namhyung@kernel.org" <namhyung@kernel.org>,
"jolsa@kernel.org" <jolsa@kernel.org>,
"Hunter, Adrian" <adrian.hunter@intel.com>,
"wangnan0@huawei.com" <wangnan0@huawei.com>,
"mark.rutland@arm.com" <mark.rutland@arm.com>
Subject: Re: [PATCH 13/14] perf tools: warn on high overhead
Date: Fri, 25 Nov 2016 12:42:33 -0800 [thread overview]
Message-ID: <20161125204233.GG26852@two.firstfloor.org> (raw)
In-Reply-To: <37D7C6CF3E00A74B8858931C1DB2F07750CA27D8@SHSMSX103.ccr.corp.intel.com>
On Wed, Nov 23, 2016 at 10:03:24PM +0000, Liang, Kan wrote:
> > Perhaps we need two separate metrics here:
> >
> > - cost of perf record on its CPU (or later on if it gets multi threaded
> > more multiple). Warn if this is >50% or so.
>
> What's the formula for cost of perf record on its CPU?
> The cost only includes user space overhead or all overhead?
> What is the divisor?
It would be all the overhead in the process. Accounting overhead in
kernel threads or interrupts caused by IO is difficult, we could leave
that out for now.
Sum of:
For each perf thread:
thread cpu time / monotonic wall time
I guess Sum is better than average here because the perf threads are
likely running (or could be) on the same CPU. If perf record was changed to be
more aggressively flush buffers on the local CPUs this would need to
change, but I presume it's good enough for now.
>
>
> > - average perf collection overhead on a CPU. The 10% threshold here
> > seems appropiate.
> For the average, do you mean add all overheads among CPUs together
> and divide the CPU#?
Right. Possibly also max of all too.
>
> To calculate the rate, the divisor is wall clock time, right?
monotonic wall clock time yes.
-Andi
next prev parent reply other threads:[~2016-11-25 20:42 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-11-23 9:44 [PATCH 00/14] export perf overheads information kan.liang
2016-11-23 9:44 ` [PATCH 01/14] perf/x86: Introduce PERF_RECORD_OVERHEAD kan.liang
2016-11-23 20:11 ` Peter Zijlstra
2016-11-23 20:13 ` Peter Zijlstra
2016-11-23 23:41 ` Jiri Olsa
2016-11-24 13:45 ` Liang, Kan
2016-11-24 13:50 ` Peter Zijlstra
2016-11-24 13:56 ` Liang, Kan
2016-11-24 14:27 ` Jiri Olsa
2016-11-24 14:39 ` Liang, Kan
2016-11-24 14:47 ` Jiri Olsa
2016-11-24 18:28 ` Andi Kleen
2016-11-24 18:58 ` Peter Zijlstra
2016-11-24 19:02 ` Andi Kleen
2016-11-24 19:08 ` Peter Zijlstra
2016-11-23 9:44 ` [PATCH 02/14] perf/x86: output NMI overhead kan.liang
2016-11-23 20:06 ` Peter Zijlstra
2016-11-24 16:19 ` Mark Rutland
2016-11-24 19:02 ` Peter Zijlstra
2016-11-24 19:40 ` Liang, Kan
2016-11-24 23:26 ` Namhyung Kim
2016-11-23 9:44 ` [PATCH 03/14] perf/x86: output multiplexing overhead kan.liang
2016-11-23 20:06 ` Peter Zijlstra
2016-11-23 20:09 ` Liang, Kan
2016-11-23 9:44 ` [PATCH 04/14] perf/x86: output side-band events overhead kan.liang
2016-11-23 20:06 ` Peter Zijlstra
2016-11-24 16:21 ` Mark Rutland
2016-11-24 19:40 ` Liang, Kan
2016-11-23 9:44 ` [PATCH 05/14] perf tools: handle PERF_RECORD_OVERHEAD record type kan.liang
2016-11-23 22:35 ` Jiri Olsa
2016-11-23 22:58 ` Jiri Olsa
2016-11-23 9:44 ` [PATCH 06/14] perf tools: show NMI overhead kan.liang
2016-11-23 22:51 ` Jiri Olsa
2016-11-24 13:37 ` Liang, Kan
2016-11-24 15:27 ` Jiri Olsa
2016-11-24 23:20 ` Namhyung Kim
2016-11-24 23:45 ` Jiri Olsa
2016-11-25 0:21 ` Andi Kleen
2016-11-23 22:52 ` Jiri Olsa
2016-11-23 22:52 ` Jiri Olsa
2016-11-23 9:44 ` [PATCH 07/14] perf tools: show multiplexing overhead kan.liang
2016-11-23 9:44 ` [PATCH 08/14] perf tools: show side-band events overhead kan.liang
2016-11-23 9:44 ` [PATCH 09/14] perf tools: make get_nsecs visible for buildin files kan.liang
2016-11-23 9:44 ` [PATCH 10/14] perf tools: introduce PERF_RECORD_USER_OVERHEAD kan.liang
2016-11-23 9:44 ` [PATCH 11/14] perf tools: record write data overhead kan.liang
2016-11-23 23:02 ` Jiri Olsa
2016-11-23 23:06 ` Jiri Olsa
2016-11-23 9:44 ` [PATCH 12/14] perf tools: record elapsed time kan.liang
2016-11-23 9:44 ` [PATCH 13/14] perf tools: warn on high overhead kan.liang
2016-11-23 20:25 ` Andi Kleen
2016-11-23 22:03 ` Liang, Kan
2016-11-25 20:42 ` Andi Kleen [this message]
2016-11-23 9:44 ` [PATCH 14/14] perf script: show overhead events kan.liang
2016-11-23 23:17 ` Jiri Olsa
2016-11-23 23:18 ` Jiri Olsa
2016-11-23 23:19 ` Jiri Olsa
2016-11-23 23:22 ` Jiri Olsa
2016-11-24 4:27 ` [PATCH 00/14] export perf overheads information Ingo Molnar
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=20161125204233.GG26852@two.firstfloor.org \
--to=andi@firstfloor.org \
--cc=acme@kernel.org \
--cc=adrian.hunter@intel.com \
--cc=alexander.shishkin@linux.intel.com \
--cc=jolsa@kernel.org \
--cc=kan.liang@intel.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mingo@redhat.com \
--cc=namhyung@kernel.org \
--cc=peterz@infradead.org \
--cc=tglx@linutronix.de \
--cc=wangnan0@huawei.com \
/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;
as well as URLs for NNTP newsgroup(s).