From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jiri Olsa <jolsa@redhat.com>,
linux-kernel@vger.kernel.org,
Arnaldo Carvalho de Melo <acme@ghostprotocols.net>,
Ingo Molnar <mingo@elte.hu>, Paul Mackerras <paulus@samba.org>,
Corey Ashford <cjashfor@linux.vnet.ibm.com>,
Frederic Weisbecker <fweisbec@gmail.com>,
Andi Kleen <andi@firstfloor.org>, David Ahern <dsahern@gmail.com>,
Namhyung Kim <namhyung@kernel.org>
Subject: Re: [RFC 00/12] perf diff: Factor diff command
Date: Thu, 6 Sep 2012 14:25:57 -0700 [thread overview]
Message-ID: <20120906212557.GD2448@linux.vnet.ibm.com> (raw)
In-Reply-To: <1346956869.18408.66.camel@twins>
On Thu, Sep 06, 2012 at 08:41:09PM +0200, Peter Zijlstra wrote:
> On Thu, 2012-09-06 at 17:46 +0200, Jiri Olsa wrote:
> > The 'perf diff' and 'std/hist' code is now changed to allow computations
> > mentioned in the paper. Two of them are implemented within this patchset:
> > 1) ratio differential profiling
> > 2) weighted differential profiling
>
> Seems like a useful thing indeed, the explanation of the weighted diff
> method doesn't seem to contain a why. I know I could go read the paper
> but... :-)
Or you could ask the author. ;-)
Ratio can be fooled by statistical variations on profiling buckets with
few counts. So if you are looking for a 10% difference in execution
overhead somewhere in a large program, ratio will unhelpfully sort a
bunch of statistical 2x or 3x noise to the top of the list.
So you could use the difference in buckets instead of the ratio, but this
has problems in the case where the two runs being compared got different
amounts of work done, as is usually the case for timed benchmark runs
or throughput-based benchmark runs. In these cases, you use the work
done (or the measured throughput, as the case may be) as weights.
The weighted difference will then pinpoint the code that suffered the
greatest per-unit-work increase in overhead between the two runs.
But ratio is still the method of choice when you are looking at pure
scalability issues, especially when the offending code increased in
overhead by a large factor.
Thanx, Paul
next prev parent reply other threads:[~2012-09-06 21:27 UTC|newest]
Thread overview: 40+ messages / expand[flat|nested] mbox.gz Atom feed top
2012-09-06 15:46 [RFC 00/12] perf diff: Factor diff command Jiri Olsa
2012-09-06 15:46 ` [PATCH 01/12] perf diff: Make diff command work with evsel hists Jiri Olsa
2012-09-08 11:41 ` [tip:perf/core] " tip-bot for Jiri Olsa
2012-09-06 15:46 ` [PATCH 02/12] perf tools: Replace sort's standalone field_sep with symbol_conf.field_sep Jiri Olsa
2012-09-08 11:42 ` [tip:perf/core] perf tools: Replace sort' s " tip-bot for Jiri Olsa
2012-09-06 15:46 ` [PATCH 03/12] perf hists: Add struct hists pointer to struct hist_entry Jiri Olsa
2012-09-06 15:46 ` [PATCH 04/12] perf diff: Refactor diff displacement possition info Jiri Olsa
2012-09-08 0:56 ` Arnaldo Carvalho de Melo
2012-09-06 15:46 ` [PATCH 05/12] perf diff: Refactor stdio ui data columns output Jiri Olsa
2012-09-07 2:55 ` Namhyung Kim
2012-09-07 9:20 ` Jiri Olsa
2012-09-08 12:35 ` Jiri Olsa
2012-09-08 12:50 ` Arnaldo Carvalho de Melo
2012-09-08 14:37 ` Namhyung Kim
2012-09-08 15:10 ` Arnaldo Carvalho de Melo
2012-09-08 15:12 ` Arnaldo Carvalho de Melo
2012-09-08 15:21 ` Arnaldo Carvalho de Melo
2012-09-06 15:47 ` [PATCH 06/12] perf diff: Add -b option for perf diff to display paired entries only Jiri Olsa
2012-09-06 15:47 ` [PATCH 07/12] perf diff: Add ratio computation way to compare hist entries Jiri Olsa
2012-09-07 5:45 ` Namhyung Kim
2012-09-07 9:26 ` Jiri Olsa
2012-09-07 15:33 ` Arnaldo Carvalho de Melo
2012-09-07 15:41 ` Namhyung Kim
2012-09-06 15:47 ` [PATCH 08/12] perf diff: Add option to sort entries based on diff computation Jiri Olsa
2012-09-06 15:47 ` [PATCH 09/12] perf diff: Add weighted diff computation way to compare hist entries Jiri Olsa
2012-09-07 5:58 ` Namhyung Kim
2012-09-07 9:28 ` Jiri Olsa
2012-09-07 13:33 ` Namhyung Kim
2012-09-07 15:26 ` Peter Zijlstra
2012-09-07 15:31 ` Arnaldo Carvalho de Melo
2012-09-07 16:08 ` Peter Zijlstra
2012-09-06 15:47 ` [PATCH 10/12] perf diff: Add -p option to display period values for " Jiri Olsa
2012-09-06 15:47 ` [PATCH 11/12] perf diff: Add -F option to display formula for computation Jiri Olsa
2012-09-07 6:02 ` Namhyung Kim
2012-09-07 9:30 ` Jiri Olsa
2012-09-06 15:47 ` [PATCH 12/12] perf diff: Add -F option for ratio computation Jiri Olsa
2012-09-06 17:31 ` [RFC 00/12] perf diff: Factor diff command Jiri Olsa
2012-09-06 18:41 ` Peter Zijlstra
2012-09-06 21:25 ` Paul E. McKenney [this message]
2012-09-07 7:05 ` Peter Zijlstra
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=20120906212557.GD2448@linux.vnet.ibm.com \
--to=paulmck@linux.vnet.ibm.com \
--cc=a.p.zijlstra@chello.nl \
--cc=acme@ghostprotocols.net \
--cc=andi@firstfloor.org \
--cc=cjashfor@linux.vnet.ibm.com \
--cc=dsahern@gmail.com \
--cc=fweisbec@gmail.com \
--cc=jolsa@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=namhyung@kernel.org \
--cc=paulus@samba.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;
as well as URLs for NNTP newsgroup(s).