linux-kernel - Re: [RFC] Add --show-total-period for perf annotate

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <5564685D.1020204@suse.cz>
Date:	Tue, 26 May 2015 14:34:37 +0200
From:	Martin Liška <mliska@...e.cz>
To:	Andi Kleen <andi@...stfloor.org>
CC:	linux-perf-users@...r.kernel.org, linux-kernel@...r.kernel.org,
	Arnaldo Carvalho de Melo <acme@...nel.org>
Subject: Re: [RFC] Add --show-total-period for perf annotate

On 05/25/2015 05:14 PM, Andi Kleen wrote:
>>> perf diff does not handle this? Especially with the differential
>>> profiling options it should.
>>
>> It does not work if you, in my case, compare ICC and GCC, where ICC uses a different mangling
>> scheme for fortran modules. Moreover, situation can be more complicated if a compiler performs
>> a bit different inlining decisions.
>
> I suppose it could be enhanced with an input file that describes
> equivalent functions. But yes wouldn't work for inlining.
>
>> Good point. Can you please help me how to compute a function percentage usage in perf annotate ;) ?
>
> I wouldn't use time at all. Just sum up periods and then compute the
> percentage. The period sum computation already happens in the main view, and
> is displayed there. So you only need to save that value somewhere and
> then use it in the annotate display for another column.
>
> # Samples: 24  of event 'cycles'
> # Event count (approx.): 8856637
>
> -Andi
>

Hello.

Are you talking about summing cycles and compute global percentage for each isntruction?

Anyway, attached patch is capable of displaying milliseconds approximation for each instruction.

Example:

time ./perf record ./a.out
[ perf record: Woken up 7 times to write data ]
[ perf record: Captured and wrote 1.743 MB perf.data (45386 samples) ]

real	0m11.465s
user	0m11.424s
sys	0m0.045s

$ perf report --stdio

     32.69%  a.out    libc-2.19.so       [.] __random_r
     26.66%  a.out    libc-2.19.so       [.] __random
     12.55%  a.out    a.out              [.] foo
     10.14%  a.out    a.out              [.] bar
      9.27%  a.out    a.out              [.] baz
      7.22%  a.out    libc-2.19.so       [.] rand
      1.37%  a.out    a.out              [.] rand@plt

$ perf annotate --stdio

  Percent |      Source code & Disassembly of libc-2.19.so for cycles
--------------------------------------------------------------------
          :
          :
          :
          :      Disassembly of section .text:
          :
          :      0000000000038890 <random_r>:
          :      __random_r():
    11.12 :        38890:       test   %rdi,%rdi


$ perf annotate --stdio --show-total-period
  Percent |      Source code & Disassembly of libc-2.19.so for cycles
--------------------------------------------------------------------
          :
          :
          :
          :      Disassembly of section .text:
          :
          :      0000000000038890 <random_r>:
          :      __random_r():
      413 :        38890:       test   %rdi,%rdi

First `test` instruction: 11465*0.3269*0.1112 ~ 413ms.

Thanks for suggestions,
Martin



View attachment "0001-First-semi-working-version.patch" of type "text/x-patch" (6031 bytes)