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: <11256670.HWKrRQARBr@agathebauer>
Date:   Mon, 08 May 2017 10:45:18 +0200
From:   Milian Wolff <milian.wolff@...b.com>
To:     Arnaldo Carvalho de Melo <acme@...hat.com>
Cc:     linux-kernel@...r.kernel.org, linux-perf-users@...r.kernel.org,
        David Ahern <dsahern@...il.com>,
        Namhyung Kim <namhyung@...nel.org>,
        Peter Zijlstra <a.p.zijlstra@...llo.nl>,
        Yao Jin <yao.jin@...ux.intel.com>
Subject: Re: [PATCH v2] perf report: distinguish between inliners in the same function

On Mittwoch, 3. Mai 2017 23:35:36 CEST Milian Wolff wrote:
> When different functions get inlined into the same function, we
> want to show them individually in the reports. But when we group by
> function, we would aggregate all IPs and would only keep the first
> one in that function. E.g. for C++ code like the following:
> 
> ~~~~~
> #include <cmath>
> #include <random>
> #include <iostream>
> 
> using namespace std;
> 
> int main()
> {
>     uniform_real_distribution<double> uniform(-1E5, 1E5);
>     default_random_engine engine;
>     double s = 0;
>     for (int i = 0; i < 10000000; ++i) {
>         s += uniform(engine);
>     }
>     cout << "random sum: " << s << '\n';
>     return 0;
> }
> ~~~~~
> 
> Building it with `g++ -O2 -g` and recording some samples with
> `perf record --call-graph dwarf` yields for me:
> 
> ~~~~~
> $ perf report --stdio --inline
> # Overhead  Command    Shared Object      Symbol
> # ........  .........  .................  ................................
> #
>     99.40%    99.11%  a.out    a.out                [.] main
> 
>              --99.11%--_start
>                        __libc_start_main
>                        main
> ...
> ~~~~~
> 
> Note how no inlined frames are actually shown, because the first
> sample in main points to an IP that does not correspond to any
> inlined frames.
> 
> With this patch applied, we instead get the following, much more
> meaningful, reports.
> 
> ~~~~~
> $ perf report --stdio --inline --no-children
> # Overhead  Command    Shared Object      Symbol
> # ........  .........  .................  ................................
> #
>     99.11%  a.out    a.out             [.] main
> 
>             |--48.15%--main
>             |
>             |          std::__detail::_Adaptor<std::linear_congruential_engi
>             |          ne<unsigned long, 16807ul, 0ul, 2147483647ul>,
>             |          double>::operator() (inline)
>             |          std::uniform_real_distribution<double>::operator()<s
>             |          td::linear_congruential_engine<unsigned long,
>             |          16807ul, 0ul, 2147483647ul> > (inline)
>             |          std::uniform_real_distribution<double>::operator()<s
>             |          td::linear_congruential_engine<unsigned long,
>             |          16807ul, 0ul, 2147483647ul> > (inline) main (inline)
>             |          __libc_start_main
>             |          _start
>             |
>             |--47.61%--main
>             |
>             |          std::__detail::__mod<unsigned long, 2147483647ul,
>             |          16807ul, 0ul> (inline)
>             |          std::linear_congruential_engine<unsigned long,
>             |          16807ul, 0ul, 2147483647ul>::operator() (inline)
>             |          std::generate_canonical<double, 53ul,
>             |          std::linear_congruential_engine<unsigned long,
>             |          16807ul, 0ul, 2147483647ul> > (inline)
>             |          std::__detail::_Adaptor<std::linear_congruential_eng
>             |          ine<unsigned long, 16807ul, 0ul, 2147483647ul>,
>             |          double>::operator() (inline)
>             |          std::uniform_real_distribution<double>::operator()<s
>             |          td::linear_congruential_engine<unsigned long,
>             |          16807ul, 0ul, 2147483647ul> > (inline)
>             |          std::uniform_real_distribution<double>::operator()<s
>             |          td::linear_congruential_engine<unsigned long,
>             |          16807ul, 0ul, 2147483647ul> > (inline) main (inline)
>             |          __libc_start_main
>             |          _start
> 
>              --3.35%--main
>                       
> std::uniform_real_distribution<double>::operator()<std::linear_congruential
> _engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inline) main (inline)
>                        __libc_start_main
>                        _start
> ...
> 
> $ perf report --stdio --inline
> # Children      Self  Command  Shared Object        Symbol
> # ........  ........  .......  ................... 
> ........................................................ #
>     99.40%    99.11%  a.out    a.out                [.] main
> 
>              --99.11%--_start
>                        __libc_start_main
> 
>                        |--70.51%--main
>                        |
>                        |          main (inline)
>                        |          std::uniform_real_distribution<double>::op
>                        |          erator()<std::linear_congruential_engine<u
>                        |          nsigned long, 16807ul, 0ul, 2147483647ul>
>                        |          > (inline)
>                        |          std::uniform_real_distribution<double>::o
>                        |          perator()<std::linear_congruential_engine<
>                        |          unsigned long, 16807ul, 0ul, 2147483647ul>
>                        |          > (inline)
>                        |          std::__detail::_Adaptor<std::linear_congr
>                        |          uential_engine<unsigned long, 16807ul,
>                        |          0ul, 2147483647ul>, double>::operator()
>                        |          (inline)                       |
>                        |--25.25%--main
>                        |
>                        |          main (inline)
>                        |          std::uniform_real_distribution<double>::op
>                        |          erator()<std::linear_congruential_engine<u
>                        |          nsigned long, 16807ul, 0ul, 2147483647ul>
>                        |          > (inline)
>                        |          std::uniform_real_distribution<double>::o
>                        |          perator()<std::linear_congruential_engine<
>                        |          unsigned long, 16807ul, 0ul, 2147483647ul>
>                        |          > (inline)
>                        |          std::__detail::_Adaptor<std::linear_congr
>                        |          uential_engine<unsigned long, 16807ul,
>                        |          0ul, 2147483647ul>, double>::operator()
>                        |          (inline) std::generate_canonical<double,
>                        |          53ul,
>                        |          std::linear_congruential_engine<unsigned
>                        |          long, 16807ul, 0ul, 2147483647ul> >
>                        |          (inline)
>                        |          std::linear_congruential_engine<unsigned
>                        |          long, 16807ul, 0ul,
>                        |          2147483647ul>::operator() (inline)
>                        |          std::__detail::__mod<unsigned long,
>                        |          2147483647ul, 16807ul, 0ul> (inline)
>                         --3.35%--main
>                                   main (inline)
>                                  
> std::uniform_real_distribution<double>::operator()<std::linear_congruential
> _engine<unsigned long, 16807ul, 0ul, 2147483647ul> > (inline) ~~~~~
> 
> Note that the latter top-down call graphs require us to traverse the
> full inliner call chain when comparing for equality. Without it,
> we would potentially combine the three distinct branches again, since
> they share an equal first inline frame "main (inline)".

Ping? Any chance that I could get a review on this one please? It works really 
well for me and greatly improves perf's usability for C++ code bases.

Thanks
-- 
Milian Wolff | milian.wolff@...b.com | Software Engineer
KDAB (Deutschland) GmbH&Co KG, a KDAB Group company
Tel: +49-30-521325470
KDAB - The Qt Experts
Download attachment "smime.p7s" of type "application/pkcs7-signature" (5903 bytes)

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ