From: "Jin, Yao" <yao.jin@linux.intel.com>
To: Jiri Olsa <jolsa@redhat.com>
Cc: acme@kernel.org, jolsa@kernel.org, peterz@infradead.org,
mingo@redhat.com, alexander.shishkin@linux.intel.com,
Linux-kernel@vger.kernel.org, ak@linux.intel.com,
kan.liang@intel.com, yao.jin@intel.com
Subject: Re: [PATCH v2] perf diff: Report noisy for cycles diff
Date: Tue, 6 Aug 2019 19:46:38 +0800 [thread overview]
Message-ID: <3efff55b-511c-0748-432c-e7ef4718c506@linux.intel.com> (raw)
In-Reply-To: <20190806083421.GH7695@krava>
On 8/6/2019 4:34 PM, Jiri Olsa wrote:
> On Thu, Jul 25, 2019 at 06:14:32AM +0800, Jin Yao wrote:
>> This patch prints the stddev and hist for the cycles diff of
>> program block. It can help us to understand if the cycles diff
>> is noisy or not.
>>
>> This patch is inspired by Andi Kleen's patch
>> https://lwn.net/Articles/600471/
>>
>> We create new option '-n or --noisy'.
>>
>> Example:
>>
>> perf record -b ./div
>> perf record -b ./div
>> perf diff -c cycles
>>
>> # Event 'cycles'
>> #
>> # Baseline [Program Block Range] Cycles Diff Shared Object Symbol
>> # ........ ...................................................................... ................. ................................
>> #
>> 46.42% [div.c:40 -> div.c:40] 0 div [.] main
>> 46.42% [div.c:42 -> div.c:44] 0 div [.] main
>> 46.42% [div.c:42 -> div.c:39] 0 div [.] main
>> 20.72% [random_r.c:357 -> random_r.c:394] -2 libc-2.27.so [.] __random_r
>> 20.72% [random_r.c:357 -> random_r.c:380] -1 libc-2.27.so [.] __random_r
>> 20.72% [random_r.c:388 -> random_r.c:388] 0 libc-2.27.so [.] __random_r
>> 20.72% [random_r.c:388 -> random_r.c:391] 0 libc-2.27.so [.] __random_r
>> 17.58% [random.c:288 -> random.c:291] 0 libc-2.27.so [.] __random
>> 17.58% [random.c:291 -> random.c:291] 0 libc-2.27.so [.] __random
>> 17.58% [random.c:293 -> random.c:293] 0 libc-2.27.so [.] __random
>> 17.58% [random.c:295 -> random.c:295] 0 libc-2.27.so [.] __random
>> 17.58% [random.c:295 -> random.c:295] 0 libc-2.27.so [.] __random
>> 17.58% [random.c:298 -> random.c:298] 0 libc-2.27.so [.] __random
>> 8.33% [div.c:22 -> div.c:25] 0 div [.] compute_flag
>> 8.33% [div.c:27 -> div.c:28] 0 div [.] compute_flag
>> 4.80% [rand.c:26 -> rand.c:27] 0 libc-2.27.so [.] rand
>> 4.80% [rand.c:28 -> rand.c:28] 0 libc-2.27.so [.] rand
>> 2.14% [rand@plt+0 -> rand@plt+0] 0 div [.] rand@plt
>>
>> When we enable the option '-n' or '--noisy', the output is
>>
>> perf diff -c cycles -n
>>
>> # Event 'cycles'
>> #
>> # Baseline [Program Block Range]/Cycles Diff/stddev/Hist Shared Object Symbol
>> # ........ ................................................................................ ................. ................................
>> #
>> 46.42% [div.c:40 -> div.c:40] 0 ± 40.2% ▂███▁▂▁▁ div [.] main
>> 46.42% [div.c:42 -> div.c:44] 0 ±100.0% ▁▁▁▁█▁▁▁ div [.] main
>> 46.42% [div.c:42 -> div.c:39] 0 ± 15.3% ▃▃▂▆▃▂█▁ div [.] main
>> 20.72% [random_r.c:357 -> random_r.c:394] -2 ± 20.1% ▁▄▄▅▂▅█▁ libc-2.27.so [.] __random_r
>> 20.72% [random_r.c:357 -> random_r.c:380] -1 ± 20.9% ▁▆▇▁█▅▇█ libc-2.27.so [.] __random_r
>> 20.72% [random_r.c:388 -> random_r.c:388] 0 ± 0.0% libc-2.27.so [.] __random_r
>> 20.72% [random_r.c:388 -> random_r.c:391] 0 ± 88.0% ▁▁▁▁▁▁▁█ libc-2.27.so [.] __random_r
>> 17.58% [random.c:288 -> random.c:291] 0 ± 29.3% ▁████▁█▁ libc-2.27.so [.] __random
>> 17.58% [random.c:291 -> random.c:291] 0 ± 29.3% ▁████▁▁█ libc-2.27.so [.] __random
>> 17.58% [random.c:293 -> random.c:293] 0 ± 29.3% ▁████▁▁█ libc-2.27.so [.] __random
>> 17.58% [random.c:295 -> random.c:295] 0 ± 0.0% libc-2.27.so [.] __random
>> 17.58% [random.c:295 -> random.c:295] 0 ± 0.0% libc-2.27.so [.] __random
>> 17.58% [random.c:298 -> random.c:298] 0 ± 0.0% libc-2.27.so [.] __random
>> 8.33% [div.c:22 -> div.c:25] 0 ± 29.3% ▁████▁█▁ div [.] compute_flag
>> 8.33% [div.c:27 -> div.c:28] 0 ± 48.8% ▁██▁▁▁█▁ div [.] compute_flag
>> 4.80% [rand.c:26 -> rand.c:27] 0 ± 29.3% ▁████▁█▁ libc-2.27.so [.] rand
>> 4.80% [rand.c:28 -> rand.c:28] 0 ± 0.0% libc-2.27.so [.] rand
>> 2.14% [rand@plt+0 -> rand@plt+0] 0 ± 0.0% div [.] rand@plt
>
> I'm getting some unaligned lines:
>
> # Event 'cycles'
> #
> # Baseline [Program Block Range]/Cycles Diff/stddev/Hist Shared Object Symbol
> # ........ ................................................................................ ................... ..............................................
> #
> 11.87% [do_syscall_64+0 -> do_syscall_64+30] -22 ± 91.8% █▁ [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+0 -> do_syscall_64+47] -3 ± 84.9% ▁▁▁▁▁▁█▂ [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+91 -> do_syscall_64+118] -1 ± 33.5% ▁▁▅▃█▁▅█ [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+0 -> do_syscall_64+286] 0 ±100.0% ▁▁█▁▁▁▁▁ [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+0 -> do_syscall_64+332] 0 ± 0.0% [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+53 -> do_syscall_64+86] 0 ± 0.0% [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+124 -> do_syscall_64+148] 0 ± 26.3% ▁▆▂█▆▁██ [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+181 -> do_syscall_64+219] 0 ± 38.9% █▄▄▂▁▁▁▁ [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+278 -> do_syscall_64+281] 0 ± 0.0% [kernel.kallsyms] [k] do_syscall_64
> 11.87% [do_syscall_64+291 -> do_syscall_64+294] 0 ±100.0% ▁▁▁▁▁▁▁█ [kernel.kallsyms] [k] do_syscall_64
> 6.76% [psi_task_change+421 -> psi_task_change+440] -5 ± 45.6% ▄▂▁▁▄▁█▁ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+0 -> psi_task_change+60] -3 ± 47.9% ▁▃▁▂▁▁█▁ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+126 -> psi_task_change+225] -2 ± 22.8% ▂▅█▇▃▁▃▅ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+126 -> psi_task_change+172] -1 ± 48.8% ▁██▁█▁▁▁ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+476 -> psi_task_change+488] -1 ± 39.9% █▁▁█▁▁▄█ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+80 -> psi_task_change+118] 0 ± 50.0% █▁▂▂▁▁▄▁ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+80 -> psi_task_change+129] 0 ± 48.9% ▁▁▁▇▁█▁█ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+126 -> psi_task_change+142] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+230 -> psi_task_change+252] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+230 -> psi_task_change+265] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+254 -> psi_task_change+324] 0 ±100.0% ▁█▁▁▁▁▁▁ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+298 -> psi_task_change+307] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+309 -> psi_task_change+332] 0 ± 37.8% ▁▁█▁▁███ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+326 -> psi_task_change+370] 0 ± 79.5% ▁█▁▁▁▁▁▁ [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+341 -> psi_task_change+348] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+446 -> psi_task_change+456] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+461 -> psi_task_change+475] 0 ± 0.0% [kernel.kallsyms] [k] psi_task_change
> 6.76% [psi_task_change+493 -> psi_task_change+497] 0 ± 77.8% █▁▁▁▁▁▃▁ [kernel.kallsyms] [k] psi_task_change
> 5.27% [syscall_return_via_sysret+0 -> syscall_return_via_sysret+81] -21 ± 28.1% ▂▄█▄▅▂▁▁ [kernel.kallsyms] [k] syscall_return_via_sysret
> 5.27% [syscall_return_via_sysret+0 -> syscall_return_via_sysret+66] 0 ± 48.7% ▁▇▂▁▁▁█▁ [kernel.kallsyms] [k] syscall_return_via_sysret
> 5.27% [syscall_return_via_sysret+83 -> syscall_return_via_sysret+114] 0 ± 12.8% █▅▆▅▆▁▅▅ [kernel.kallsyms] [k] syscall_return_via_sysret
> 4.22% [native_write_msr+0 -> native_write_msr+11] 0 ± 24.6% █▅▄▄▂▄▁▁ [kernel.kallsyms] [k] native_write_msr
> 2.84% [enqueue_task_fair+257 -> enqueue_task_fair+281] -168 ± 94.5% ▁█ [kernel.kallsyms] [k] enqueue_task_fair
> 2.84% [enqueue_task_fair+904 -> enqueue_task_fair+916] -122 ± 0.0% [kernel.kallsyms] [k] enqueue_task_fair
> 2.84% [enqueue_task_fair+93 -> enqueue_task_fair+98] 27 ± 8.6% ▁█ [kernel.kallsyms] [k] enqueue_task_fair
> 2.84% [enqueue_task_fair+286 -> enqueue_task_fair+289] 5 ± 85.5% ▁▁█ [kernel.kallsyms] [k] enqueue_task_fair
> 2.84% [enqueue_task_fair+0 -> enqueue_task_fair+120] -3 ± 19.5% ▁█▁▃▂▄▆▃ [kernel.kallsyms] [k] enqueue_task_fair
>
> thanks,
> jirka
>
Line is too long. While putting the histogram in a separate column may
have better display. Let me try it.
Thanks
Jin Yao
next prev parent reply other threads:[~2019-08-06 11:46 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-24 22:14 [PATCH v2] perf diff: Report noisy for cycles diff Jin Yao
2019-08-06 0:16 ` Jin, Yao
2019-08-06 8:34 ` Jiri Olsa
2019-08-06 11:39 ` Jin, Yao
2019-08-06 8:34 ` Jiri Olsa
2019-08-06 11:46 ` Jin, Yao [this message]
2019-08-06 8:34 ` Jiri Olsa
2019-08-06 11:32 ` Jin, Yao
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=3efff55b-511c-0748-432c-e7ef4718c506@linux.intel.com \
--to=yao.jin@linux.intel.com \
--cc=Linux-kernel@vger.kernel.org \
--cc=acme@kernel.org \
--cc=ak@linux.intel.com \
--cc=alexander.shishkin@linux.intel.com \
--cc=jolsa@kernel.org \
--cc=jolsa@redhat.com \
--cc=kan.liang@intel.com \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
--cc=yao.jin@intel.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).