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 v3 0/7] perf: Stream comparison
Date: Tue, 28 Apr 2020 16:10:12 +0800 [thread overview]
Message-ID: <d2fc66e5-d33a-0ed2-e46b-6cfbc4a941a6@linux.intel.com> (raw)
In-Reply-To: <20200427101044.GA1457790@krava>
Hi Jiri,
On 4/27/2020 6:10 PM, Jiri Olsa wrote:
> On Mon, Apr 20, 2020 at 09:04:44AM +0800, Jin Yao wrote:
>
> SNIP
>
>> compute_flag div.c:25 compute_flag div.c:25
>> compute_flag div.c:22 compute_flag div.c:22
>> main div.c:40 main div.c:40
>> main div.c:40 main div.c:40
>> main div.c:39 main div.c:39*
>>
>> [ Hot chains in old perf data only ]
>>
>> hot chain 1:
>> cycles: 2, hits: 4.08%
>> --------------------------
>> main div.c:42
>> compute_flag div.c:28
>>
>> [ Hot chains in new perf data only ]
>>
>> hot chain 1:
>> cycles: 36, hits: 3.36%
>> --------------------------
>> __random_r random_r.c:357
>> __random random.c:293
>> __random random.c:293
>> __random random.c:291
>> __random random.c:291
>> __random random.c:291
>> __random random.c:288
>> rand rand.c:27
>> rand rand.c:26
>> rand@plt
>> rand@plt
>> compute_flag div.c:25
>> compute_flag div.c:22
>> main div.c:40
>> main div.c:40
>>
>> Now we can see, following streams pair is moved to another section
>> "[ Hot chains in old perf data but source line changed (*) in new perf data ]"
>>
>> cycles: 1, hits: 26.80% cycles: 1, hits: 27.30%
>> --------------------------- --------------------------
>> main div.c:39 main div.c:39*
>> main div.c:44 main div.c:44
>>
>
>
> so I tried following:
>
> # ./perf record -e cycles:u -b ./perf bench sched pipe
> # ./perf record -e cycles:u -b ./perf bench sched pipe
> # ./perf diff -f --stream --before $PWD --after $PWD >out 2>&1
>
> and the out file looks like this:
>
> [ Matched hot chains between old perf data and new perf data ]
>
> [ Hot chains in old perf data but source line changed (*) in new perf data ]
>
> [ Hot chains in old perf data only ]
>
> hot chain 1:
> cycles: 0, hits: 4.20%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 2:
> cycles: 0, hits: 4.11%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 3:
> cycles: 0, hits: 8.22%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 4:
> cycles: 0, hits: 5.54%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 5:
> cycles: 0, hits: 6.10%
> --------------------------
> 0xffffffff89c00163
>
> [ Hot chains in new perf data only ]
>
> hot chain 1:
> cycles: 0, hits: 5.21%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 2:
> cycles: 0, hits: 4.79%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 3:
> cycles: 0, hits: 5.44%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 4:
> cycles: 0, hits: 5.50%
> --------------------------
> 0xffffffff89c00163
>
> hot chain 5:
> cycles: 0, hits: 7.14%
> --------------------------
> 0xffffffff89c00163
>
>
> I'd expected more common paths, from what I can see from 'perf report --branch-history'
> on bpth perf.data and perf.data.old
>
> jirka
>
I used the same command line but I can see more callchain entries.
perf record -e cycles:u -b perf bench sched pipe
perf record -e cycles:u -b perf bench sched pipe
perf diff --stream
[ Matched hot chains between old perf data and new perf data ]
hot chain pair 1:
cycles: 0, hits: 7.95% cycles: 0, hits: 6.61%
--------------------------- --------------------------
__libc_read read.c:27 __libc_read read.c:27
0xffffffffa9800163 0xffffffffa9800163
[ Hot chains in old perf data but source line changed (*) in new perf data ]
[ Hot chains in old perf data only ]
hot chain 1:
cycles: 49, hits: 4.98%
--------------------------
worker_thread sched-pipe.c:64
worker_thread sched-pipe.c:63
__libc_read read.c:28
__libc_read read.c:27
0xffffffffa9800163
hot chain 2:
cycles: 0, hits: 6.68%
--------------------------
0xffffffffa9800163
hot chain 3:
cycles: 0, hits: 6.57%
--------------------------
0xffffffffa9800163
hot chain 4:
cycles: 60, hits: 5.20%
--------------------------
worker_thread sched-pipe.c:67
worker_thread sched-pipe.c:60
worker_thread sched-pipe.c:70
worker_thread sched-pipe.c:70
__libc_read read.c:28
__libc_read read.c:27
0xffffffffa9800163
[ Hot chains in new perf data only ]
hot chain 1:
cycles: 68, hits: 7.83%
--------------------------
worker_thread sched-pipe.c:68
__libc_write write.c:28
__libc_write write.c:27
0xffffffffa9800163
__libc_write write.c:27
write@plt
write@plt
worker_thread sched-pipe.c:67
worker_thread sched-pipe.c:60
worker_thread sched-pipe.c:70
worker_thread sched-pipe.c:70
__libc_read read.c:28
hot chain 2:
cycles: 70, hits: 4.34%
--------------------------
worker_thread unistd.h:44
worker_thread sched-pipe.c:61
worker_thread sched-pipe.c:65
__libc_write write.c:28
__libc_write write.c:27
0xffffffffa9800163
__libc_write write.c:27
write@plt
write@plt
worker_thread sched-pipe.c:64
worker_thread sched-pipe.c:63
__libc_read read.c:28
hot chain 3:
cycles: 0, hits: 5.67%
--------------------------
0xffffffffa9800163
hot chain 4:
cycles: 0, hits: 5.47%
--------------------------
0xffffffffa9800163
It's interesting that some leaked kernel address are displayed in callchains
even we use the -e cycles:u. Should be the skid issue. I have a patch for
processing the kernel leaked samples but have not posted it.
But I'm no idea why only the leaked kernel address displayed in your example. :(
Thanks
Jin Yao
next prev parent reply other threads:[~2020-04-28 8:10 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-20 1:04 [PATCH v3 0/7] perf: Stream comparison Jin Yao
2020-04-20 1:04 ` [PATCH v3 1/7] perf util: Create source line mapping table Jin Yao
2020-04-27 10:11 ` Jiri Olsa
2020-04-28 8:27 ` Jin, Yao
2020-04-20 1:04 ` [PATCH v3 2/7] perf util: Create streams for managing top N hottest callchains Jin Yao
2020-04-27 10:10 ` Jiri Olsa
2020-04-28 8:12 ` Jin, Yao
2020-04-20 1:04 ` [PATCH v3 3/7] perf util: Return per-event callchain streams Jin Yao
2020-04-20 1:04 ` [PATCH v3 4/7] perf util: Compare two streams Jin Yao
2020-04-20 1:04 ` [PATCH v3 5/7] perf util: Calculate the sum of all streams hits Jin Yao
2020-04-20 1:04 ` [PATCH v3 6/7] perf util: Report hot streams Jin Yao
2020-04-20 1:04 ` [PATCH v3 7/7] perf diff: Support hot streams comparison Jin Yao
2020-04-27 10:10 ` [PATCH v3 0/7] perf: Stream comparison Jiri Olsa
2020-04-28 8:10 ` Jin, Yao [this message]
2020-04-27 10:29 ` Jiri Olsa
2020-04-28 8:29 ` 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=d2fc66e5-d33a-0ed2-e46b-6cfbc4a941a6@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