> On 三月 8, 2016, 1:35 a.m., Ben Mahler wrote:
> > src/linux/perf.cpp, lines 430-431
> > <https://reviews.apache.org/r/44379/diff/1/?file=1280476#file1280476line430>
> >
> > Couple of thoughts here:
> >
> > (1) This comment is no longer accurate since there is a special case
> > for 3.10. Would be better to re-structure the last block of this else flow
> > to better reflect the new logic.
> >
> > (2) What about 3.11? Is it the same as 3.10? It would be surprising if
> > 3.11 and < 3.10 are the same, but 3.10 is different.
>
> haosdent huang wrote:
> There is a exist patch may related to this.
> https://reviews.apache.org/r/43283/
To Ben Mahler:
Thanks for the the comments, you got it, I gave vanilla 3.10 kernel a try, it
turns out as you expected:
<not counted>,cycles,mesos/0ca020e9-2db3-4a47-87f0-b4c831409ef7
I use Centos 7 with kernel 3.10.0-123.el7.x86_64, might with perf enhancement.
I will rework the comments anyway if you don't have any other comments.
To Haosdent:
I also use Centos 7, and tried 3.10.0-{123 OR 229}.el7.x86_64, perf stat has
below format:
value,unit,event,cgroup,running,ratio
May I know which kernel version do you use?
thanks
- fan
-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://reviews.apache.org/r/44379/#review122455
-----------------------------------------------------------
On 三月 4, 2016, 6:28 a.m., fan du wrote:
>
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://reviews.apache.org/r/44379/
> -----------------------------------------------------------
>
> (Updated 三月 4, 2016, 6:28 a.m.)
>
>
> Review request for mesos.
>
>
> Bugs: MESOS-4705
> https://issues.apache.org/jira/browse/MESOS-4705
>
>
> Repository: mesos
>
>
> Description
> -------
>
> Correctly parse perf stat format for 3.10 kernel.
>
>
> Diffs
> -----
>
> src/linux/perf.cpp 1c113a2b3f57877e132bbd65e01fb2f045132128
>
> Diff: https://reviews.apache.org/r/44379/diff/
>
>
> Testing
> -------
>
> {Found and Test} with Serenity, ema filter could get perf event statistics
> correctly as expected.
>
>
> Thanks,
>
> fan du
>
>
