Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1456226 > unrolled thread

Re: [lkp] [sched/fair] e210bffd39: perf-stat.cpu-migrations -36.0% decrease

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [lkp] [sched/fair]  e210bffd39: perf-stat.cpu-migrations -36.0%  decrease Peter Zijlstra <peterz@infradead.org> - 2016-08-04 11:10 +0200
    Re: [lkp] [sched/fair] e210bffd39: perf-stat.cpu-migrations -36.0% decrease Vincent Guittot <vincent.guittot@linaro.org> - 2016-08-04 11:20 +0200
    Re: [lkp] [sched/fair]  e210bffd39: perf-stat.cpu-migrations -36.0%  decrease Ye Xiaolong <xiaolong.ye@intel.com> - 2016-08-04 16:10 +0200

#1456226 — Re: [lkp] [sched/fair] e210bffd39: perf-stat.cpu-migrations -36.0% decrease

On Thu, Aug 04, 2016 at 01:24:23PM +0800, kernel test robot wrote:
> 
> FYI, we noticed perf-stat.cpu-migrations -36.0% decrease due to commit:
> 
> commit e210bffd39d01b649c94b820c28ff112673266dd ("sched/fair: Fix and optimize the fork() path")
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> 
> in testcase: unixbench
> on test machine: 16 threads Broadwell-DE with 8G memory
> with following parameters:
> 
> 	runtime: 300s
> 	nr_task: 1
> 	test: shell8
> 	cpufreq_governor: performance
> 
> 

Urgh, I so hate these reports. They're near useless :/

It took me 3 readings to realize it was the unixbench testcase (why
isn't that in the subject!?). But then I still haven't found the primary
performance metric. Does the test improve, degrade or not change
overall?

[toc] | [next] | [standalone]

#1456242 — Re: [lkp] [sched/fair] e210bffd39: perf-stat.cpu-migrations -36.0% decrease

On 4 August 2016 at 11:01, Peter Zijlstra <peterz@infradead.org> wrote:
> On Thu, Aug 04, 2016 at 01:24:23PM +0800, kernel test robot wrote:
>>
>> FYI, we noticed perf-stat.cpu-migrations -36.0% decrease due to commit:
>>
>> commit e210bffd39d01b649c94b820c28ff112673266dd ("sched/fair: Fix and optimize the fork() path")
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
>>
>> in testcase: unixbench
>> on test machine: 16 threads Broadwell-DE with 8G memory
>> with following parameters:
>>
>>       runtime: 300s
>>       nr_task: 1
>>       test: shell8
>>       cpufreq_governor: performance
>>
>>
>
> Urgh, I so hate these reports. They're near useless :/
>
> It took me 3 readings to realize it was the unixbench testcase (why
> isn't that in the subject!?). But then I still haven't found the primary
> performance metric. Does the test improve, degrade or not change
> overall?

Yes, i have the same question

>
>

[toc] | [prev] | [next] | [standalone]

#1456417

On 08/04, Peter Zijlstra wrote:
>On Thu, Aug 04, 2016 at 01:24:23PM +0800, kernel test robot wrote:
>> 
>> FYI, we noticed perf-stat.cpu-migrations -36.0% decrease due to commit:
>> 
>> commit e210bffd39d01b649c94b820c28ff112673266dd ("sched/fair: Fix and optimize the fork() path")
>> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
>> 
>> in testcase: unixbench
>> on test machine: 16 threads Broadwell-DE with 8G memory
>> with following parameters:
>> 
>> 	runtime: 300s
>> 	nr_task: 1
>> 	test: shell8
>> 	cpufreq_governor: performance
>> 
>> 
>
>Urgh, I so hate these reports. They're near useless :/
>
>It took me 3 readings to realize it was the unixbench testcase (why
>isn't that in the subject!?). But then I still haven't found the primary

Will add test name in subject in later performance reports.

>performance metric. Does the test improve, degrade or not change
>overall?
>
Sorry, I should have described it more clearly, the primary performance
metric of unixbench is not changed, while perf-stat.cpu-migrations has a
-36.0% decrease which should improve cache locality.

We'd expect such stats changes report would help the developer to
evaluate impacts of his patch, anyway, we'll make it more clear that
it's just a FYI report, not some regression report. 

Thanks,
Xiaolong

>

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web