Re: [LKP] [mm, memcg] 1e577f970f: will-it-scale.per_process_ops -7.2% regression
From: Michal Hocko
Date: Tue Sep 10 2019 - 03:31:25 EST
On Tue 10-09-19 15:08:41, Huang, Ying wrote:
> Michal Hocko <mhocko@xxxxxxxxxx> writes:
> > On Mon 09-09-19 10:15:44, kernel test robot wrote:
> >> Greeting,
> >> FYI, we noticed a -7.2% regression of will-it-scale.per_process_ops due to commit:
> > What is the memcg setup for this test?
> Except enabling memcg in kernel config, we don't do any special
> setup for memcg for the test.
OK, this is really interesting because the local events introduced by
this commits do not apply to the root memcg. So this smells like another
case when a microbenchmark is sensitive on the memcg layout changes.