Re: [mm] f3344adf38: fxmark.hdd_btrfs_DWAL_63_bufferedio.works/sec -52.4% regression

[Michal Hocko]

On Mon 15-03-21 13:42:50, Linus Torvalds wrote:
> If somebody can actually figure out what happened there, that would be
> good, but for now it goes into my "archived as a random outlier"
> folder.

This is not something new. We have seen reports like that in the past
already. In many cases there was no apparent reason except for potential
code alignment: e.g. http://lkml.kernel.org/r/20200409135029.GA2072@xsang-OptiPlex-9020
There were some other memcg related which didn't really show any
increase in cache misses or similar. I cannot find those in my notes
right now. There were others like https://lore.kernel.org/lkml/20201102091543.GM31092@shao2-debian/
with analysis http://lkml.kernel.org/r/20201125062445.GA51005@xxxxxxxxxxxxxxxxxxxxxxx
which actually pointed to something legit actually.

For this particular report I do not see any real relation to the patch
either.
-- 
Michal Hocko
SUSE Labs