On Wed, Oct 11, 2017 at 10:34:47AM -0700, Dave Hansen wrote:
> On 10/11/2017 01:06 AM, Jan Kara wrote:
> >>> when rebasing our enterprise distro to a newer kernel (from 4.4 to 4.12) we
> >>> have noticed a regression in bonnie++ benchmark when deleting files.
> >>> Eventually we have tracked this down to a fact that page cache truncation got
> >>> slower by about 10%. There were both gains and losses in the above interval of
> >>> kernels but we have been able to identify that commit 83929372f629 "filemap:
> >>> prepare find and delete operations for huge pages" caused about 10% regression
> >>> on its own.
> >> It's odd that just checking if some pages are huge should be that
> >> expensive, but ok ..
> > Yeah, I was surprised as well but profiles were pretty clear on this - part
> > of the slowdown was caused by loads of page->_compound_head (PageTail()
> > and page_compound() use that) which we previously didn't have to load at
> > all, part was in hpage_nr_pages() function and its use.
>
> Well, page->_compound_head is part of the same cacheline as the rest of
> the page, and the page is surely getting touched during truncation at
> _some_ point. The hpage_nr_pages() might cause the cacheline to get
> loaded earlier than before, but I can't imagine that it's that expensive.
Profiles appear to disagree but regardless of the explanation, the fact
is that the series improves truncation quite a bit on my tests. From three
separate machines running bonnie, I see the following gains.
4.14.0-rc4 4.14.0-rc4
vanilla janbatch-v1r1
Hmean SeqCreate del 21313.45 ( 0.00%) 24963.95 ( 17.13%)
Hmean RandCreate del 19974.03 ( 0.00%) 23377.66 ( 17.04%)
4.14.0-rc4 4.14.0-rc4
vanilla janbatch-v1r1
Hmean SeqCreate del 4408.80 ( 0.00%) 5074.91 ( 15.11%)
Hmean RandCreate del 4161.52 ( 0.00%) 4879.15 ( 17.24%)
4.14.0-rc4 4.14.0-rc4
vanilla janbatch-v1r1
Hmean SeqCreate del 11639.73 ( 0.00%) 13648.20 ( 17.26%)
Hmean RandCreate del 10979.90 ( 0.00%) 12818.99 ( 16.75%)
--
Mel Gorman
SUSE Labs
