[Ksummit-2013-discuss] [ATTEND] Transparent huge pagecache

[Aneesh Kumar K.V]

"Kirill A. Shutemov" <kirill at shutemov.name> writes:

> Hi,
>
> Current Transparent Huge Pages implementation is limited to anonymous
> memory mapping. I work on adding support of huge pages in pagecache and
> enabling it for most simple filesystem -- ramfs. The patchset is mostly
> ready, although it hasn't got much review yet.
>
> I would like to discuss upstreaming of the feature and future steps --
> enabling it for more useful filesystems: tmpfs and, probably, some
> filesystem with backing storage (ext4?).
> One other possible application is THP as backend for hugetlb interface.
>

I would be interested in transparent hugepage discussion. It would be
interesting to discuss how architectures like ppc64 that have block size
< page size can do this better ?. Do we need to split hugepage to
regular page and then to block size for writepage ? . Also to improve
worst case transparent hugepage performance (application touching one
byte of every hugepage) does it make sense to actually have a mode in
which we don't allocate hugepage on fault but only do collapse based
hugepage allocation. If so will this help in page cache hugepage so that
we don't allocate hugepage on fault ?

-aneesh