[PATCH] hugepages: cache THP mode and page size [BZ #34083]

[Shamil Abdulaev]

> I will post my bz #34083 patch after my v6 THP patch is merged.

Understood, withdrawing mine.  Thanks.


On Sat, May 02, 2026 11:41 AM, "H.J. Lu" <hjl.tools@gmail.com> wrote:

> On Sat, May 2, 2026 at 4:35 PM Shamil Abdulaev <ashamil435@gmail.com>
> wrote:
> >
> > On Sat, May 2, 2026 at 6:27 AM H.J. Lu <hjl.tools@gmail.com> wrote:
> > > I took a different approach:
> > >
> https://gitlab.com/x86-glibc/glibc/-/commits/users/hjl/hugepage/pr34083
> > >
> > > since 2 THP optimizations can be controlled independently.   They share
> > > the same THP mode only if both are enabled.
> >
> > Thanks for the pointer.
> >
> > Your "Support THP segment load with madvise enabled THP" series is at
> > v6 and under review by Carlos -- great, that covers BZ #34079 / BZ #34080
> > well, and I am not trying to compete there.
> >
> > For BZ #34083 specifically, the malloc commit b7b121c on your branch
> > ("malloc: Don't call __get_thp_mode/__get_thp_size twice") does not
> > appear to have been posted to libc-alpha.  A search of the archive for
> > that subject returns no results, so currently the only public patch
> > addressing #34083 is mine.
> >
> > If you plan to post b7b121c as a follow-up to the v6 series, please let
> > me know the timing and I am happy to wait, withdraw, or rebase.
>
> I will post my bz #34083 patch after my v6 THP patch is merged.
>
> > Otherwise, could my standalone caching fix for #34083 land?  It is
> > independent of the madvise rework, and your malloc commit could rebase
> > on top trivially if you decide to post it later.
> >
> > Thanks,
> > Shamil
>
>
>
> --
> H.J.
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://sourceware.org/pipermail/libc-alpha/attachments/20260502/25bb2982/attachment.htm>