Re: [PATCH 2.5.43-mm2] New shared page table patch

From: Gerrit Huizenga (gh@us.ibm.com)
Date: Tue Oct 22 2002 - 14:06:29 EST


In message <3DB59DA7.453F89E2@digeo.com>, > : Andrew Morton writes:
> Dave McCracken wrote:
> >
> > And
> > 3) The current large page implementation is only for applications
> > that want anonymous *non-pageable* shared memory. Shared page
> > tables reduce resource usage for any shared area that's mapped
> > at a common address and is large enough to span entire pte pages.
> > Since all pte pages are shared on a COW basis at fork time, children
> > will continue to share all large read-only areas with their
> > parent, eg large executables.
> >
>
> How important is that in practice?
>
> Seems that large pages are the preferred solution to the "Oracle
> and DB2 use gobs of pagetable" problem because large pages also
> reduce tlb reload traffic.
>
> So once that's out of the picture, what real-world, observed,
> customers-are-hurting problem is solved by pagetable sharing?

If the shared pte patch had mmap support, then all shared libraries
would benefit. Might need to align them to 4 MB boundaries for best
results, which would also be easy for libraries with unspecified
attach addresses (e.g. most shared libraries).

gerrit
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Wed Oct 23 2002 - 22:01:00 EST