linux-kernel - Re: Superpages Project - sourceforge.net/projects/linuxsuperpages

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-Id: <1224075647.28131.6.camel@twins>
Date:	Wed, 15 Oct 2008 15:00:47 +0200
From:	Peter Zijlstra <peterz@...radead.org>
To:	"Wildman, Tom" <tom.wildman@...com>
Cc:	"linux-mm@...ck.org" <linux-mm@...ck.org>,
	"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
	"alan@...hat.com" <alan@...hat.com>, Mel Gorman <mel@...net.ie>
Subject: Re: Superpages Project  -  sourceforge.net/projects/linuxsuperpages

On Wed, 2008-10-15 at 01:51 +0000, Wildman, Tom wrote:
> A new project has been created at SourceForge with an implementation
> of the Rice University's Superpages FreeBSD prototype that has been
> ported to the 2.6 Linux kernel for IA64, x86-64, and x86-32.
> 
> The project can be found at:
> http://sourceforge.net/projects/linuxsuperpages
> 
> The major benefit of supporting Superpages is increased memory reach
> of the processor's TLB, which reduces the number of TLB misses in
> applications that have large data sets.  Some benchmarks have been
> improved 20% in execution time.
> 
> Reference www.cs.rice.edu/~jnavarro/superpages/ for more information
> about the Rice University's Superpages project.
> 
> The project is being made available to the Open Source community to
> share the implementation and knowledge.  With the enhancements to the
> x86 architectures to support multiple and large page sizes there
> should be increased interest in this functionality.

How are you proposing to address the fun issues like online compaction
etc. ?

Furthermore, what's the added advantage of dynamic superpages over
exlpicit superpage support, eg. though the use of libhugetlb?

Unless you realize online compaction and add a kind of extend allocation
to the page allocator there will hardly ever be a situation where you
can promote a page.

All of which is rather expensive to do, changing the application might
be easier and deliver better performance gains.

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