lkml.org 
[lkml]   [2008]   [Jul]   [30]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [RFC] [PATCH 0/5 V2] Huge page backed user-space stacks
On (30/07/08 10:34), Andrew Morton didst pronounce:
> On Wed, 30 Jul 2008 18:23:18 +0100 Mel Gorman <mel@csn.ul.ie> wrote:
>
> > On (30/07/08 01:43), Andrew Morton didst pronounce:
> > > On Mon, 28 Jul 2008 12:17:10 -0700 Eric Munson <ebmunson@us.ibm.com> wrote:
> > >
> > > > Certain workloads benefit if their data or text segments are backed by
> > > > huge pages.
> > >
> > > oh. As this is a performance patch, it would be much better if its
> > > description contained some performance measurement results! Please.
> > >
> >
> > I ran these patches through STREAM (http://www.cs.virginia.edu/stream/).
> > STREAM itself was patched to allocate data from the stack instead of statically
> > for the test. They completed without any problem on x86, x86_64 and PPC64
> > and each test showed a performance gain from using hugepages. I can post
> > the raw figures but they are not currently in an eye-friendly format. Here
> > are some plots of the data though;
> >
> > x86: http://www.csn.ul.ie/~mel/postings/stack-backing-20080730/x86-stream-stack.ps
> > x86_64: http://www.csn.ul.ie/~mel/postings/stack-backing-20080730/x86_64-stream-stack.ps
> > ppc64-small: http://www.csn.ul.ie/~mel/postings/stack-backing-20080730/ppc64-small-stream-stack.ps
> > ppc64-large: http://www.csn.ul.ie/~mel/postings/stack-backing-20080730/ppc64-large-stream-stack.ps
> >
> > The test was to run STREAM with different array sizes (plotted on X-axis)
> > and measure the average throughput (y-axis). In each case, backing the stack
> > with large pages with a performance gain.
>
> So about a 10% speedup on x86 for most STREAM configurations. Handy -
> that's somewhat larger than most hugepage-conversions, iirc.
>

It is a bit. Usually, I expect around 5%.

> Do we expect that this change will be replicated in other
> memory-intensive apps? (I do).
>

I expect so. I know SpecCPU has some benchmarks that are stack-dependent and
would benefit from this patchset. I haven't experimented enough yet with other
workloads to give a decent estimate. I've added Andrew Hastings to the cc as
I believe he can make a good estimate on what sort of gains had by backing
the stack with huge pages based on experiments along those lines. Andrew?

With Erics patch and libhugetlbfs, we can automatically back text/data[1],
malloc[2] and stacks without source modification. Fairly soon, libhugetlbfs
will also be able to override shmget() to add SHM_HUGETLB. That should cover
a lot of the memory-intensive apps without source modification.

[1] It can partially remap non-hugepage-aligned segments but ideally the
application would be relinked

[2] Allocated via the morecore hook in glibc

--
Mel Gorman
Part-time Phd Student Linux Technology Center
University of Limerick IBM Dublin Software Lab


\
 
 \ /
  Last update: 2008-07-30 21:33    [W:0.172 / U:0.720 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site