Sounds like the problem comes down to just 32-bit systems that fault on unaligned 8-byte loads.
That would be SPARC, IA64 and MIPS.  For IB only SPARC is relevant.

So perhaps alignment>2 should be conditional on 32-bit SPARC target.
Additionally, an experiment to see if 4-byte alignment is "good enough" (instead of 8) might allow one to save a little space.

-Paul


On Wed, Feb 20, 2013 at 1:51 PM, Nathan Hjelm <hjelmn@lanl.gov> wrote:
I talked to Pasha about the change. He suggests fragments are 2-byte aligned to save space. I suspect that on 64-bit platforms the fragment size is already a multiple of 8 bytes so this change will likely only affect 32-bit systems (which is where the bus error is occurring).

-Nathan

On Wed, Feb 20, 2013 at 09:39:09PM +0000, Joshua Ladd wrote:
> I would hold off, if possible, until I can investigate the issue. I don't, off-hand, know why the 2-byte alignment, although I would suspect it's for performance reasons.
>
>
> Josh
>
>
> -----Original Message-----
> From: devel-bounces@open-mpi.org [mailto:devel-bounces@open-mpi.org] On Behalf Of Jeff Squyres (jsquyres)
> Sent: Wednesday, February 20, 2013 4:35 PM
> To: <devel@open-mpi.org>
> Subject: [OMPI devel] openib fragment alignment
> Importance: High
>
> Arrgh.  I think you're going to make me eat my words (http://www.open-mpi.org/community/lists/devel/2013/02/12143.php).
>
> I just recently lost my access to InfiniBand test gear, so I can't test this myself.  Hypothetically, it should be fine.  But throwing in an untested change literally right before a release without IB vendor say-so really, really gives me pause...
>
> Mellanox?
>
>
> On Feb 20, 2013, at 4:27 PM, Open MPI <bugs@open-mpi.org> wrote:
>
> > #3519: Move r28083 to v1.6 branch
> > -----------------------------------+--------------------------
> > Reporter:  hjelmn                  |      Owner:  hjelmn
> >    Type:  changeset move request  |     Status:  new
> > Priority:  major                   |  Milestone:  Open MPI 1.6
> > Version:  trunk                   |
> > -----------------------------------+--------------------------
> > (In [28083]) btl/openib: don't align fragments on 2 byte boundaries
> > (changed to 8)
> >
> > cmr:v1.6,v1.7
> >
> > --
> > Ticket URL: <https://svn.open-mpi.org/trac/ompi/ticket/3519>
> > Open MPI <http://www.open-mpi.org/>
> >
> > _______________________________________________
> > bugs mailing list
> > bugs@open-mpi.org
> > http://www.open-mpi.org/mailman/listinfo.cgi/bugs
>
>
> --
> Jeff Squyres
> jsquyres@cisco.com
> For corporate legal information go to: http://www.cisco.com/web/about/doing_business/legal/cri/
>
>
> _______________________________________________
> devel mailing list
> devel@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/devel
>
> _______________________________________________
> devel mailing list
> devel@open-mpi.org
> http://www.open-mpi.org/mailman/listinfo.cgi/devel
_______________________________________________
devel mailing list
devel@open-mpi.org
http://www.open-mpi.org/mailman/listinfo.cgi/devel



--
Paul H. Hargrove                          PHHargrove@lbl.gov
Future Technologies Group
Computer and Data Sciences Department     Tel: +1-510-495-2352
Lawrence Berkeley National Laboratory     Fax: +1-510-486-6900