On Thu, 7 Jul 2011, Kirk, Benjamin (JSC-EG311) wrote:

> Maybe it isn't actually us?  I believe there has been a bugfix
> release of parmetis beyond the version we include, but I haven't had
> time to try packaging it.

I gave the newest METIS/ParMETIS a crack today, but got the same crash
in the same place.  I'm not going to bother committing unless anyone
else wants, though; right now I'm not in the mood to risk breaking
what works unless doing so adds a new feature or fixes something
that's broken.

It's a subtle problem, whatever it is.  I can't seem to reproduce it
on other systems, either.  And only thing particularly odd about this
laptop is that I've got an i686 instead of an x86_64 install on it
right now.

It might not be a parmetis problem, though.  I've also seen a couple
wonky incorrect results from libMesh calls to MPI_Allgather using the
default openmpi installation on this thing.  If there's a problem with
my MPI stack then that could definitely screw up parmetis too.  I'll
try replacing the default Ubuntu 11.04 i686 packages with a
built-from-scratch stack some time later and see if that fixes things.
---
Roy

------------------------------------------------------------------------------
All of the data generated in your IT infrastructure is seriously valuable.
Why? It contains a definitive record of application performance, security 
threats, fraudulent activity, and more. Splunk takes this data and makes 
sense of it. IT sense. And common sense.
http://p.sf.net/sfu/splunk-d2d-c2
_______________________________________________
Libmesh-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-devel

Reply via email to