On Tue, 11 Sep 2012, Cody Permann wrote:
So we do still have one test that's failing. John coded up the
example in pure libMesh code and it works so the problem doesn't
appear to be on your end. However there are a few differences in
the way MOOSE solves it and Johns example: for one we solve every
problem with a transient nonlinear system using JFNK which is of
course different from the simple example John has already coded up
using a linear system. We were down to looking at differences in
the system matrix to try to figure out why things were different,
but will need more time to narrow down the issue.
Is it possible that you're combining PBC-generated constraints with
user-generated or Dirichlet-generated constraints on the same node?
That's the one case I haven't tested well enough yet and wouldn't
trust a priori.
I think we've decided that we are going to put this one problem to
the side for now since it has been a hefty time sink and move
forward until we have some more downtime. The remaining PBCs issues
appear to be solved using your latest (uncommitted?) patch.
Good to hear. I'll change that from uncommitted to committed now,
then.
We are going to work on updating all of our users to the current
head of libMesh barring any other issues.
Thanks! If we're no longer forked, then I'll soon also be committing
a few other simple-but-voluminous changes (e.g. the
libmesh_assert_equal macro I mentioned on -devel a while ago) that I'd
been holding off on.
Thanks for working on this for us!
No trouble. It wasn't pure altruism. The race condition in threaded
PBCs has been causing our BuildBot tests fits for a while, and
although in hindsight it might have been more time-efficient for me to
simply disable those tests, at the time it didn't look like making
asynchronous PBC calculations robust was going to be so hard...
---
Roy
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Libmesh-devel mailing list
Libmesh-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/libmesh-devel
