I've got an easy-to-reproduce but hard-to-fix runtime failure with
--enable-parmesh: just run ex0 on 4 processors.  Or change the number
of elements in the initial mesh it builds to 5, and run it with 5
processors.  Etc.

The resulting memory corruption triggers a bad free deep in ParMETIS
code, and on first examination I can't see why there's a problem.
ParMETIS can't handle having any processors which lack initial
elements, but we already test for that and fall back on METIS if
that's the case.  (and indeed, running ex0 on 5 or more processors
with 4 elements seems to work fine).

Any ideas?  I don't have time to check this out in depth myself right
now, but I'd rather not stick in a "switch to METIS in this corner
case" workaround unless we really have to.
---
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