Until it's fixed, beware the combination of:
Nodal FEM bases (e.g. first order LAGRANGE/HIERARCHIC, or third order
HERMITE)
Second-order geometric elements (e.g. TRI6, HEX27)
Parallel non-conforming meshes
The asymmetric_constraint_rows=false option to DofMap::constrain_*
(whether you call it yourself or simply use FEMSystem)
If you use all four of those things in the same simulation, it is
currently possible to get a subtlely incorrect solution on any hanging
node on the partition boundary between fine elements on a
higher-numbered processor id and coarse elements on a lower-numbered
processor id. The solution error is subtle, on the order of your
approximation error, but because it involves a non-satisfied
constraint, the error in derived terms like jump error estimators can
be large.
If you have an application where this combination seems unavoidable
(e.g. large simulations with Taylor-Hood elements in FEMSystem), you
might want to temporarily hack your library copy to turn
asymmetric_constraint_rows off.
I'll post again when we're sure the bug is fixed in SVN, and unless
anyone has found any other outstanding bugs, we'll release 0.6.4
immediately afterward.
---
Roy
------------------------------------------------------------------------------
Come build with us! The BlackBerry® Developer Conference in SF, CA
is the only developer event you need to attend this year. Jumpstart your
developing skills, take BlackBerry mobile applications to market and stay
ahead of the curve. Join us from November 9-12, 2009. Register now!
http://p.sf.net/sfu/devconf
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
