Ben had the brilliant idea of doing boundary data output by creating a boundary mesh, solving a local system to project functions of interior data onto that mesh, then outputting the results.
Unfortunately, with certain partitionings, this can put *all* your boundary dofs on a single processor. This processor wants to create matrices and vectors. It now creates an MPI matrix (thanks to a bugfix of Ben's and one of my own; bad things happen if you try to do parallel operations on sequential matrices or mix them)... but how does it know to create an MPI vector? It sees n_local == n_global, it knows that sometimes we like creating sequential vectors on parallel runs... I think we need another argument to init. Probably a boolean with a sane default value so as not to break API compatibility, but I'm not sure on the details yet. Ben, you can see the current state of our merged code in ~roystgnr/fins.devel *on the ICES filesystem* - but I'm posting to libmesh-devel because anyone's opinions would be welcome. --- Roy ------------------------------------------------------------------------------ This SF.net email is sponsored by: SourcForge Community SourceForge wants to tell your story. http://p.sf.net/sfu/sf-spreadtheword _______________________________________________ Libmesh-devel mailing list Libmesh-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libmesh-devel
