Libmesh uses one of several partitioning schemes to assign relatively
contiguous blocks of elements to each processor. By default a multilevel graph
partitioner is used.
The degrees or freedom then are pretty well owned by the same processors that
own the elements, the exception being at proce
Hi,
As far as I understand, libmesh assembles the global stiffness matrix in
parallel due to the fact that each local performs the assembly only on its
own elements. And also in PETSc, the matrix assembly procedure will be
efficient only when the entries are generated on the 'right' process. Thus
