Data distribution and grid overlap details are stored in several Mesh fields. More...

Variables
integer, dimension(:), pointer	Mesh_class::Mesh::eptnmap
	Partition map: elements (actually nodes) to partitions (process numbers)
type(indlist), dimension(:), pointer	Mesh_class::Mesh::ax_recvidx
type(indlist), dimension(:), pointer	Mesh_class::Mesh::ax_sendidx
type(indlist), dimension(:), pointer	Mesh_class::Mesh::ol_outer
	overlap with each neighbour process in its region
type(indlist), dimension(:), pointer	Mesh_class::Mesh::ol_inner
	overlap with each neighbour process in my region
type(indlist), dimension(:), pointer	Mesh_class::Mesh::ol_solve
	overlap with each neighbour process in all regions (its,my,or third)

Detailed Description

Data distribution and grid overlap details are stored in several Mesh fields.

Let

- number of processes
- number of subdomains on all processes
- number of fine aggregates on all processes

First, there are process regions $ U_i $ nodes that each process contains after initial partitioning in the distribution phase. Then, the expanded regions $\tilde U_i$ include the overlap, that is needed for preconditioners, particularly for the subdomains $\tilde V_k, k = 1 \dots n_d$ in first level preconditioner, and coarse space node basis function supports $\tilde W_k, k = 1 \dots n_a$ . For each process i, there are 3 kinds of overlap with process j: