I am running a simulation in parallel in ANSYS CFX 11 which features a moving mesh with a GGI interface.
I noticed that the overlap nodes are very high for my parallel partitions and I can't reduce this by changing to
a different partitioning method.
If I run the case without moving mesh, the overlap nodes are much reduced.
What's going on?
|
If the mesh on either side of an interface is moving, then all nodes on that interface are marked as overlap. `Overlap` simply means that a node (and it`s associated mesh faces) is a candidate for interface re-intersection. This is required since, with moving mesh, we never know a-priori where (or how) the mesh on each side of the interface will actually intersect; we must presume that point `A` on one side of the interface could intersect with *any* point on the other side of the interface. Although the memory usage will be increased due to the high overlap, there shouldn't be too large a hit in terms of the CPU time required for the solution. |
||
|