Page 20 PREVIOUS PAGE?

MPI_CART Routines

In principle, you are ready to write your code in parallel using domain decomposition. You now know the basic issues to consider when writing a parallel code using domain decomposition. Together with basic MPI communication routines (e.g. mpi_send mpi_recv, etc) you have all the tools you need to write your parallel code.

If you ask:

Do I have to design a "map" of neighbors for each process/block?

The answer is: Not Necessarily.

You can use MPI_CART routines to manage domain decomposition for you.

Minimal set of routines:

MPI_CART_CREATE
MPI_CART_GET
MPI_CART_RANK

Very Useful MPI_CART routines:

MPI_CART_SHIFT
MPI_CART_COORDS