Introduction to MPI

An intensive one day course with practicals to encourage learning how to exploit high levels of parallelism available within a simulation or analysis by using multiple cores and nodes of a supercomputer. This introduction to “distributed memory” programming will give participants sufficient knowledge to start writing MPI programmes in FORTRAN and to appreciate how codes such as NEMO, UM and WRF have used MPI to reduce run times by several orders of magnitude. Specifically, we shall cover the programming model, point-to-point communications and collectives, but shall not cover RDMA, spawning or other advanced topics.