Intro to Code Parallelization with MPI

Course Material: Last Updated:   See Also:

[download here] Sep 18, 2007   [additional reference material] (off-site link)

The goal of this four-session (2 hours each) sequel is to teach basic and, to some extent, intermediate MPI material. Specifically, the course covers the message-passing model for parallel programming as implemented by the MPI standard. Within that framework, it covers point-to-point MPI process communication as well as collective communications between processes. To a good extent the presentation of the material is example driven. A list of the specific topics includes the following:

• message passing concepts
• layout of an MPI program
• compilation and running of MPI programs
• loop iteration scheduling and parallelization
• point-to-point communication calls, mpi_send & mpi_recv in blocking and unblocking mode
• ancillary routines for point-to-point calls: mpi_test; mpi_wait; etc.
• collective communications: mpi_bcast; mpi_gather; etc.
• communication deadlocks
• examples of various levels of difficulty are given for all of the above

Throughout the presentation an interactive environment with the lecturer is encouraged.