MPI code

Parallel distributed computing is always done through the Message Passing Interface (MPI). MPI is the parallel computing standard. As such, almost all distributed computing applications rely on MPI.

If your application requires the use of MPI, we strongly encourage that you should use C or fortran as an implementation. However, we recognize that this requires a large learning curve associated with learning a lower level langauge like C. Our thoughts are generally that if you require this amount of performance, other factors beside simply message passing is often required to make the application tenable.

MPI using R

However, some smaller applications can get away with using R for MPI. The pbdMPI package is currently the best MPI implementation in R. However, the package is some what poorly documentated. In that they expect some familiarity with the MPI standard to understand what their functions do.

C/fortran MPI implementations

There are several implementations of the MPI standard for C and fortran. They all offer mostly the same functionality, however, there are some differences. Largely, for most programmers they are interchangable.

• openMPI
• MPICH
• MVAPICH