In this chapter, we looked in some detail at the MPI standard, along with a number of its implementations, specifically Open MPI, and we looked at how to set up a cluster. We also saw how to use OpenMP to easily add multithreading to existing codes.
At this point, the reader should be capable of setting up a basic Beowulf or similar cluster, configuring it for MPI, and running basic MPI applications on it. How to communicate between MPI processes and how to define custom data types should be known. In addition, the reader will be aware of the potential pitfalls when programming for MPI.
In the next chapter, we will take all our knowledge of the preceding chapters and see how we can combine it in the final chapter, as we look at general-purpose computing on videocards (GPGPU).
