Implementation of Message Passing Interface on the Sparrow Cluster

This project examines the performance and methodology of diierent communication paradigms, shared memory communication and BSD socket communication. Goals for this work include installation of Message Passing Interface, enabling shared memory via MPI devices, and benchmarking the performance of a system with shared memory and without shared memory. We will show the compilation method required by MPI for taking advantage of multiprocessor resources, and the impact these resources have on computational eeciency. We provide a description of the MPI device that handles the hardware, and show results of benchmark testing.