Realities associated with parallel processing

Abstract

The Density Functional Program, which is used to study the electronic structure of semiconductors, consumes the most CPU time of all production applications at the IBM T. J. Watson Research Center. Thus, it was decided to undertake the parallelization of this code not only with the hope of reducing elapsed time, improving turnaround on the IBM 3090, and conserving other system resources such as memory and disk space, but also in an attempt to test the IBM Parallel Fortran Compiler on a production program while developing an understanding of the impact of parallel jobs from a system standpoint. Numerically, the program is constructed around two standard procedures involving matrix diagonalization and fast Fourier transform (FFT). The parallelization strategy is described, hotspot analysis and dynamic load balancing are discussed, and parallel loop implementation is examined. The results are reported and discussed in detail.