Double standards: bringing task parallelism to HPF via the message passing interface

  • Authors:

  • Ian Foster;David R. Kohr, Jr.;Rakesh Krishnaiyer;Alok Choudhary

  • Affiliations:

  • Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, IL;Mathematics and Computer Science Division, Argonne National Laboratory, Argonne, IL;Department of Computer and Information Science, Syracuse University, Syracuse, NY;Department of Computer and Information Science, Syracuse University, Syracuse, NY

  • Venue:
  • Supercomputing '96 Proceedings of the 1996 ACM/IEEE conference on Supercomputing
  • Year:
  • 1996

Quantified Score

Hi-index 0.00

Visualization

Abstract

High Performance Fortran (HPF) does not allow efficient expression of mixed task/data-parallel computations or the coupling of separately compiled data-parallel modules. In this paper, we show how a coordination library implementing the Message Passing Interface (MPI) can be used to represent these common parallel program structures. This library allows data-parallel tasks to exchange distributed data structures using calls to simple communication functions. We present microbenchmark results that characterize the performance of this library and that quantify the impact of optimizations that allow reuse of communication schedules in common situations. In addition, results from two-dimensional FFT, convolution, and multiblock programs demonstrate that the HPF/MPI library can provide performance superior to that of pure HPF. We conclude that this synergistic combination of two parallel programming standards represents a useful approach to task parallelism in a data-parallel framework, increasing the range of problems addressable in HPF without requiring complex compiler technology.