Experiences on Data-Parallel Programming

To parallelize a scientific application with a data-parallel compiler requires certain structural properties in the source program, or conversely, the absence of others. We have encountered a number of difficulties in applying Fortran D to Gromos, a popular dusty-deck program for molecular dynamics, that probably are neither limited to Gromos nor do they seem likely to be addressed by improved compiler technology in the near future. This parallelization effort motivated this correspondence where we present some guidelines for engineering data-parallel applications that are compatible with Fortran D or High Performance Fortran compilers. Our experience with Gromos suggests a number of points to keep in mind when developing software that may at some time in its life cycle be parallelized with a data-parallel compiler. One concern often not foremost in a scientific programmer's mind at the outset of software development is parallelization. Yet, even for scientific applications developed ..