Software development issues for parallel processing

Parallel architectures for computers give rise to several issues with respect to software development. The issue to be resolved depends on the life-cycle phase in which the problem appears. During problem analysis and program design, the issue is best resolved by using an intelligent language environment, exemplified by the Rational R1000 Ada machine. During the post-implementation phase, an existing program can be restructured by using an expert tool that generates directed graph models of the program, analyzes the models and the source to identify potentially parallelizable code sequences, and through workstation window-oriented displays guides the user through the process of restructuring a program to run efficiently on a parallel architecture machine. An eight-step methodology for reworking existing Fortran programs for more efficient execution on parallel architecture machines is presented.<<ETX>>