Theory, techniques, and experiments in solving recurrences in computer programs.

摘要

The study of theoretical and practical issues in automatic parallelization across application and language boundaries is an appropriate and timely task. In this work, we discuss theory and techniques that we have determined useful in solving recurrences in computer programs. In chapter two we present a framework for understanding parallelism in computer applications based on an approach which models loop bodies as coalescing loop operators. In chapter three, we perform a case study of a modern C{dollar}sp{lcub}++{rcub}{dollar} semantic retrieval application drawn from the digital library field based on the model presented in chapter two. In the fourth chapter we present a summary of several techniques that we believe can be applied in the automatic recognition and solution of recurrences. The techniques have been developed through performing a manual analysis of applications from benchmark suites which include sparse, irregular, and regular Fortran codes. In chapter five we discuss the application of the techniques developed in chapter four on a suite of Fortran codes representative of sparse and irregular computations which we have developed as part of this work. In the sixth chapter, we consider the application of these same techniques focused on obtaining parallelism in outer time-stepping loops. In the final chapter, we draw this work to a conclusion and discuss future directions in parallelizing compiler technology.