An Integer Set Framework for HPF Analysis and Code Generation

The core of the Rice dHPF compiler for High Performance Fortran is a practical, executable, equational framework for data parallel program analysis and optimization. This framework, based on abstract operations on sets of integers, greatly simplifies the implementation of many key analysis, optimization and code generation tasks (even with a general computation partitioning model), and yet enhances their generality and flexibility. This paper describes the equational framework, tradeoffs in its implementation, the formulation and implementation of important analyses and optimizations within the framework, and a hierarchical code generation strategy to support the framework. 1 Introduction Data-parallel languages like High-Performance Fortran (HPF)[21] hold out the promise of a high-level, portable, parallel programming model usable by non-computer-scientists. To achieve wide acceptance, such a language will require parallelizing compilers that can extract consistently high performance..