Data parallelism with high performance C

This paper describes a preliminary design of the High-Performance C (HPC) language [2]. HPC, a machine-independent language extension to C, allows the user to write programs for distributed-memory systems using global addresses. HPC includes high-level features for specifying processor arrays and both static and dynamic data distributions across processors and for formulating explicitly loops with independent iterations. The language is based on the data parallel paradigm, which exploits the parallelism inherent in many scientific applications. This paper focuses on HPC features that are beyond High-Performance Fortran (HPF) in the areas of sequential and data parallel extensions.