| Summary: | We investigate the capabilities and shortcomings of Orca, a Modulalike parallel programming language supporting shared data objects on distributed memory platforms, by examining implementations of five nontrivial parallel applications: game tree searching, active chart parsing, image skeletonization, simulation of a chaotic predator/prey system, and polygon overlay. 1 Introduction Twenty years ago, a small number of visionaries believed that massive parallelism was the future of high-performance computing. Five years ago, that view had become commonplace; today, users are increasingly sceptical of such claims. The main reason is that parallel computers have remained very difficult to program. Many new programming systems have been developed to support parallel programming [7]. However, while comparisons of the performance of highlyparallel hardware are common [33], there have been only a few comparisons or assessment of the strengths and weaknesses of parallel programming systems. [1].
|
|---|