Effective openMP extensions for irregular applications on cluster environments

摘要

Sparse and unstructured computations are widely used in Scientific and Engineering Applications. Such problem inherent in sparse and unstructured computations is called irregular problem. In this paper, we propose some extensions to OpenMP directives, aiming at efficient irregular OpenMP codes to be executed in parallel. These OpenMP directives include scheduling for irregular loops, inspector/executor for parallelizing irregular reduction, and eliminating ordered loops. We also introduce implementation strategies with respect to these extensions.