Integrating task and data parallelism with taskHPF

Many applications exhibit a large amount of potential parallelism that can be exploited at both data and task levels. In this paper, we consider applications which can be structured as ensembles of independent data parallel HPF modules (here-after HPF tasks), which interact according to static and predictable patterns. In order to make easy and effective to program such applications, we devised taskHPF, a coordination language in which programmers can define the interaction patterns among HPF tasks in a declarative way. We examine a small example application to discuss the benefits of our approach, and we show how taskHPF programs can be translated into efficient message-passing code.