Technical Report Series in Computer Science

Dipartimento di Informatica
Università Ca' Foscari di Venezia

Technical Report Series in Computer Science

CS-99-9

Mixed Data and Task Parallelism with HPF and PVM

S. Orlando, P. Palmerini, and R. Perego

Abstract

We present a framework to design efficient and portable HPF applications which exploit a mixture of task and data parallelism. According to the framework proposed, data parallelism is restricted within HPF modules, and task parallelism is achieved by the concurrent execution of several data--parallel modules cooperating through COLT_HPF, a coordination/communication layer implemented on top of PVM. COLT_HPF is portable and can be used independently of the HPF compilation system exploited, it allows instances of cooperating HPF tasks to be created either at loading time or at run-time. We claim that COLT_HPF can be easily used to implement HPF applications which exploit structured or unstructured forms of task parallelism on either homogeneous or heterogeneous computing environments. To validate our approach we present some experimental results obtained by running a physics application implemented by means of COLT_HPF on a cluster of Linux SMPs.

Return to technical report list.