Portable Runtime Support and Exploitation of Nested Parallelism in OpenMP

In this paper, we present an alternative implementation of the NANOS OpenMP runtime library (NthLib) that targets portability and efficient support of multiple levels of parallelism. We have implemented the runtime libraries of available open-source OpenMP compilers on top of NthLib, reducing thus their overheads and providing them with inherent support for nested parallelism. In addition, we present an experimental implementation of the workqueuing model and the parallelization of a data clustering algorithm using OpenMP directives. The asymmetry and non-determinism of this algorithm necessitate the exploitation of its nested loop-level parallelism. The experimental results on a SMP server with four processors demonstrate our efficient OpenMP runtime support.