Portable parallel programming on emerging platforms

Clusters of symmetric shared memory multiprocessors (SMPs) are fast becomin g a highly available platform for parallel computing. There is a need for a uniform programming paradigm that allows users to transparently extend par allelism across multiple SMP nodes. A shared memory paradigm leverages the available hardware to handle sharing within an SMP, in addition to providin g programming ease. Software distributed shared memory systems support the illusion of shared memory across the cluster via a software runtime layer b etween the application and the hardware. This approach can potentially prov ide a cost-effective alternative to larger hardware shared memory systems f or executing certain classes of workloads. We describe here one such system and discuss its interface, performance and portability through an example real-world application from the scientific domain.