Our method, which is fully parallelized, uses conventional two-sided message passing and thus avoids the overhead of a standard implementation of direct remote memory access based on one-sided communication.
The main prerequisite is that the given program is structured in a BSP-compliant way. NestStep is a parallel programming language for the BSP (bulk-synchrono= us parallel) programming model. We describe the concept of distributed sha= red arrays in NestStep and its implementation for an MPI-based PC cluster. In particular, we describe a novel method for run-time scheduling of direct remote access to distributed shared array sections. This method allows to replace a standard implementation of direct remote memory access based on one-sided communication by using the more efficient two-sided communication statements, provided that the given program is structured in a BSP-compliant way.