The bulk-synchronous parallel random access machine

The model of bulk-synchronous parallel (BSP) computation is intended to provide a simple and realistic framework for general-purpose parallel computing. Originally, BSP was defined as a distributed memory model. In this paper we present a new model, called BSPRAM, which is a variant of BSP based on a mixture of shared and distributed memory. The two models are equivalent for some important classes of algorithms. We identify two such classes: oblivious and coarse-block algorithms. Finally, we present BSPRAM algorithms for dense matrix multiplication and Fast Fourier Transform.