An Architecture of a Parallel Grid Multiprocessor

Abstract

An architecture is described of a specialized PGM (Parallel Grid Multiprocessor) supercomputer directed toward efficient solution of large scale numerical problems by grid methods for partial differential equations of different types. The project is based on intermediate mapping of the structure of computations and data streams characteristic for grid methods into the hardware architecture. The organization of two important PGM subsystems is described, namely the distributed interconnection network, and the set of parallel functional modules.