Parallel computers used to be, for the most part, one-of-a-kind systems which were extremely difficult to program portably. With SMP architectures, the advent of the POSIX thread API and OpenMP gave developers ways to portably exploit on-the-box shared memory parallelism. Since these architectures didnt scale cost-effectively, distributed memory clusters were developed. The associated MPI message passing libraries gave these systems a portable paradigm too. Having programmers effectively use this paradigm is a somewhat different question. Distributed data has to be explicitly transported via the messaging system in order for it to be useful. This paper will present pyMPI, a distributed implementation of Python extended with an MPI interface. The tool makes it easy to write parallel Python scripts for system administration, data exploration, file post-processing, and even for writing full blown scientific simulations. Parallel Python also allows developers to prototype the data distribution for parallel algorithms in a easy, interactive, and intuitive manner without having to compile code, build specialized MPI types, and build serialization mechanisms. pyMPI supports most of the MPI API. It allows access to sends, receives, barriers, asynchronous messaging, communicators, requests, and status. In short, it provides a fully functional parallel environment coupled with a powerful scripting engine. The combination simplifies the generation of large scale, distributed tools for clusters.
PDF
download
878987797
028-85220240
