A multi-microprocessor system for parallel computations

Abstract

This paper is concerned with the design of a multi-microprocessor system that exploits parallelism in ordinary high-level language programs so as to increase their execution speed, without adding too much complexity to the system. The notion of Parallel Execution Strings (PES) as a representation of expressions for parallel execution is first discussed. We then apply the PES scheme to detect the parallelism at both the statement and the block levels. Finally, the design of a multi-microprocessor system suitable for executing the programs compiled with the PES notion is presented. It consists of a number of microprocessors (Am2900) with central control of moderate complexity.