Distributed Computing for Signal Processing: Modeling of Asynchronous Parallel Computation. Appendix B. Design of the Operating System for the PASM Parallel Processing System

摘要

As a result of advances in microcomputer technology, it is now feasible to build large-scale parallel processing systems capable of performing image processing tasks more rapidly than previously possible. Such parallel processing systems add levels of complexity for both the operating system and the application software. They impose constraints that make a direct transplantation of conventional (multiprogrammed) operating systems extremely inefficient (Bae80). This thesis considers the design of PASMOS, a distributed operating system for the PASM parallel processing system. PASM is a reconfigurable multimicrocomputer system which is being designed at Purdue University for image processing and pattern recognition applications. The special purpose nature of PASM has been exploited in the design of PASMOS. PASMOS has a hierarchical structure and is distributed throughout the hardware components of PASM. It utilizes the PASM hardware to create an execution environment (virtual machine) for parallel processing tasks. Facilities provided by PASMOS include those for task management and scheduling, memory management, user interaction, process communication and synchronization, and protection. The general performance of the PASM system has been examined both analytically and via simulation.