Architectural considerations, software support, and compiler issues in multi-computer implementation. Final report, 5 September 1985-31 October 1988
Dharma Prakash Agrawal, Jon Mauney · OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) · 1988
The authors studied the problem of efficient execution of programs on parallel computers -- and on the B-HIVE architecture in particular --- and the software support required, concentrating on programming language implementation. In a loosely-coupled multiprocessing environment, in which processors communicate over a generalized hypercube or similar network, the cost of sharing data among processes is quite high. Techniques of minimizing the total communication cost in parallel programs were studied. The authors also looked into ways of introducing parallelism in combinatorial problems and observed the impact of randomization on the system speedup. Besides the speedup, the network connectivity and data distribution in parallel and distributed systems play a very important role in determining the system performance. Reliability of such systems under different environments was computed. In addition, the design of topologies with limited connections, which could be appropriate for both LAN and MAN applications was studied.