9626215 Zheng Massively parallel computation is a frontier computing field. An extremely important aspect of parallel computation is data communication among processing/memory components. Much research has been centered around building point-to-point networks using the VLSI technology. The VLSI technology is not suitable for interconnection intensive circuits due to its two- dimensionality, I/O constraints, and electrical properties such as resistance, capacitance, and inductance. The emerging optical technology has drastically changed the landscape of interconnection schemes. There are many desirable characteristics of optical interconnects. Optics can utilize free-space interconnects as well as guided wave technology, neither of which has the problems mentioned for the VLSI technology. The characteristics of optical interconnects have significant system configuration and complexity implications. Architectural freedom and constraints have shifted when optical interconnects are incorporated. The conventional graph theory, which has been successfully used to characterize point-to-point networks, is no longer adequate for modeling these interconnect technologies. Recently, based on the hypergraph theory, the P.I. has introduced a new class of network architectures, the hypernetworks, that are particularly suitable for optical interconnects. This project will focus on the feasibilties of implementing large-scale optical interconnection networks using current and future optical technologies. Topological, communication, computation, fault-tolerance, and implementation aspects of optical hypernetworks will be investigated. During this project, the P.I. will interact with researchers in the field of optical engineering, and develop a joint research plan for further investigations. It is possible that the proposed research will have a significant impact not only on the design, implementation and analysis of new generation massively parallel computing systems based on current and future V LSI and optical technologies, but also on the theory and practice of massively parallel computation as a whole. ***
