Novel semiconductor device approach for finding acceptable solutions to classes of graph-based computing problems not otherwise possible

Abstract

The framework and practical applications of ongoing efforts to design novel classes of semiconductor devices making it possible to address highly complex problems that can not be currently solved in real-time, is presented. Important classes of such problems are identified and characterized by massively parallel and distributed graph-based architectures of ever-increasing complexity, national, and global scope. The etiology of the inability of current digital microprocessor-based computing schemes and devices to address such complex real-time problems is briefly reviewed. An overview of novel classes of information processing semiconductor devices, designated as “system-on-a-chip” devices that overcome these limitations, is given. Emanating from the application of a new graph-based methodology, they are designed to find acceptable domain-specific solutions in real-time. Examples and simulation results illustrate key concepts and capabilities of these devices for planning classes of problems.