VLSI Architectures for Supercomputing

Abstract

Very large scale integrated (VLSI) circuit chips containing 3 million transistors are being implemented, and microprocessor chips containing over 1 million transistors are commercially available. This chapter reviews some suitable architectural models which are being used for VLSI implementations. Then four categories of implementations (VLSI processors, VLSI arrays, wafer scale integration arrays, and Gallium arsenide computer systems) are reviewed to bring up to date current interests in VLSI applications to supercomputing. The VLSI architectural developments to be presented here render their execution model into one or more of those decentralized architectures in one way or another. The chapter discusses each in terms of execution sequence control. VLSI arrays have been a special kind of VLSI architectures for effectively solving supercomputing problems. VLSI architectures are the most fundamental element needed for engineering supercomputing. Centralized and decentralized architectures have been proposed, and some are implemented on various VLSI system formations in different technologies.