A Capability-Based Massive Memory Computer

Abstract

ABSTRACf Conventional supercomputers gain their speed from the use of complicated and expensive multi-stage processor designs and/or the employment of a large number of simple processors working in parallel. This paper investigates another approach, based on using a massive main memory (in the order of gigabytes). The paper discusses the advantages of this approach in supporting database applications, VLSI applications and many other applications working on large volumes of data. It is shown how the architecture of the MONADS-PC system, a capability-based computer developed in Australia by the authors, can be adapted to support such a large memory. The architectural design of a new machine based on MONADS-PC is given, with special emphasis on the addressing and address translation issue.