Requirements for a Research-oriented IC Design System

Abstract

Computer-aided design techniques for integrated circuits grown in an incremental way, responding to various perceived needs, so that today there are a number of useful programs for logic generation, simulation at various levels, test preparation, artwork generation and analysis (including design rule checking), and interactive graphical editing. While the design of many circuits has benefitted from these programs, when industry wants to produce a high-volume part, the design and layout are done manually, followed by digitizing and perhaps some graphic editing before it is converted to pattern generation format, leading to the often heard statement that computer-aided design of integrated circuits doesn't work. If progress is to be made, it seems clear that the entire design process has to be thought through in basic terms, and much more attention must be paid to the way in which computational techniques can complement the designer's abilities. Currently, it is appropriate to try to characterize the design process in abstract terms, so that implementation and technological biases don't cloud the view of a desired system. In this paper, we briefly describe the conversion of algorithms to masks at a very general level, and then describe several projects at MIT which aim to provide contributions to an integrated design system. It is emphasized that no complete system design exists now at MIT, and that we believe that general design considerations must constantly be tested by building (and rebuilding) the various subcomponents, the structure of which is guided by our view of the overall design process.