SHF: Small: Rethinking CAD for System-Level Design via Interactivity, Learning, and Collaboration

Columbia University, New York NY

Investigators

Abstract

In a zeal to pack more transistors into a single integrated circuit, the semiconductor industry has produced many generations of electronic products that are increasingly more powerful and less expensive resulting in massive expansion of information and communication technologies with economic benefits that are estimated as a large part of global productivity growth during the recent decades. A continuous stream of innovations across the three main semiconductor technology areas lies at the root of this success story: (i) device and manufacturing, (ii) circuits and architectures, and (iii) Computer-Aided Design (CAD) methodologies and tools. Recent innovations in the first two areas make it possible to envision new generations of products that are stronger in computational power and richer in functional capabilities. The research enabled by this project, attempts to undertake corresponding innovations in the third area mentioned above. A core part of this program is the development of a new senior-level undergraduate course that will engage the students in working together on a semester-long project with the infrastructure support for collaborative engineering provided by the new CAD environment. Current effort of designing state-of-the-art integrated circuits, suffers from shortcomings that characterize current CAD flows, e.g., the lack of support for exploring the design space across its hardware and software components in search of better solutions; the limited capabilities in optimizing simultaneously multiple components at the system level; and the big learning curve and long execution times that many tools require. This project will investigate a new generation of CAD methodologies and tools. Specifically, the research team will develop a new environment for System-Level Design that is highly interactive, exalts the designers' experience, inspires their creativity, and promotes continuous collaboration across the engineering team. The proposed approach leverages recent work on supervised and compositional design-space exploration, and the application of learning methods to assist designers in the use of advanced high-level synthesis tools.