Transferring software engineering methods to VLSI-design: a statistical approach

Abstract

The possibilities of VLSI technology grow by a factor of 100 per decade. In parallel, design efficiency has been increased only by a factor 30 per decade, so major problems in the development of complex systems have been shifted from controlling new technologies to the management of the design process itself. Classical design methods tend to fail for VLSI circuits with increasing complexity, often due to communication problems. To avoid redesigns, the adoption of methods of neighbour discipline software engineering seems to be a good choice. This paper discusses the influence of these methods on the quality of VLSI designs in terms of errors per design and costs per error detection. Statements and experiences are based on statistical data collected during several developments of complex systems including VLSI components.<>