The era of embedded computing

Moore's Law, which states that microprocessor performance doubles approximately every 18 months (for a fixed cost), also implies that the cost of computing declines by about 35% per year (for a fixed performance). Over the past few decades, this has led to an inexorable downward drift in the center of gravity of computing, from mainframes down to PCs. In turn, this has led to huge, but cyclic, changes in computer architecture, accompanied by major dislocations in the computer industry. With the advent of system level integration (SLI) and system-onchip (SOC), the center of gravity of the computer industry is once again moving: from personal computing into embedded computing. The opportunities, needs and constraints of this next generation of computing are quite different from those to which we have got accustomed in general-purpose computing. This will lead to significantly different computer architectures, at both the system and the processor levels, and a rich diversity of off-theshelf and custom designs. Furthermore, embedded computing will introduce a new theme into computer architecture: the automation of architecture. In this talk, I shall discuss the rather predictable dynamics, both technology and market related, that have shaped the computer industry over the past few decades and that will, by extension, affect the next generation of embedded computing. I will also share my thoughts on the architecture of embedded computer systems and the automation of their Design.