Chips, Architectures and Algorithms: Reflections on the Exponential Growth of Digital Signal Processing Capability
Mark A. Richards, Gary A. Shaw · 2003
The celebrated 1965 prediction by Gordon Moore regarding exponential improvements in integrated circuit density is so widely known, and has proven so accurate, that it has been elevated to the status of a “law”. Some pundits have even touted the unprecedented growth in IC performance as a major contributor to the sustained economic boom of the late twentieth century. Less well known and appreciated is the fact that many areas of computation have benefited equally from progress in algorithms. In this paper we compare and contrast the contributions to progress in signal processing capability of “hardware” (integrated circuits and computer architectures) and “software” (functionality and algorithms). While acknowledging the enormous impact that Moore’s Law growth in hardware performance has had on signal processing applications, we make the case that algorithm improvements, in the form of improved functionality and implementation, have contributed significantly to past progress as well. We discuss important differences in the ways that progress in the hardware and software realms contribute to signal processing capability, and describe how signal processing application developers tend to draw successively on each different domain of improvement as their product designs evolve over time. Finally, we argue that sustaining exponential growth in signal processing applications in the future depends more than ever on sustaining innovation in software and algorithm development.