Hardware Constraints in Computing - Exploiting Locality, Parallelism, and Operation Specialization

Explore fundamental hardware design constraints in this conference talk from the Simons Institute's Algorithmic Foundations for Emerging Computing Technologies Boot Camp. Discover the often-overlooked factors that limit computing hardware, from the fundamental properties of CMOS gates including their magic of no static power consumption and cost considerations like minimum operating voltages, to the reality that building domain-specific hardware doesn't automatically improve application efficiency. Learn how both performance and energy efficiency depend on exploiting three key characteristics in order of importance: locality, parallelism, and operation specialization, which apply to both individual logic gates and overall system design. Understand why applications and algorithms lacking these characteristics cannot benefit from domain-specific hardware, and examine how achieving desired performance and energy efficiency gains requires modifying applications to provide greater locality and parallelism for hardware to exploit. Gain insights into fundamental constraints on hardware design across any technology, from devices using electric fields to control current flow to large-scale machine learning accelerators, while exploring how cost considerations affect hardware design decisions in almost all systems.