The Endgame for Moore's Law: Architecture, Algorithm, and Application Challenges

Abstract

Single processor clock speed scaling ended a decade ago, and transistor sizes will approach atomic scales in the next decade. With no abatement in the ideas of how to use more computing in science, engineering and business applications, and new performance drivers coming from increased density, speed and ubiquity of data collection devices, how will the demands for computing be met? Future computing system designs will be constrained by power density and total system energy, and the two performance approaches are likely to be parallelism and specialization. Data movement already dominates running time and energy costs, making communication cost reduction the primary optimization criteria for compilers and programmers.