Embedded Scalable Platforms for Heterogeneous Computing

Speaker: Luca Carloni, Columbia University

Abstract

Information technology has entered the age of heterogeneous computing. Across a variety of application domains, computing systems rely on highly heterogeneous architectures that combine multiple general-purpose processors with specialized hardware accelerators. The complexity of these systems, however, threatens to widen the gap between the capabilities provided by semiconductor technologies and the productivity of computer engineers. Embedded Scalable Platforms (ESP) is a new approach to the design and programming of heterogeneous computing systems that raises the level of abstraction in the design process and provides a framework for collaboration between software developers and hardware designers. ESP combines a flexible tile-based socketed architecture with a companion system-level design methodology. By relying on scalable communication infrastructures, the architecture simplifies the integration of heterogeneous components. By leveraging high-level synthesis, the methodology simplifies the development of system prototypes that speed up both design validation and software development. In this lecture, I present first the foundations of ESP and then the latest advances, including dynamic support of heterogeneous cache-coherence models for accelerators.