Nanophotonic interconnection networks for multicore embedded computing systems

Abstract

Today's microprocessors are driven in performance by increases in the number of on-chip parallel computational cores rather than by the sheer acceleration of transistor speeds. These emerging chip multiprocessors (CMPs) are fundamentally different from traditional unicore microprocessors with architectures that essentially resemble highly parallel computing systems on-chip. Achieving maximum utilization of compute resources as the number of functional parallel units scales, falls increasingly on the efficiency of the information exchange among those resources. In this new communication-bound paradigm, the realization of a system-wide scalable communications infrastructure that can meet the enormous bandwidths, capacities, and stringent latency requirements in an energy efficient manner is a key goal for scaling future computation performance. We explore how recent advances in nanoscale silicon photonic technologies can be exploited for developing optical interconnection networks that address the critical bandwidth and power challenges presented for CMP computing systems.