Ant Colony Optimization Based NoCs for Flexible Spatial Isolation in Mixed Criticality Systems

Abstract

Multiple applications of different criticality are increasingly being executed on the same System-on-Chip platform to reduce cost. To reduce influence of non-critical applications on critical applications, popular techniques utilized include spatial partitioning of resources. The Network-on-Chip (NoC) in such systems can aid by spatially isolating network traffic within the partitions. Topologies in such regions can be regular or irregular and can require determining of non-minimal paths. Thereby making computation of routes at runtime challenging. In this paper, we present a self-learning NoC consisting of a flexible adaptive routing algorithm based on Ant Colony Optimization (ACO) metaheuristic. The NoC discovers minimal and non-minimal routes within a region at runtime and optimizes for latency. The proposed solution does not incur software overhead and is topology agnostic. We present the potential of utilizing such an algorithm to discover new routes at hardware level and aid in spatial isolation of traffic.