An Analytical Model for Hypercube Network-On-Chip Systems with Wormhole Switching and Fully Adaptive Routing

Abstract

In this paper, we present a novel performance analysis model for predicting average message latency of hypercube network-on-chip (NoC) systems that employ wormhole switching and fully adaptive routing method with uniformly distributed traffic. Unlike previous works, our model calculates the service rate that is provided by a physical link for the incoming traffic at a particular communicating node by obtaining the reversal service rate that is provided by the downstream nodes, when the network gets saturated. This model has simple closed-form calculations, and can produce accurate analytic results when the NoC system is operating in stable state. The proposed model is validated by comparing the results that are calculated by the model with the results that are obtained through simulations with different network configurations and traffic loads.