A comparison between two optimisation alternatives for mapping in wireless network on chip

Abstract

Network on Chip (NoC) is a well known approach that aims at improving the performance of many-core systems. The design of such systems involves the optimal mapping of tasks to nodes, and the corresponding scheduling of the tasks at every node, which results in a challenging optimisation problem considering the constraints that need to be respected. In this paper, after formalising the problem and elaborating on its complexity, we present an AI approach to solve the problem and evaluate it against a MIP approach. Our empirical evaluation shows that the AI approach is able to obtain solutions of good quality very quickly.