GA-MMAS: an Energy-aware Mapping Algorithm for 2D Network-on-Chip
Ning Wu, Yifeng Mu, Zhou Fang, Fen Ge · Lecture notes in computer science · 2011
IP core mapping for Network-on-Chip (NoC) at system level significantly impacts the communication energy consumption of the system. Optimization of IP mapping can lead to significant energy savings. In this paper, a new mapping algorithm named GA-MMAS is proposed based on Genetic Algorithm (GA) and MAX-MIN Ant System Algorithm (MMAS), to optimize energy consumption for NoC. Firstly in this algorithm, we improve MMAS. We obtain the elicitation information via priority mapping of IP core with larger communication volume, instead of using heuristics, to improve the optimal solution of MMAS. Then with the combination of MMAS and GA, the advantage of speed in GA makes compensation to the lack of pheromone in the early stage of MMAS, and in turn enhancing the accuracy of optimal solution, which leads to lower energy consumption. The experiments performed on various random benchmarks and a complex video/audio application to conform the efficiency of the algorithm. Experimental results show that the algorithm saves about 36%-60% of energy consumption compared to random mapping, and saves about 3%-25%, 10%-30% and 3%-30% compared to algorithm GA, Ant Colony Algorithm (ACA) and MMAS respectively. MAX-MIN Ant System Algorithm (MMAS) in (6) are applied to optimize the energy consumption for IP mapping. However ACA or MMAS will find the solution very slowly because of the lack of initial pheromone. In this paper, an algorithm named GA-MMAS is proposed for the problem of low energy consumption IP mapping based on researching on GA and MMAS. GA-MMAS firstly improves MMAS. Then it combines MMAS with GA to make compensation to their disadvantages by their advantages. Several experiments are carried out to verify the efficiency of the algorithm. The rest of the paper is organized as follows: Section II describes the problem of IP core mapping; Section III illustrates the improved MMAS. Section IV illustrates our algorithm; Section V presents the experimental results; and finally, Section VI concludes the paper and outlines some directions for future work.