Cluster-based Simulated Annealing for Mapping Cores onto 2D Mesh Networks on Chip

Authors:
Zhonghai Lu;Lei Xia;Axel Jantsch
Affiliations:
Dept. of Electronic, Computer and Software Systems, Royal Institute of Technology (KTH), Stockholm, Sweden, zhonghai@kth.se;Dept. of Electronic, Computer and Software Systems, Royal Institute of Technology (KTH), Stockholm, Sweden, leix@kth.se;Dept. of Electronic, Computer and Software Systems, Royal Institute of Technology (KTH), Stockholm, Sweden, axel@kth.se
Venue:
DDECS '08 Proceedings of the 2008 11th IEEE Workshop on Design and Diagnostics of Electronic Circuits and Systems
Year:
2008

Citing 0
Cited 5

A power-aware mapping approach to map IP cores onto NoCs under bandwidth and latency constraints

ACM Transactions on Architecture and Code Optimization (TACO)
Analytical derivation of traffic patterns in cache-coherent shared-memory systems

Microprocessors & Microsystems
t(k)-SA: accelerated simulated annealing algorithm for application mapping on networks-on-chip

Proceedings of the 14th annual conference on Genetic and evolutionary computation
A survey on application mapping strategies for Network-on-Chip design

Journal of Systems Architecture: the EUROMICRO Journal
Developing Domain-Knowledge Evolutionary Algorithms for Network-on-Chip Application Mapping

Microprocessors & Microsystems

Quantified Score

Hi-index	0.00

Visualization

Abstract

In Network-on-Chip (NoC) application design, core-to-node mapping is an important but intractable optimization problem. In the paper, we use simulated annealing to tackle the mapping problem in 2D mesh NoCs. In particular, we combine a clustering technique with the simulated annealing to speed up the convergence to near-optimal solutions. The clustering exploits the connectivity and distance relation in the network architecture as well as the locality and bandwidth requirements in the core communication graph. The annealing is cluster-aware and may be dynamically constrained within clusters. Our experiments suggest that simulated annealing can be effectively used to solve the mapping problem with a scalable size, and the combined strategy improves over the simulated annealing in execution time by up to 30% without compromising the quality of solutions.