Cluster - based Simu ate nnea ing for apin ores onto esh etworks on C 1

In Network-on-Chip (NoC) application design, coresearch, genetic algorithms and neural networks [1] are based to-node mapping is an important but intractable optimization on empirical techniques in nature. Simulated annealing (SA) problem. In the paper, we use simulated annealing to tackle the [7] is a generic probabilistic meta-algorithm for the global mapping problem in 2D mesh NoCs. In particular, we combine a clustering technique with the simulated annealing to speed up toptimizationpo lm amellocatina good aroximatio the convergence to near-optimal solutions. The clustering exploits to the global optimum of a given function in a large search the connectivity and distance relation in the network architecture space. It is inspired by annealing in metallurgy a technique as well as the locality and bandwidth requirements in the core involving heating and controlled cooling of a material to communication graph. The annealing is cluster-aware and may be increase the size of its crystals and reduce their defects. The dynamically constrained within clusters. Our experiments suggest t .n.a that simulated annealing can be effectively used to solve the heatcass th atoms obce untuck from thi inita mapping problem with a scalable size, and the combined strategy positions (a local minimum of the internal energy) and wander improves over the simulated annealing in execution time by up randomly through states of higher energy; the slow cooling to 30% without compromising the quality of solutions. gives them more chances of finding configurations with lower