Optimized SCC Processor by Using Parallel Genetic Algorithms

In this paper, we show how to optimize the communications between processes on the SCC processor. The Scc, Single-chip Cloud Computer was designed by Intel, it is a 48-core processor. The performances of the communications between the cores depend on theirs positions on the processor. For a same parallel application on the SCC, the distribution of the processes on the cores could impact the general performance. The goal of this paper is to show the possibilities of optimizing the communications by changing the positions of processes on the cores. For this, we use benchmarked applications by experts and meta-heuristic optimization methods like parallel genetic algorithms. This paper details also how the genetic algorithm's parameters influence the convergence ant how to improve the performances by compressing the data.