high performance computing on graphics processing units: hgpu.org

Much work has recently been reported in parallel GPU-based particle swarm optimization (PSO). Motivated by the encouraging results of these investigations, while also recognizing the limitations of GPU-based methods for big problems using a large amount of data, this paper explores the efficacy of employing other types of parallel hardware for PSO. Most commodity systems feature a variety of architectures whose high-performance capabilities can be exploited. In this paper, high-dimensional problems and those that employ a large amount of external data are explored within the context of heterogeneous systems. Large problems are decomposed into constituent components, and analyses are undertaken of which components would benefit from multi-core or GPU parallelism. The current study therefore provides another demonstration that "supercomputing on a budget" is possible when subtasks of large problems are run on hardware most suited to these tasks. Experimental results show that large speedups can be achieved on high dimensional, data-intensive problems. Cost functions must first be analysed for parallelization opportunities, and assigned hardware based on the particular task.