Abstract:

We used a new architecture of graphic processing unit (Fermi GPU) to re\|evaluate the feasibility and effectiveness of using programmable GPU to accelerate the standard particle swarm optimization (SPSO) algorithm and made a systematical analysis on the rules of harnessing the power of Fermi GPU, and implemented the parallel version of SPSO on such an architecture. In the experiment, we achieved a good speedup via  the best use of the new architecture of graphic processing unit,  compared to using the other parallel version of the algorithm.

Key words: standard particle swarm optimization (SPSO) algorithm, Fermi architecture, graphic processing unit (GPU)