Evaluation of the Cell Broadband Engine running Continuous Estimation of Distribution Algorithms

Current consumer-grade computers and game devices incorporate very powerful processors that can be used to accelerate many classes of scien- tific codes. However, programming multi-core chips, hybrid multi-processors or graphical processing units is not an easy task for those programmers that deal mainly with sequential codes. In this paper, we ex- plore the ability of the Cell Broadband Engine to run a particular Estimation of Distribution Algorithm (Univariate Marginal Distribution Algorithm for the Continuous domain). From an initial sequential ver- sion, we develop a multi-threaded one that is after- wards reworked to run on a Cell. Both versions of the code show significant improvements in performance, compared to the sequential version. We analyze the results obtained and provide some clues about the performance/cost characteristics of the tested plat- forms.