On parallel random number generation for accelerating simulations of communication systems

[摘要] Powerful compute clusters and multi-core systems have become widely availablein research and industry nowadays. This boost in utilizable computationalpower tempts people to run compute-intensive tasks on those clusters, eitherfor speed or accuracy reasons. Especially Monte Carlo simulations with theirinherent parallelism promise very high speedups. Nevertheless, the quality ofMonte Carlo simulations strongly depends on the quality of the employedrandom numbers. In this work we present a comprehensive analysis ofstate-of-the-art pseudo random number generators like the MT19937 or the WELLgenerator used for parallel stream generation in different settings. Theserandom number generators can be realized in hardware as well as in softwareand help to accelerate the analysis (or simulation) of communicationssystems. We show that it is possible to generate high-quality parallelrandom number streams with both generators, as long as some configurationconstraints are met. We furthermore depict that distributed simulations withthose generator types are viable even to very high degrees of parallelism.

[学科分类] 电子、光学、磁材料