BSP Sorting: An experimental Study

The Bulk-Synchronous Parallel model of computation has been used for the architecture independent design and analysis of parallel algorithms whose performance is expressed not only in terms of problem size n but also in terms of parallel machine properties. In this paper the performance of implementations of deterministic and randomized BSP sorting algorithms is examined. The randomized algorithm is a variant of the algorithm in [19], and the deterministic algorithm is the one in [20, 26]; both algorithms are designed to handle duplicate keys with optimal efficiency. The two algorithms have been implemented in ANSI C and their performance (scalability and efficiency issues) has been studied on a distributed memory machine, a Cray T3D. The experimental results obtained from these implementations are reported in this work. The validity of the theoretical model is also test; based on the theoretical performance of each algorithm under the BSP model and the BSP parameters of a Cray T3D, it is possible to estimate the actual performance of the implementations based on the theoretical performance of the designed algorithms. The experimental results reinforce the claims that the BSP model can be usefully used as a programming model or, indeed, as a kind of a programming