(9th workshop on parallel systems and algorithms – workshop of the GI/ITG special interest groups PARS and PARVA, 2008) Keller, Jörg; Kessler, Christoph; König, Kalle; Heenes, Wolfgang
Sorting large data sets has always been an important application, and hence has been one of the benchmark applications on new parallel architectures. We present a parallel sorting algorithm for the Cell processor that combines elements of bitonic sort and merge sort, and reduces the bandwidth to main memory by pipelining. We present runtime results of a partial prototype implementation and simulation results for the complete sorting algorithm, that promise performance advantages over previ- ous implementations.