Efficient oblivious parallel sorting on the MasPar MP-1
Klaus Josef Brockmann, Rolf Wanka · 2002
The authors address the problem of sorting a large number N of keys on a MarPar MP-1 parallel SIMD machine of moderate size P where the processing elements (PEs) are interconnected as a toroidal mesh and have 16 KB local storage each. They present a comparative study of implementations of the following deterministic oblivious sorting methods: bitonic sort, odd-even merge sort, and fastsort. They successfully use the guarded split&merge operation introduced by Rub (1995). The experiments and investigations in a simple, parameterized analytical model show that, with this operation, from a certain ratio N/P upwards both odd-even merge sort and fastsort become faster on average than the up to the present fastest, sophisticated implementation of bitonic sort by Prins (1991). Though it is not as efficient as odd-even merge sort, fastsort is, to their knowledge, the first method specially tailored to the mesh architecture that can be, when implemented, competitive on average with a mesh-adaptation of bitonic sort for large N/P.