Heuristics for 1D Rectilinear Partitioning as a Low Cost and High Quality Answer to Dynamic Load Balancing

Several algorithms have been proposed for computing the optimal rectilinear partitioning of data to a linear array of processors. We introduce two fully parallel heuristics that compute sub-optimal partitions, in a more efficient way than the best known optimal algorithm. The goal of this paper is to compare our heuristics to an optimal partitioning, both in terms of execution time and accuracy of the partition. We give some very interesting theoritical bounds on the quality of our heuristics and we report results on numerical experiments and real applications.