A Parallel Heuristic for Bandwidth Reduction Based on Matrix Geometry

This paper proposes a parallel hybrid heuristic aiming the reduction of the bandwidth of sparse matrices. Mainly based on the geometry of the matrix, the proposed method uses a greedy selection of rows/columns to be interchanged, depending on the nonzero extremities and other parameters of the matrix. Experimental results obtained on an IBM Blue Gene/P supercomputer illustrate the fact that the proposed parallel heuristic leads to better results, with respect to time efficiency, speedup, efficiency and quality of solution, in comparison with serial variants and of course in comparison with other reported results.