Anais do Simpósio Brasileiro de Pesquisa Operacional
Anais do Simpósio Brasileiro de Pesquisa Operacional

Previous publications evaluated a large number of heuristics for bandwidth and profile reductions and selected 14 heuristics as promising low--cost heuristics for these problems. Based on this experience, this paper evaluates these heuristics when applied to matrices derived from finite--volume discretizations of the Laplace equation. Specifically, with the support of extensive numerical experiments, this paper shows that state--of--the--art low--cost heuristics for bandwidth and profile reductions do not reduce execution times of the zero--fill incomplete Cholesky--preconditioned conjugate gradient method when applied to matrices (contained in systems of linear equations) formerly ordered using a Sierpi\'nski-like curve.