An experimental analysis of Hager’s exchange methods in conjunction with heuristics for bandwidth and profile reductions applied to accelerate the ICCG method

Authors

  • Sanderson L. Gonzaga de Oliveira
  • Júnior Assis Barreto Bernardes

DOI:

https://doi.org/10.5540/03.2018.006.02.0248

Keywords:

Sparse matrices, graph labeling, graph algorithm, bandwidth reduction, profile reduction, conjugate gradient method.

Abstract

This paper evaluates the original Hager’s adjacent exchange methods in conjunction with heuristics for bandwidth and profile reductions with the objective of reducing computing times of the zero–fill incomplete Cholesky–preconditioned conjugate gradient method. The numerical results obtained in this computational experiment show that the original Hager’s exchange methods, although capable of reducing the profile of the instances, are not useful when reducing processing times of the zero–fill incomplete Cholesky–preconditioned conjugate gradient method.

Downloads

Download data is not yet available.

Published

2018-12-19

Issue

Section

Trabalhos Completos