Linear Time Implementation of Frontal Increase Minimization (FIM) Strategy Algorithms for the Profile Reduction of Sparse Matrices

摘要

An improved implementation of Frontal Increase Minimization strategy is described. The aim is to reduce computer time utilization, which is a major drawback of this type of algorithm. An insight into the numbering process and best appropriate data structures is found. It is shown that the time complexity of this implementation is linear with respect to the number of nonzero elements for all algorithms of this type. Numerical results highlight the improvement and the efficiency of the method for small problems.