Sparse Approximate Inverse Preconditioner With Parametric Sparsity Pattern Applied to the Macrobasis Function Methods

An efficient approach for the generation of sparse preconditioners, tailored for the application of macrobasis function methods, is described in this letter. The preconditioner is highly scalable and generated by the computing groups of rows that share a common sparsity pattern, calculated by using a distance threshold between geometrical blocks that controls the amount of data to be considered. This approach is suitable for the analysis of very large and realistic problems, as shown in some test cases included. Due to the clustering nature of this technique, the preconditioners can be generated in a substantially lower amount of time than that required by conventional approaches.