It was proposed that a robust and efficient parallelizable preconditioner for solving general sparse linear systems of equations, in which the use of sparse approximate inverse (AINV) techniques in a multi-level block...It was proposed that a robust and efficient parallelizable preconditioner for solving general sparse linear systems of equations, in which the use of sparse approximate inverse (AINV) techniques in a multi-level block ILU (BILUM) preconditioner were investigated. The resulting preconditioner retains robustness of BILUM preconditioner and has two advantages over the standard BILUM preconditioner: the ability to control sparsity and increased parallelism. Numerical experiments are used to show the effectiveness and efficiency of the new preconditioner.展开更多
A new cyclic pseudo-elimination(CPE,in brief)preconditioner,which combines the pseudoelimination(PE)technique and LU factorization together,is proposed for a kind of cyclic structured matrices.For the case of M-matric...A new cyclic pseudo-elimination(CPE,in brief)preconditioner,which combines the pseudoelimination(PE)technique and LU factorization together,is proposed for a kind of cyclic structured matrices.For the case of M-matrices,some theoretical results of convergence and estimation of the condition number are presented.Numerical experiments show that the CPE preconditioner performs the best with respect to the reduction of number of iterations.Moreover,it costs much less time than the ILUT and block Jacobi(BJ)preconditioners in a whole in all tested cases.展开更多
文摘It was proposed that a robust and efficient parallelizable preconditioner for solving general sparse linear systems of equations, in which the use of sparse approximate inverse (AINV) techniques in a multi-level block ILU (BILUM) preconditioner were investigated. The resulting preconditioner retains robustness of BILUM preconditioner and has two advantages over the standard BILUM preconditioner: the ability to control sparsity and increased parallelism. Numerical experiments are used to show the effectiveness and efficiency of the new preconditioner.
基金Supported by the National Natural Science Foundation of China(Grant Nos.61472462,11671049)Science Challenge Project(Grant No.TZ2016002)the CAEP Developing Fund of Science Technology(Grant No.CX20210021)。
文摘A new cyclic pseudo-elimination(CPE,in brief)preconditioner,which combines the pseudoelimination(PE)technique and LU factorization together,is proposed for a kind of cyclic structured matrices.For the case of M-matrices,some theoretical results of convergence and estimation of the condition number are presented.Numerical experiments show that the CPE preconditioner performs the best with respect to the reduction of number of iterations.Moreover,it costs much less time than the ILUT and block Jacobi(BJ)preconditioners in a whole in all tested cases.
