Parallel block ILUT/ILDLT preconditioning for sparse eigenproblems and sparse linear systems
Achim Basermann · Numerical Linear Algebra with Applications · 2000
In this paper, a combination of parallel computing techniques and incomplete block LU preconditioning methods with threshold (ILUT) are presented to reduce the execution times of iterative solvers for sparse eigenproblems as well as sparse linear systems. The symmetric variant of ILUT, an incomplete block LDLT preconditioner (ILDLT), is investigated in addition. The parallelization of the solvers is based on matrix and vector partitioning with a data distribution and a communication scheme exploiting the sparsity of the matrix. The efficiency of the parallel preconditioned eigenproblem and sparse linear system solvers is demonstrated on the parallel systems NEC Cenju-4 as well as on a PC cluster. Copyright © 2000 John Wiley & Sons, Ltd.