Combinatorial preconditioning for accelerating the convergence of the parallel block Jacobi method for the symmetric eigenvalue problem

Abstract

In this paper, we propose combinatorial preconditioning to accelerate the convergence of the parallel block Jacobi method for the symmetric eigenvalue problem. The idea is to gather matrix elements of large modulus near the diagonal prior to each annihilation by permutation of rows and columns and annihilate them at once, thereby leading to large reduction of the off-diagonal norm. Numerical experiments show that the resulting method can actually speedup the convergence and reduce the execution time of the parallel block Jacobi method.