Parallel performance of block ILU preconditioners for a block-tridiagonal matrix (Q1395724)

The parallel implementation for Krylov subspace methods of the block ILU preconditioners for block-tridiagonal matrices proposed by the author [BIT 40, 583-605 (2000; Zbl 0961.65040)] is discussed. Especially an efficient implementation for a five-point discretisation of an elliptic second-order partial differential equation is discussed, and more generally, the effect of such a method on the parallelization of the Krylov method on a distributed memory machine is described. Several numerical examples show the effectiveness of the method and compare it with other methods.