Efficient implementation of a generalized Cholesky factorization for Symmetric Galerkin Boundary Element Methods.

In the present paper a natural extension of the Cholesky factorization, adapted to the peculiar matrix-structure of the solving system of Symmetric Galerkin Boundary Element Methods is presented. The implementation is described in some detail and preliminary timing figures are given, showing that the proposed approach is considerably faster than the Bunch-Kaufman factorization for non-definite matrices. Thread-based parallelism is exploited to show the scalability of the algorithm on moderately parallel shared memory multiprocessors. [ABSTRACT FROM AUTHOR]