Abstract de la publi numéro 5737

We consider the direct solution of sparse symmetric indefinite matrices. We develop new pivoting strategies that combine numerical and static pivoting. Furthermore, we propose original approaches that are designed for parallel distributed factorization. We show that our pivoting strategies are numerically robust and that the factorization is significantly faster because of this static/numerical combination. A key point of our parallel implementation is the cheap and reliable estimation of the growth factor. This estimation is based on an approximation of the off-diagonal entries and does not require any supplementary messages.