Generalizing block FSAI preconditioning to unsymmetric indefinite matrices

The efficient solution to non-symmetric linear systems is still an open issue on parallel computers. In this short paper we generalize to the non-symmetric case the Block Factorized Sparse Approximate Inverse (BFSAI) preconditioner which has already proved very effective on symmetric positive definite (SPD) problems. The proposed algorithm is experimented with in a number of large size matrices showing both a good robustness and scalability.