Motivated by the paper , where the authors proposed a method to solve a symmetric positive definite (SPD) system Ax = b via a sparse-sparse iterative-based projection method,
we extend this method to nonsymmetric linear systems and propose a modified method to construct
a sparse approximate inverse preconditioner by using the Frobenius norm minimization
technique in this paper. Numerical experiments indicate that this new preconditioner appears
more robust and takes less time of constructing than the popular parallel sparse approximate
inverse preconditioner (PSM) proposed in .