I have a system of linear equations in matrix form A*x=B. A is square, sparse and large (several million rows). The matrix A is complex, nonsymmetric, and indefinite. I am trying to solve this system by iterative methods - QMR, BiCG. The A matrix is somewhat reasonably well conditioned, but it could be better conditioned (I assume). What would be a good preconditioner for such a matrix? Is preconditioning of such a matrix even possible? Due to the size of the A matrix I cannot use any kind of factorization because factorization returns a half-full matrix and this gives me "out of memory" problem. Any ideas on how to make the methods converge faster?
Than you in advance.