>
> > what is clear is that your "A" is indefinite, so GMRES or common
> > iterative solvers can not works.
>
> That's not true, of course: GMRES works just fine for indefinite systems,
> as do many other iterative solvers. However, CG does not work with
> indefinite systems.



W, thanks for your correction :)


I just want to add this fact that: your preconditioners may not handle
diagonal singularities, e.g. with diagonal preconditioner zero diag. may
cause problem (scaling by 1/zero), further, ILU without pivoting also fails
for the case of zero diagonals.


Cheers

RT
_______________________________________________

Reply via email to