Convergence criteria for MLMG::BottomSolver::bicgstab

[ZiyangHuang]

Greetings,

I have a question of convergence of the MLMG::BottomSolver::bicgstab. The matrix I have is strictly diagonal dominant, and I know that the Gauss-Seidel iteration is guaranteed to converge. I wonder if MLMG::BottomSolver::bicgstab has the same property.

If I want to use the Gauss-Seidel iteration as the bottom solver, is the code below a correct way to do so?
mlmg.setBottomSolver(MLMG::BottomSolver::smoother);
where mlmg is an instance of MLMG.

Thanks!

[WeiqunZhang]

I don't know if bicgstab is in theory guaranteed to converge for diagonally dominant matrix. I suspect it is not. In my experience, if the initial guess is not good enough and the matrix is big, it may not converge.

Yes, that's the way to use a smoother at the bottom.