[petsc-users] number of pre- and post- smoothing in multigrid as a preconditioner

[Barry Smith]

On May 23, 2012, at 6:04 PM, coco at dmi.unict.it wrote:

> Dear all,
> 
> I am using multigrid as a preconditioner and I noticed that if I try to perform 2 or more pre (or post) smoothing I obtain the same results as I perform just 1 smoothing. The ksp_view output is the following:
> 
> [...]
> 
> Down solver (pre-smoother) on level 1 -------------------------------
>    KSP Object:    (mg_levels_1_)     1 MPI processes
>      type: richardson
>        Richardson: damping factor=1
>      maximum iterations=2, initial guess is zero
>      tolerances:  relative=0, absolute=0, divergence=inf
>      left preconditioning
>      using PRECONDITIONED norm type for convergence test
>    PC Object:    (mg_levels_1_)     1 MPI processes
>      type: sor
>        SOR: type = local_symmetric, iterations = 1, local iterations = 1, omega = 1
>      linear system matrix = precond matrix:
>      Matrix Object:       1 MPI processes
>        type: shell
>        rows=18522, cols=18522
> 
> [...]
> 
> Even if maximum iterations=2 and the tolerances are zeros, it performs just one iteration. Anyway, the converged reason for the down smoother is 4, then it seems it converges with the maximum iterations, but the results I obtain are the same as in the case I set 1 pre smoothing.
> I implemented the relaxation in a routine UserSOR and I registered it as MATOP_SOR for the matrix of the kspSmootherDown context (together with a MATOP_MULT).
> 
> Finally, the output of the KSPMonitorTrueResidualNorm() for the first iteration is:
> 1 KSP preconditioned resid norm 0.000000000000e+00 true resid norm 5.094557710929e+02 ||r(i)||/||b|| 2.001351267432e+00

   With that 0 residual norm it has to stop. Something is wrong that the preconditioned residual is 0 while unpreconditioned is nonzero. That cannot be true to since the smoother is suppose to be a linear operator. Investigate why the it thinks the preconditioned residual is 0 but the nonpreconditioned is not.

   Barry

> 
> Thank you.
> Armando
>