[petsc-users] Preconditioning Matrix-free SNES, preconditioner update, PCView

[Matthew Knepley]

On Tue, Mar 31, 2015 at 9:03 PM, Juris Vencels <juris.vencels at gmail.com>
wrote:

>  Hello PETSc Users,
>
> I am using Matrix-Free SNES solver with the following options
>    PETSC_OPTIONS="ksp_type gmres -snes_monitor_short -snes_mf
> -user_precond"
>
> and shell-ILU preconditioner that I create as following:
>

This is not the canonical PETSc way to do this. We would suggest using the
matrix A
as the preconditioner matrix which you would give back in FormJacobian().
Then all
the reforming would be handled automatically.

  Matt


> ! ===============================
>
>       call PCSetType(pc,PCSHELL,ierr)
>       call PCShellSetUp
>       call PCShellSetApply(pc,PCShellApply,ierr)
>       call PCShellSetName(pc,"ShellPC",ierr)
>
> ! ===============================
>
>   subroutine PCShellSetUp()
>     implicit none
>
>      call SetInitialMatrix   ! Assambles matrix A
>
>      call PCCreate(PETSC_COMM_WORLD,mf_prec,ierr)
>      call PCSetType(mf_prec,PCILU,ierr)
>      call PCSetOperators(mf_prec,A,A,ierr)
>      call PCSetUp(mf_prec,ierr)
>      call PCShellSetName(mf_prec,"LinPC",ierr)
>
>   end subroutine PCShellSetUp
>
> ! ===============================
>
>   subroutine PCShellApply(pc_tmp,x_tmp,y_tmp,ierr)
>     implicit none
>     PC  :: pc_tmp
>     Vec :: x_tmp,y_tmp
>     PetscErrorCode :: ierr
>
>     call PCApply(mf_prec,x_tmp,y_tmp,ierr)
>
>   end subroutine PCShellApply
>
> ! ===============================
>
> * If matrix A changes during a simulation, how can I update
> preconditioner? Is calling "PCSetUp(mf_prec,ierr)" enough?
>
> * Function PCView prints out information about type and size of
> preconditioner. Can I view/get preconditioner matrix itself? I want to
> compare it with reference matrix.
>
> If I try to MatView factored matrix:
>     call PCFactorGetMatrix(mf_prec,pcmat,ierr)
>     call MatView(pcmat, PETSC_VIEWER_STDOUT_SELF,ierr)
>
> Then I get the following error:
>     [0]PETSC ERROR: No viewers for factored matrix except ASCII info or
> info_detailed
>
> Thanks!
>
>
>
>


-- 
What most experimenters take for granted before they begin their
experiments is infinitely more interesting than any results to which their
experiments lead.
-- Norbert Wiener
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150331/55c82d2b/attachment-0001.html>