[petsc-users] KSP changes for successive solver

[Mark Adams]

On Thu, Jul 16, 2015 at 5:53 PM, Michele Rosso <mrosso at uci.edu> wrote:

>  Hi,
>
> I am performing a series of solves inside a loop. The matrix for each
> solve changes but not enough to justify a rebuilt of the PC at each solve.
> Therefore I am using  KSPSetReusePreconditioner to avoid rebuilding unless
> necessary. The solver is CG + MG with a custom  PC at the coarse level.
>

Ah, OK this is the problem.  The code I just sent gets fooled but your
custom PC, probably.

We should just remove this paternal code:

  PetscBool      preonly,lu,redundant,cholesky,svd

   ...


  /*
      If coarse solver is not direct method then DO NOT USE preonly
  */
  ierr =
PetscObjectTypeCompare((PetscObject)mglevels[0]->smoothd,KSPPREONLY,&preonly);CHKERRQ(ierr);
  if (preonly) {
    ierr =
PetscObjectTypeCompare((PetscObject)cpc,PCBJACOBI,&bjaclu);CHKERRQ(ierr);
    if (bjaclu) {
      KSP *k2;
      PetscInt ii,first;
      ierr = PCBJacobiGetSubKSP(cpc,&ii,&first,&k2);CHKERRQ(ierr);
      if (ii==1) {
        PC pc2;
        ierr = KSPGetPC(k2[0],&pc2);CHKERRQ(ierr);
        ierr =
PetscObjectTypeCompare((PetscObject)pc2,PCLU,&bjaclu);CHKERRQ(ierr);
      }
      else bjaclu = PETSC_FALSE;
    }
    ierr = PetscObjectTypeCompare((PetscObject)cpc,PCLU,&lu);CHKERRQ(ierr);
    ierr =
PetscObjectTypeCompare((PetscObject)cpc,PCREDUNDANT,&redundant);CHKERRQ(ierr);
    ierr =
PetscObjectTypeCompare((PetscObject)cpc,PCCHOLESKY,&cholesky);CHKERRQ(ierr);
    ierr =
PetscObjectTypeCompare((PetscObject)cpc,PCSVD,&svd);CHKERRQ(ierr);
    if (!lu && !redundant && !cholesky && !svd && !bjaclu) {
      ierr = KSPSetType(mglevels[0]->smoothd,KSPGMRES);CHKERRQ(ierr);
    }
  }




> If KSP is not updated each time, everything works as it is supposed to.
> When instead I allow the default PETSc  behavior, i.e. updating PC every
> time the matrix changes, the coarse level KSP , initially set to PREONLY,
> is changed into GMRES
> after the first solve. I am not sure where the problem lies (my PC or
> PETSc), so I would like to have your opinion on this.
> I attached the ksp_view for the 2 successive solve and the options stack.
>
> Thanks for your help,
> Michel
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20150719/632fcc6d/attachment.html>