[petsc-dev] DMGetMatrix --> DMGetMatrices?

[Jed Brown]

On Fri, Feb 10, 2012 at 19:51, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:

> How does this explanation square with the following code from KSPSetUp?
>
>  if (ksp->dmActive && !ksp->setupstage) {
>     /* first time in so build matrix and vector data structures using DM */
>     if (!ksp->vec_rhs) {ierr =
> DMCreateGlobalVector(ksp->dm,&ksp->vec_rhs);CHKERRQ(ierr);}
>     if (!ksp->vec_sol) {ierr =
> DMCreateGlobalVector(ksp->dm,&ksp->vec_sol);CHKERRQ(ierr);}
>     ierr = DMCreateMatrix(ksp->dm,MATAIJ,&A);CHKERRQ(ierr);
>     ierr = KSPSetOperators(ksp,A,A,stflg);CHKERRQ(ierr);
>     ierr = PetscObjectDereference((PetscObject)A);CHKERRQ(ierr);
>  }
>

Look down the page a little, DMComputeJacobian() is what actually assembles.

    if (ksp->setupstage != KSP_SETUP_NEWRHS) {
      ierr = KSPGetOperators(ksp,&A,&A,PETSC_NULL);CHKERRQ(ierr);
      ierr = DMComputeJacobian(ksp->dm,PETSC_NULL,A,A,&stflg);CHKERRQ(ierr);
      ierr = KSPSetOperators(ksp,A,A,stflg);CHKERRQ(ierr);
    }
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20120210/01ad119b/attachment.html>