[petsc-dev] Creation of snes->vec_func

[Jed Brown]

On Wed, Dec 21, 2011 at 19:16, Barry Smith <bsmith at mcs.anl.gov> wrote:

> You could do this. Sounds reasonable.


Done:

http://petsc.cs.iit.edu/petsc/petsc-dev/rev/918daef122af


Now what is going on with this (apparent) debugging code virs.c:424 and
virsaug.c:1183 (http://petsc.cs.iit.edu/petsc/petsc-dev/rev/31c8ce):

    ierr = MatFindNonzeroRows(jac_inact_inact,&keptrows);CHKERRQ(ierr);
    if (0 && keptrows) {
      PetscInt       cnt,*nrows,k;
      const PetscInt *krows,*inact;
      PetscInt       rstart=jac_inact_inact->rmap->rstart;

      ierr = MatDestroy(&jac_inact_inact);CHKERRQ(ierr);
      ierr = ISDestroy(&IS_act);CHKERRQ(ierr);

      ierr = ISGetLocalSize(keptrows,&cnt);CHKERRQ(ierr);
      ierr = ISGetIndices(keptrows,&krows);CHKERRQ(ierr);
      ierr = ISGetIndices(IS_inact,&inact);CHKERRQ(ierr);
      ierr = PetscMalloc(cnt*sizeof(PetscInt),&nrows);CHKERRQ(ierr);
      for (k=0; k<cnt; k++) {
        nrows[k] = inact[krows[k]-rstart];
      }
      ierr = ISRestoreIndices(keptrows,&krows);CHKERRQ(ierr);
      ierr = ISRestoreIndices(IS_inact,&inact);CHKERRQ(ierr);
      ierr = ISDestroy(&keptrows);CHKERRQ(ierr);
      ierr = ISDestroy(&IS_inact);CHKERRQ(ierr);

      ierr =
ISCreateGeneral(PETSC_COMM_WORLD,cnt,nrows,PETSC_OWN_POINTER,&IS_inact);CHKERRQ(ierr);
      ierr =
ISComplement(IS_inact,F->map->rstart,F->map->rend,&IS_act);CHKERRQ(ierr);
      ierr =
MatGetSubMatrix(snes->jacobian,IS_inact,IS_inact,MAT_INITIAL_MATRIX,&jac_inact_inact);CHKERRQ(ierr);
    }
    ierr = DMSetVI(snes->dm,IS_inact);CHKERRQ(ierr);
    /* remove later */


This (a) leaks keptrows and (b) doesn't do anything useful because of the
if (0 && ...). What's the idea here?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20111221/e2e35657/attachment.html>