On Sat, May 14, 2011 at 22:31, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote:
> Via Vec/MatSetValuesLocal? I would agree with that. > What I don't like about that is that now there will be two different > ways of treating splits that happen to be nonoverlapping: > via some clever XXXSetValuesLocal and via the nonoverlapping splits > mechanism inside PetscLayout. > I think MatSetValuesLocal() should be the canonical way to modify "part" of a Mat. For Vec, you can get a local vector to modify with array access or use VecSetValuesLocal() depending on your needs. Top-down addressing requires some amount of global knowledge and it makes information flow in an unnatural way (for code and for performance). -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20110514/2435feec/attachment.html>
