yeah, I meant that when "stencilwise" access is required, then VecGetDM(v,&dm); DMVecGetStencilIterator(dm,v,&itor); <something to dereference the itor> etc In a way, this is similar to DAVecGetArray(), which generalizes VecGetArray to use the geometric information about the function space (DA) the Vec belongs to.
On Tue, Dec 7, 2010 at 3:42 PM, Jed Brown <jed at 59a2.org> wrote: > On Tue, Dec 7, 2010 at 22:40, Dmitry Karpeev <karpeev at mcs.anl.gov> wrote: > >> And yet DMDA already defines this local fine-granularity "stencil" access. >> Where would it go, if not in the DM? In my opinion this is part of the >> function space definition: >> it generalizes VecGetArray to something more geometric. >> > > I just think it should be a DM-level call. Not VecGetUberCustomIterator. > Perhaps I'm wrong, but I think it belongs in the DM. > > Not all DMs have to implement this. >> > > Maybe that's what you were saying too. > > Jed > -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20101207/49cefba4/attachment.html>
