[petsc-dev] Add natural-to-global and global-to-natural operations for DM?

[Jed Brown]

On Wed, Mar 20, 2013 at 12:49 AM, Lisandro Dalcin <dalcinl at gmail.com> wrote:

> IMHO, global<->natural mapping does make sense on sparse problems.
> What about making global<->natural to be the identity map (i.e, just
> use VecCopy or VecAXPY depending on InsertMode) for any DM where the
> global<->natural does not make sense?
>

What can you do generically with a Natural vector?

With Local vectors, we can implement DMCompositeScatter(), which is really
the only functional reason to have promoted DMGlobalToLocalBegin/End. All
other uses were already specialized on DM type.

What is the equivalent functionality that motivates promoting the "Natural"
concept?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/petsc-dev/attachments/20130320/6445a1c3/attachment.html>