[mpas-developers] Zoltan support in MPAS

[Michael Duda]

Hi, All.

I've created a branch, zoltan_cleaning, from which I've removed all of
the code related to Zoltan. All of the cores compile, and I get
bit-identical results for a short test of the non-hydrostatic core;
neither of these outcomes are surprising, of course.

In case anyone else would like to review the changes in the branch, I'll
wait a couple of days before merging this back to the trunk.

Thanks,
Michael


On Mon, Apr 15, 2013 at 05:55:22PM +0000, Ringler, Todd D wrote:
> 
> this was an early project with Sandia to optimize the reordering of cell indices for data locality.
> 
> this is definitely something that we will have to revisit, but we would want to rethink how we connected to zoltan anyway.
> 
> so i think removing the connection is fine.
> 
> On Apr 15, 2013, at 11:42 AM, Doug Jacobsen <jacobsen.douglas at gmail.com<mailto:jacobsen.douglas at gmail.com>> wrote:
> 
> Hey Michael,
> 
> I think it's fine to remove this. I don't know of anyone here who uses Zoltan.
> 
> Doug
> 
> 
> On Mon, Apr 15, 2013 at 11:35 AM, Michael Duda <duda at ucar.edu<mailto:duda at ucar.edu>> wrote:
> I've not heard any objections to removing the Zoltan code, but just in
> case folks were particularly busy last week, I'll wait until later this
> week before starting to remove this code from the trunk.
> 
> Regards,
> Michael
> 
> 
> On Mon, Apr 08, 2013 at 05:11:50PM -0600, Michael Duda wrote:
> > Hi, All.
> >
> > The MPAS framework currently has a number of hooks to call the Zoltan
> > package to reorder cells, edges, and vertices in an attempt to improve
> > cache reuse; however, the code does not currently build with Zoltan,
> > and I suspect it has been a couple of years since anyone has tried to
> > do so. In the interest of clearing out unused or unusable code ahead of
> > the initial MPAS release, I'd like to propose the idea of eliminating
> > all of the Zoltan hooks in the MPAS code. Should there be any interest
> > in future in trying out Zoltan, I think it would probably be a good
> > idea to re-implement the functionality from scratch anyway to better
> > fit with our current initialization sequence. What do you all think
> > about getting rid of the Zoltan code in MPAS?
> >
> > Regards,
> > Michael
> _______________________________________________
> mpas-developers mailing list
> mpas-developers at mailman.ucar.edu<mailto:mpas-developers at mailman.ucar.edu>
> http://mailman.ucar.edu/mailman/listinfo/mpas-developers
> 
> _______________________________________________
> mpas-developers mailing list
> mpas-developers at mailman.ucar.edu<mailto:mpas-developers at mailman.ucar.edu>
> http://mailman.ucar.edu/mailman/listinfo/mpas-developers
>