[mpas-developers] error using POP grid

[Michael Duda]

Hi, Mat.

The code in the repository trunk isn't fully capable of handling
missing cells in a grid yet. Although parts of the code (namely,
the solvers) are set up to avoid cells/edges/vertices with index
zero, which happens when running in parallel for
cells/edges/vertices that are outside the last row of ghost cells,
the code can't handle missing cells/edges/vertices in the input
grid; as you've seen, the halo-building code (I think) fails here.

Todd and I will be working on changes to accommodate missing cells
in the input grid as part of the proposed boundary condition
implementation. Right now, we have a branch at
branches/lateral_boundary_conditions that I'm planning to make
some changes to very soon so that we can handle missing cells; in
fact, the changes you sent to Todd for module_block_decomp.F
should be included in these changes. Once I've gotten those
changes in place, it would be great if you could take a look at
the branch and see whether it is robust enough to handle the POP
grid with missing land cells.

Cheers,
Michael


On Thu, Mar 25, 2010 at 03:14:17PM -0600, Mathew Maltrud wrote:
> Hi All--
> 
> Todd asked me to broadcast the error i was getting when running with a  
> POP grid with land cells missing and zero values for missing  
> cellOnEdge that result.  here's what happens:
> 
> mm at cy-c1.lanl.gov {154}% ./sw_model.exe
>   Error in dmpar_alltoall_field: For non-dmpar, arrayIn and arrayOut  
> dims must match.
> 
> the arrays differ in size because the hash table logic interprets 0 as  
> a valid cell value.
> 
> cheers,
> -mat
> 
> _______________________________________________
> mpas-developers mailing list
> mpas-developers at mailman.ucar.edu
> http://mailman.ucar.edu/mailman/listinfo/mpas-developers