[ncl-talk] ESMF regridding is really slow

[Mary Haley]

Hi all,

Laura and I have been iterating offline about this and I wanted to update
ncl-talk.

The issue is that when you regrid unstructured data, the internal
regridding routine needs to have a mesh that defines the edges of a
triangular or some mesh structure of each cell.  If such a mesh is not
provided by the user, then ESMF_regrid will call the Fortran-based "csstri"
routine to generate a triangular mesh.

This routine can take hours if not days on large grids (this one was 6.5
million cells).

It turns out that Laura's NetCDF file already had a triangular mesh on it,
so she was able to read this mesh off the file and set it to the special
"SrcTriangularMesh" option.

With this option set, we were able to get the generation of the weights
file down to 6 minutes.

Once you have the weights file, you can use this to regrid future
variables, and this only took about 1 second on my Mac.

--Mary


On Wed, Apr 8, 2015 at 11:24 AM, Mary Haley <haley at ucar.edu> wrote:

> Hi Laura,
>
> I don't think I've tried regridding an MPAS grid of this size.
>
> Can you point me to your script and file on geyser, and I'll have a look?
> You can send me this information offline if you like.
>
> --Mary
>
>
> On Wed, Apr 8, 2015 at 10:42 AM, Laura Fowler <laura at ucar.edu> wrote:
>
>> Hello:
>>
>> I have been trying to regrid a MPAS unstructured variable-resolution
>> grid that has a number of cells equal 6848514 to a rectangular 0.25
>> deg latitude-longitude grid. I use the function ESMF_regrid to create
>> generic source, destination, and weights files that I can use for my
>> MPAS runs.
>>
>> The issue that I have is that it is really really slow and I cannot
>> fit the regretting within the 12 hour limit on geyser. Is there
>> anything else that I can do to speed up that process?
>>
>> Thanks,
>> Laura
>>
>>
>> --
>>
>> !-------------------------------------------------------------------------------------------------------------
>> Laura D. Fowler
>> Mesoscale and Microscale Meteorology Division (MMM)
>> National Center for Atmospheric Research
>> P.O. Box 3000, Boulder CO 80307-3000
>>
>> e-mail: laura at ucar.edu
>> phone: 303-497-1628
>>
>>
>> !-------------------------------------------------------------------------------------------------------------
>> _______________________________________________
>> ncl-talk mailing list
>> List instructions, subscriber options, unsubscribe:
>> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.ucar.edu/pipermail/ncl-talk/attachments/20150428/eda1225a/attachment.html