In filter_topo added extrapolation to halo points #571
Conversation
|
Thanks @RatkoVasic-NOAA. Does routine 'read_topo_file` have the same problem. What is in the halo region of arrays 'oro' and 'mask'? |
The only arrays (in regional standalone mode) that were using halo points are geolat_c and geolon_c. All other arrays are not using halo points. |
|
@RatkoVasic-NOAA I tested your branch at b602411 on my canned test case. I compiled your branch on Hera with CMAKE_BUILD_TYPE equal to "Release" and "Debug". I got big differences along the halo. I will provide details. |
|
Here are the differences between "Debug" and "Release" mode for my test case: I visualized the files. The differences are confined to the halo. If you do not already know, you can change the compile type using this variable: |
|
I'll test it. |
|
@RatkoVasic-NOAA I tried 1ec6689 on my canned case. The intel/release and intel/debug options gave identical results. The gnu/release option gives a different answer from the intel options. And the gnu/release and gnu/debug options give different answers. Differences are confined to the left edge of the grid. Here are the gnu differences: |
|
I guess I'll have to work on this when I come back 😐. (Friday)Sent from my Galaxy
-------- Original message --------From: GeorgeGayno-NOAA ***@***.***> Date: 8/26/21 10:14 AM (GMT-05:00) To: NOAA-EMC/UFS_UTILS ***@***.***> Cc: RatkoVasic-NOAA ***@***.***>, Mention ***@***.***> Subject: Re: [NOAA-EMC/UFS_UTILS] In filter_topo added extrapolation to halo points (#571)
@RatkoVasic-NOAA I tried 1ec6689 on my canned case. The intel/release and intel/debug options gave identical results. The gnu/release option gives a different answer from the intel options. And the gnu/release and gnu/debug options give different answers. Differences are confined to the left edge of the grid. Here are the gnu differences:
$ nccmp -dmfqS oro.C424.tile7.nc.ratko.O3.gnu oro.C424.tile7.nc.ratko.debug.gnu
Variable Group Count Sum AbsSum Min Max Range Mean StdDev
orog_filt / 21 77.3102 97.5196 -7.94759 24.5957 32.5433 3.68144 7.90504
—You are receiving this because you were mentioned.Reply to this email directly, view it on GitHub, or unsubscribe.Triage notifications on the go with GitHub Mobile for iOS or Android.
|
|
Yes, this is a frustrating bug. Also, I am writing a simple unit test for your new routine. I will check it in to your branch today. |
to allow for using it in unit tests. Create simple unit test for that routine. Fixes ufs-community#105
|
@RatkoVasic-NOAA I added a unit test stub for you at f87f1f5. |
|
@RatkoVasic-NOAA I see that your unit test was failing with a memory leak error. In your unit test, try deallocating "testdata" at the end of program execution. |
|
It works! |
|
@JeffBeck-NOAA may want to run a quick test with the regional workflow. @RatkoVasic-NOAA is this ready to merge? |
I replied by email, I guess it didn't go through. |
The 'grid_gen' regression tests will need to be run. Since results will likely change for the regional tests, new baseline files will need to be generated. @RatkoVasic-NOAA can you run the tests on Hera or WCOSS? |
|
And merge to your branch the latest changes from 'develop'. |
Thanks. I have updated the baseline data on all machines. |
This commit is related to the issue #105
Fields' geolat_c and geolon_c halo points are used to calculate dxc and dyc, but halo points are not initialized for regional standalone case. I added extrapolation for regional standalone similar way to regional nest and global application.