ungridded dimension for pstokes #1591
Conversation
…n-20220902 update to MOM6 main 20220902 commit
…n-20221112 update to main 20221112 commit
use ungridded dimension for pstokes
jiandewang
requested review from
alperaltuntas,
breichl,
kshedstrom,
sanAkel and
abozec
There was a problem hiding this comment.
@jiandewang These files are outside the scope of what we use (since they pertain to nuopc). For that reason, I'll decline reviewing this PR.
@marshallward would you please "adjust" that from 5 reviews policy if not enough folks review it. Thank you!
@sanAkel fully understand, I think @kshedstrom will be on the same boat as you. Let me add @gustavo-marques |
|
@sanAkel your review is primarily an acknowledgement that the change will not impact your runs. Critical feedback is greatly appreciated, of course, but it's not necessary to understand everything coming in. Which is to say, just click "approved" if it doesn't affect you :). |
|
The CI failures appear to be from mom-ocean/main, not EMC. Most likely due to software stack changes on GitHub's end. Merging dev/gfdl into this PR does appear to work, so it's likely that the problem will go away after merging dev/gfdl into main. I would feel more comfortable fixing this before accepting this PR though, so we may want to cherry-pick a fix into either main dev/emc before accepting. I can't see what the actual problem could be, however, so I'm not yet sure how to proceed. Currently, the problem is a FMS library link failure. The FMS library can be built, and module include test passes ( |
@marshallward it might be simple to do cherry-pick into this PR branch but I need your instruction on how to do the cherry-pick |
|
The fix appears to be the change from MPICH to OpenMPI. I don't recall why this was needed, but I didn't think it was related to FMS linking. Anyway, I agree with your suggestion to cherry-pick into the EMC pull request, I'll send you some instructions if needed. |
@marshallward you mean NOAA-GFDL#270, right ? |
|
Yes, thats the one. (Looks like Alistair did not know why either.) My guess: The MPI calls in FMS have some type mismatches, which is why we need the I've sent some basic instructions, but get in touch if it doesn't work. |
Testing to see if GH actions is failing due to MPI installation
Codecov Report
@@ Coverage Diff @@
## main #1591 +/- ##
==========================================
+ Coverage 37.18% 37.22% +0.03%
==========================================
Files 263 263
Lines 73035 73074 +39
Branches 13609 13608 -1
==========================================
+ Hits 27161 27204 +43
- Misses 40859 40864 +5
+ Partials 5015 5006 -9
📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more |
There was a problem hiding this comment.
These changes fall outside the scope of what we use. Therefore these changes do not impact our answers, though they look okay and I am happy to see that if (cesm_coupled) has been removed.
These sort of blocks (or equivalent ifdef directives always come in the way of making our couplers more uniform and seamless!
thanks for Marshall's help, this PR branch has cherry-picked Alistair's fixing on CI issue |
|
I am not seeing anything here that I expect would present a problem from the GFDL side, especially as all these changes are confined to the NUOPC coupler, but given the potentially strong interactions with the wave interfaces, I would like to defer formal approval from GFDL until we hear what @breichl thinks about this PR. |
I don't foresee any conflicts with our use of the wave interfaces from the proposed code updates. I recommend approving from GFDL. |
|
This has passed our CI and review; GFDL approves this PR. |
|
since we got green lights from all reviewers, I am going to merge it soon. |
this is a minor PR to implements use of ungridded dimensions for the partitioned stokes drift by removing the exchange of 6 individual fields in favor of 2 fields containing 3 ungridded dimensions each.
The WW3 mesh cap has the capability to use ungridded dimensions to transfer fields. This allows the 6 fields (3 each for x,y) used to exchange the partitioned stokes drift between WW3 and MOM6 to be simplified to only 2 fields. The MOM6 and WW3 caps used in UFS can be simplified as a result.