-
Notifications
You must be signed in to change notification settings - Fork 239
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
HR4 GWD update for FV3 #2290
base: develop
Are you sure you want to change the base?
HR4 GWD update for FV3 #2290
Conversation
All (@Qingfu-Liu @barlage @JongilHan66 @mdtoyNOAA @cenlinhe @BoYang-NOAA), please look over this list copied from cpld_control_p8_mixedmode intel |
Looks good |
@Qingfu-Liu Here is the PR into your branch to add RT logs: Qingfu-Liu#1 (also updates FV3 submodule pointer) |
update RT logs and FV3 submodule pointer
@Qingfu-Liu Can we move this out of draft and fill out the description? |
@grantfirl Thank you very much. I just updated the description and change the draft to PR |
@Qingfu-Liu can you sync up branch? We can start working on this pr. |
@jkbk2004 Thanks very much for syncing this. I didn't have permissions to do this on Qingfu's branch. I'm still out of the country until mid next week, although I'll be available off and on throughout the day to merge if necessary if no one else is. Dustin is at the same conference as me. |
@Qingfu-Liu @grantfirl @dustinswales @mdtoyNOAA can you check the following error message about mntvar !subgrid orographic statistics data ?
It crashes with rrfs_v1beta_debug intel control_p8_atmlnd_sbs intel control_p8_atmlnd intel control_p8_atmlnd_debug intel rrfs_v1beta_debug gnu. An experiment output is available at /scratch2/NAGAPE/epic/Jong.Kim/stmp2/Jong.Kim/FV3_RT/rt_3529436/rrfs_v1beta_debug_intel/err |
@jkbk2004 There is a new set of orographic data related to this PR#2290 and the new data is described in PR#2670 workflow. Can you run the regression tests using the new data? Thank you very much.: |
If there are any anticipated changes to data in the input-data-YYYYMMDD directories, it should be very specifically addressed in the template. Please make changes there. |
OK. Thanks. After look the error, I am not sure the failed tests are related to the data change. I just add Jongil who works on the code to see if he know this. |
@JongilHan66 Can you take a look of the errors @jkbk2004 mentioned in this PR#2290? Thanks |
Currently we use gwd_opt=2 (unified ugwp GWD), which is defined in config.fcst. Then the dimension of MNTVAR is increased from 10 to 24, as described in 'GFS_typedefs.F90'. |
@Qingfu-Liu If you haven't yet, please update the config.fcst files with those in the directories of gfs & gefs in /lfs/h2/emc/physics/noscrub/jongil.han/git_HR4_gwd/global-workflow/parm/config |
@JongilHan66 Those files are updated in workflow PR#2670: |
@Qingfu-Liu Did you also update the "parsing_namelists_FV3.sh" and "parsing_namelists_FV3_nest.sh" in the ush directory? |
@JongilHan66 Yes, both files are updated in the PR#2670. There are 4 files are updated in PR#2670 |
@Qingfu-Liu If gwd_opt=2 in config.fcst, the dimension of MNTVAR is 24 and the model should not complain the max of 14 which is for gwd_opt=1. |
@JongilHan66 Thanks. I think I understand the problem. |
Nick @NickSzapiro-NOAA is working on GEFSv13 RTs, I am providing a set of new input data. |
@jkbk2004 I don't understand why you need a new cpld_control_run_gfsv17? You already have the change in the cpld_control_run and all the cpld tests will use that (and the new input-data). The question is more about the tests which don't use cpld_control_run and which will not use the new input-data. Given @yangfanglin 's comment that those tests will still run w/ the older data, it may be a later PR to update those tests to the new settings when they move away from the p8 settings. |
Got it! I will finalize default_var.sh with the comment above and update tests/tests cpld cases accordingly. And we will expect another PR to update all other cases. |
So far one of the changes that is made to global coupled regression tests is the fix file update, @yangfanglin, should the fix files be updated for the global tests with all the resolutions? Also what version is the fixed file? |
@junwang-noaa The fix file version is 20240624 (see: /scratch1/NCEPDEV/global/glopara/fix/ugwd/20240624 or the original directory /scratch1/NCEPDEV/global/Qingfu.Liu/git/GWD_SHong/APR2023_SSO and global-workflow issue: NOAA-EMC/global-workflow#2716) |
So this requires the sfc input files to be regenerated from the new oro data? |
@junwang-noaa it is not required to regenerate sfc data |
Should the fix files be updated for the global tests with all the resolutions? @JessicaMeixner-NOAA |
@junwang-noaa See Denise's comments above in this thread. |
@JessicaMeixner-NOAA @yangfanglin @JongilHan66 can you take a look at /scratch1/NCEPDEV/stmp2/Jong.Kim/FV3_RT/rt_102912/cpld_control_gfsv17_intel/input.nml ? I used the requested parameters. It crashes. err file is in the experiment directory. |
@jkbk2004 At least, the settings for HR4 GWD update are correct. |
The run from the workflow has been submitted, but the queue is slow. I'll report back on that when I know more. |
@jkbk2004 suggestions for tests: 1) test with C768 resolution; 2) C96 test but with the old oro_data |
We don't have regression test case for C768. cpld_bmark_p8 case is C384. It crashes same way. cpld_control_gfsv17_intel is C96. It still blows up even if I use old oro files. BTW, crash happens around https://github.com/Qingfu-Liu/ccpp-physics/blob/00dc921754a699d361ee0f98b4f321ca021563a1/physics/GWD/drag_suite.F90#L2616. It complains as
@JongilHan66 @grantfirl @dustinswales @mdtoyNOAA sounds like memory mapping issue. array mismatch for some of those diagnostic varaiables? |
@grantfirl @dustinswales @JongilHan66 If I run with gnu and debug on: cpld_control_gfsv17_intel, it shows
|
@jkbk2004 to narrow down the cause, could you make 2 more tests with 1) do_gsl_drag_tofd=.false. and 2) do_gwd_opt_psl=.false.? |
This indicates the new dataset itself is fine. The model code https://github.com/Qingfu-Liu/fv3atm/tree/HR4-GWD-update might have issues (@dustinswales )
3). Finally I dropped the executable from step 2 to the step 1 RT directory, updated input.nml to set do_gsl_drag_ss=.false. and added do_gwd_opt_psl= .true. and psl_gwd_dx_factor= 6.0, rerun the case. The model failed at calling drag_suite.F90, like what John found out. @JongilHan66 Could you compare https://github.com/Qingfu-Liu/ccpp-physics/commits/00dc921754a699d361ee0f98b4f321ca021563a1/physics/GWD/drag_suite.F90 to what you had used ? |
My run from the workflow finally went in but after the last time I checked last night. Unfortunately I had an error about a variable not being found, not an actual forecast failure (yet). I'll keep you posted on my runs with that. |
We are going to start working on #2183 today while debugging continues in this pr. @BrianCurtis-NOAA @FernandoAndrade-NOAA @zach1221 FYI |
My run from global-workflow also segfaulted, but hopefully we can use this to confirm settings between workflow and rt.sh: Log file: https://github.com/JessicaMeixner-NOAA/global-workflow/commits/HR4-GWD-update/ |
@jkbk2004 @dustinswales @JongilHan66 I'm back from the UK, and I can start looking at this failure. |
@jkbk2004 @dustinswales @JongilHan66 FYI, I'm running a quick test for |
@grantfirl One thing to check. Would it be ok to let #2183 go first? This PR is mostly with GWD update. I think it's ok. |
If it's already being worked on, sure. Once #2183 is merged, I'll update submodule pointers for this one so that it should be good to go next. |
Given the comment history and issues with this PR, I think it is problematic to update only a subset of tests to the new input data. If we maintain different tests with different oro_data_ss* and oro_data_ls* versions, then we won't actually know if the new data works across all applications. From Fanglin's comment, it is expected that the new input data should work even for tests which retain the older/different namelist settings. I would suggest a new input-data-directory be created with the new ss and ls files to be used by all configurations. @yangfanglin @JessicaMeixner-NOAA @jkbk2004 |
@jkbk2004 @dustinswales @JongilHan66 FYI, I've found a couple of bugs in the new code in drag_suite.F90 that were causing seg faults in certain configurations. I think that I have them fixed, but I'm running final tests of the |
@jkbk2004 @dustinswales @JongilHan66 The |
Commit Queue Requirements:
Description:
Commit Message:
Priority:
Git Tracking
UFSWM:
Sub component Pull Requests:
UFSWM Blocking Dependencies:
Changes
Regression Test Changes (Please commit test_changes.list):
Input data Changes:
HR4 GWD update NOAA-EMC/global-workflow#2670
This PR#2670 includes changes for four scripts, and new orographic data.
The new orographic data temporarily stored at:
/scratch1/NCEPDEV/global/Qingfu.Liu/git/GWD_SHong/APR2023_SSO
and this new data should replace the old data: /scratch1/NCEPDEV/global/glopara/fix/ugwd/20231027
Library Changes/Upgrades:
Testing Log: