adjust some cpld v16 tests for forecast length and output type

[DeniseWorthen]

Description

• The v16 cpld benchmark tests are seen to be coming close to the wall clock time of 30mins when fhmax=12.

• The v16 cpld benchmark test on gaea fails with a pio error (see Issue cpld_bmarkfrac_v16 test fails on Gaea  #509) when the ATM output is the tiled grid.

Solution

Reduce the fhmax for the cpld v16 test types to 6 hours.
Change the output type for the cpld v16 tests to gaussian_grid

[DeniseWorthen]

@junwang-noaa I have a branch for the updated tests that I'm testing on gaea. I created a baseline and then tried to verify against it. All tests pass except for the cpld_restart_bmarkfrac_v16 test. This test is not reproducing the phyf006.nc forecast file but all other files do reproduce. The data is actually different in the phyf006.nc forecast file.

baseline dir = /lustre/f2/scratch/Denise.Worthen/FV3_RT/test_v16_baseline2/cpld_bmarkfrac_v16
working dir  = /lustre/f2/scratch/Denise.Worthen/FV3_RT/rt_9502/cpld_restart_bmarkfrac_v16
Checking test 003 cpld_restart_bmarkfrac_v16 results ....
 Comparing phyf006.nc ............ALT CHECK......NOT OK
 Comparing dynf006.nc .........OK
 Comparing RESTART/coupler.res .........OK
 Comparing RESTART/fv_core.res.tile1.nc .........OK
 Comparing RESTART/fv_core.res.tile2.nc .........OK
 Comparing RESTART/fv_core.res.tile3.nc .........OK
 Comparing RESTART/fv_core.res.tile4.nc .........OK
 Comparing RESTART/fv_core.res.tile5.nc .........OK
 Comparing RESTART/fv_core.res.tile6.nc .........OK
...

[junwang-noaa]

Is this the cpld_restart_bmarkfrac_v16 test with fhmax changed from 12 to 6 using the top of develop branch?

[DeniseWorthen]

It is a branch of the upcoming PR #533 which includes the fix for building Waves on Gaea. It has the fhmax set to 6 and the gaussian output change.

[DeniseWorthen]

I repeated the restart test for the cpld v16 test using tiles and 3h/3h/6h. That restart test also fails comparison of the phy006.tileX.nc files but all restart files between the restart and control test are b4b.

[junwang-noaa]

Denise, I did cpld v16 test with Gaussian grid 6h/6h/12h, the results are the same at fh012. /lustre/f2/scratch/emc.nemspara/FV3_RT/rt_6970/cpld_bmarkfrac_v16 /lustre/f2/scratch/emc.nemspara/FV3_RT/rt_6970/cpld_restart_bmarkfrac_v16 cmp phyf012.nc ../cpld_bmarkfrac_v16/phyf012.nc source code is at: /lustre/f2/pdata/ncep/emc.nemspara/rttest/denise_pr533/tests I got permission denied when looking your data:

ls

/lustre/f2/scratch/Denise.Worthen/FV3_RT/rt_9502/cpld_restart_bmarkfrac_v16 ls: cannot access '/lustre/f2/scratch/Denise.Worthen/FV3_RT/rt_9502/cpld_restart_bmarkfrac_v16': Permission denied May I ask where is your source code too?

…

On Mon, May 3, 2021 at 8:52 PM Denise Worthen ***@***.***> wrote: I repeated the restart test for the cpld v16 test using tiles and 3h/3h/6h. That restart test also fails comparison of the phy006.tileX.nc files but all restart files between the restart and control test are b4b. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#551 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI7D6TO3THNYNZDWO47A3LLTL5AODANCNFSM434WCPIQ> .

[DeniseWorthen]

My code is located here: /lustre/f2/pdata/ncep/Denise.Worthen/ufs-weather-model/tests

I've updated since the original post; it is now up-to-date with develop. The current cpld_bmarkfrac_v16 and cpld_restart_bmarkfrac_v16 tests are using the tiled output because that is what I tested yesterday on Hera.

I don't understand the permission issue. It shows my group as rstprod. Is that why you can't see into them?

[DeniseWorthen]

I've been able to get the 6/3/9 restart test with gaussian output to pass at c96 by also setting

export NFHOUT=3
export FDIAG=3

I did also try these with 3/3/6 but the phy006.nc file still failed to reproduce. Currently the cpld v16 tests are run for 12h. I will need to test this in the actual v16 tests if we want to add this change to PR #544.

[junwang-noaa]

You will also need to change namelist variable fhzero in gfs_physics_nml in input.nml to 3 if you would like to have reproducibility 3/3/6

…

On Wed, May 5, 2021 at 5:10 PM Denise Worthen ***@***.***> wrote: I've been able to get the 6/3/9 restart test with gaussian output to pass at c96 by also setting export NFHOUT=3 export FDIAG=3 I did also try these with 3/3/6 but the phy006.nc file still failed to reproduce. Currently the cpld v16 tests are run for 12h. I will need to test this in the actual v16 tests if we want to add this change to PR #544 <#544>. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#551 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AI7D6TIFE37L36MPM7PAKPDTMGX2JANCNFSM434WCPIQ> .