update XML defaults for theta-l dycore (V2) #3368
Conversation
also remove obsolete namelist variables
|
this touches a lot of code, so I will run the full acme_integration on sandiatoss3 to verify BFB. @oksanaguba , can you configure an ne30 and ne30pg2 F compset and see if the dycore settings are what you would use for the upcoming v2 watercycle configuration? |
| <nu_top dyn_target="theta-l" hgrid="ne240np4"> 4.3e4 </nu_top> | ||
| <nu_top dyn_target="theta-l" hgrid="ne256np4"> 4e4 </nu_top> | ||
| <nu_top dyn_target="theta-l" hgrid="ne512np4"> 2e4 </nu_top> | ||
| <nu_top dyn_target="theta-l" hgrid="ne1024np4"> 1e4 </nu_top> |
There was a problem hiding this comment.
Is if fair to say these nu_top coefs are still wip?
There was a problem hiding this comment.
yes, these are the ones that work with hsv_tom=1. since we dont know the correct value for high-res, i thought we should start with the most efficient.
| <hypervis_subcycle hgrid="ne0np4_svalbard_x8v1_lowcon" dyn_target="preqx" > 8 </hypervis_subcycle> | ||
| <hypervis_subcycle hgrid="ne0np4_sooberingoa_x4x8v1_lowcon" dyn_target="preqx" > 8 </hypervis_subcycle> | ||
| <hypervis_subcycle hgrid="ne0np4_enax4v1" dyn_target="preqx">7</hypervis_subcycle> | ||
| <hypervis_subcycle hgrid="ne0np4_twpx4v1" dyn_target="preqx"> 7 </hypervis_subcycle> |
There was a problem hiding this comment.
There are a lot of changes for resolutions that are not tested nightly, like ne120 and RRM with preqx dycore. How would we know settings for them are not modified?
There was a problem hiding this comment.
for all grid specific settings, I added a dyn_target="preqx" clause, so they will all be preserved for the preqx dycore. The the grid independent settings, the integration suite should be a good test of if any preqx settings are changed.
There was a problem hiding this comment.
Yes, some changes can be read like that, that preqx dycore was added, but some other changes are harder to parse.
|
|
||
| <entry id="nu_s" type="real" category="se" | ||
| group="ctl_nl" valid_values="" > | ||
| Hyperviscosity applied to dynamics scalars [m^4/s]. |
There was a problem hiding this comment.
... applied to temperature/pot. temperature? dp is a scalar, too.
There was a problem hiding this comment.
does this need to be changed?
|
Ran atm_integration against today's master on anvil. All is ok, except for 1 test, SMS_D_Ln5.ne4_ne4.FC5AV1C-L.anvil_intel.cam-cosplite, stack is below: Log for failed test: |
|
Default run for ne30np4 fails with This is because of _map=2 in default theta-l. I always run with _map=0 to avoid setting tolerances. What should I do here? |
|
This is what I did before: |
|
For ne30np4 run I did not have another run with all these defaults, so, comparing parameters by eye. All looks ok, now default rsplit is 2 (all our ne30 runs were probably done with 3 or 1), ftype is 2, not 4. Map is 2, all newer runs were done with 0. |
|
cubed_sphere_map issues should be fixed now there will be an issue for new np4 grids made with TR mapping files, issue #3369 |
|
testing with e3sm_integration found two more issues: need to set se_ftype=0 default with SCAM=1 (single column model) Other erros were cause by some confusion since namelist defaults are based on cfg_ref->dyn_target = "preqx" or "theta-l", while the build options are based on CAM_TARGET (stored in opts{'dyn_target'}) can have many options (preqx, preqx_acc, preqx_kokkos, theta-l) |
fix SCAM=1 default value of se_ftype make all preqx_* targets use the 'preqx' defaults dont access uniitalized variables on the non-root MPI task
|
So error |
|
yes, should be fixed by afc60ae |
|
with afc60ae, all the e3sm_integration tests pass on sandiatoss3 except for the theta dycore ones that have not yet been reblessed, and the ICG test which has been failing for a while. |
|
Ran e3sm_integration suite against master of Dec 20: A lot of NL fails, also, For hommebfb it is a build fail, for PET is was failing with the same sigterm error 15 as before, for SMS_D_Ld1 test the error looks the same as on dashboard ( 806: e3sm.exe 000000000D36E86F m_matattrvectmul_ 267 m_MatAttrVectMul.F90), for SMS.f09_..MALI it is a build fail as on cdash. |
|
Configuring ne30np4 (which ran) atm log has these values (I see nothing alarming)" |
|
Configuring ne30pg2 (which ran) I see nothing alarming: |
|
This is ready. |
…3368) Add XML defaults for theta-l dycore Remove unused/obsolete dycore variables from XML defaults Add code to allow namelist defaults to be set based on value of cam_target Removed se_ftype=2 from all use cases (and made this a default) Support new timestep interface (se_tstep, dt_tracer_factor, dt_remap_factor) for theta-l dycore Update all theta "testmods" test cases to preserve original settings [BFB] with nlfails.
|
merged to next |
…3368) Add XML defaults for theta-l dycore Remove unused/obsolete dycore variables from XML defaults Add code to allow namelist defaults to be set based on value of cam_target Removed se_ftype=2 from all use cases (and made this a default) Support new timestep interface (se_tstep, dt_tracer_factor, dt_remap_factor) for theta-l dycore Update all theta "testmods" test cases to preserve original settings [BFB] with nlfails.
Add XML defaults for theta-l dycore
Remove unused/obsolete dycore variables from XML defaults
Add code to allow namelist defaults to be set based on value of cam_target
Removed se_ftype=2 from all use cases (and made this a default)
Support new timestep interface (se_tstep, dt_tracer_factor, dt_remap_factor) for theta-l dycore
Update all theta "testmods" test cases to preserve original settings
[BFB] with nlfails.