BGCEXP_BDRD_CNPRDCTC_20TR compset fails on master and maint-1.1

[rljacob]

The following case will fail on both master and the head of the maint-1.1 branch.

--compset BGCEXP_BDRD_CNPRDCTC_20TR --res ne30_oECv3

The error messages are: ERROR: SCATTER_FIELD_TO_CHUNK error: hdim1d < hdim1_d

stack trace: 396: e3sm.exe 000000000317A84A Unknown Unknown Unknown 396: e3sm.exe 0000000002DB07B0 shr_abort_modmp 114 shr_abort_mod.F90 396: e3sm.exe 00000000005A31F5 phys_grid_mp_scat 2231 phys_grid.F90 396: e3sm.exe 0000000000551736 co2_data_fluxmp 290 co2_data_flux.F90 396: e3sm.exe 00000000005500AC co2_cycle_mpco2 232 co2_cycle.F90 396: e3sm.exe 0000000000610781 physpkg_mp_phys_i 816 physpkg.F90 396: e3sm.exe 0000000000500479 cam_comp_mp_cam_i 178 cam_comp.F90 396: e3sm.exe 00000000004F568B atm_comp_mct_mp_a 260 atm_comp_mct.F90 396: e3sm.exe 0000000000438A04 component_modmp 267 component_mod.F90 396: e3sm.exe 000000000042763B cime_comp_modmp 1178 cime_comp_mod.F90 396: e3sm.exe 0000000000435A9C MAIN__ 92 cime_driver.F90

Seen on compy with intel compiler.

[rljacob]

Not sure why I chose that compset but I was trying to make a baseline on maint-1.1 to test https://github.com/E3SM-Project/E3SM/pull/3268

[singhbalwinder]

Sometimes we see this error if we are not interpolating an input file (on FV grid) which is being read when the model is running on an SE grid.

[singhbalwinder]

Looking closely, it looks like, I fixed this error by adding a code which is turned on only with a CPP directive added to CAM_CONFIG_OPTS. That code invoked the needed interpolation. We are revisiting that code now and will likely push a fix after we complete our testing.

[rljacob]

I remember why I chose that compset. I wanted to run BGCEXP_BCRC_CNPRDCTC_1850 which is in e3sm_integration and works on compy but that compset isn't in maint-1.1. So I chose one that looked close and was in maint-1.1: BGCEXP_BDRD_CNPRDCTC_20TR. Several days later....

Is there a BGCEXP case I can run out-of-the-box on maint-1.1? I may have to modify the PIO_TYPENAME and set finidat to empty but don't want to do much else.

[singhbalwinder]

I am not very familiar with the BGC compsets. @susburrows @maltrud @beharrop might know which compset would be the best to use for this purpose.

[maltrud]

@rljacob and @singhbalwinder : the CBGC cases don't run out of the box--there are a lot of tweaks and setup stuff done in the run_e3sm script. Balwinder was right that one thing (among many others) that gets set in the script is for correct interpolation of the co2 flux file: $xmlchange_exe --id CAM_CONFIG_OPTS --append --val='-cppdefs -DCO2_BILIN_REGRID' we could probably set up a sample run_e3sm appropriate for testing if you want.

[rljacob]

Someone should work to make CBGC cases run out-of-the-box. I found --res ne30_oECv3 --compset BGCEXP_CNTL_CNPRDCTC_1850 works so will use that.

[susburrows]

@rljacob the BGCEXP_CNTL_CNPRDCTC_1850 is a good one to use for testing purposes for now. The PR Balwinder is working on should resolve the specific error identified in your initial issue report, once it is completed. Is there a timeline for when we need the BGC compsets to be working out-of-the-box, from the overall development schedule or other considerations?

[rljacob]

At the latest, they should be working out-of-the-box before any papers using them are published. Ideally, they should have been working when we made the v1.1.0 (or v1.1.1) tags.