Closed ekluzek closed 2 years ago
Note, that this test did seem to work.
PEM.f19_g17.IHistClm45BgcQianGs.cheyenne_intel.clm-realistic_fromCLM5_1850
This might just be that answers change for different processor count with CICE.
OK, I've got two tests that are passing fine:
PEM.f19_g16.F2000Cam6SlimRsGs.cheyenne_intel.clm-2000_CMIP6_AMIP_1deg_ensemble
PEM.f19_g17.IHistClm45BgcQianGs.cheyenne_intel.clm-realistic_fromCLM5_1850
I'm also running a restart test that changes processors so will see if that works.
So I'm going to label this as a WONTFIX since it's probably just the fact that CICE is used prognostically in these SOM cases and that's a compset that is known to change answers on different PE count.
The following tests are failing showing that the fill pattern for some history variables change with a different processor count.
PEM.f19_g16.H_MML_2000_CAM6.cheyenne_gnu.clm-global_uniform_g16_SOM
PEM.f19_g16.H_MML_2000_CAM6.cheyenne_intel.clm-global_uniform_g16_SOM
PEM.f19_g17.F2000Cam6SlimRsGs.cheyenne_intel.clm-2000_CMIP6_AMIP_1deg_ensemble