ctsm5.4.024: Hot fix with updated derecho_intel and a few fixes for it#3796
Conversation
… and clarify the strings with a concatenation so it's more clear and there's the needed space between the and and the namelist option name do_transient_urban
|
This is on top of #3786 so that needs to come to master first to show the true differences. |
|
There were 73 tests on Derecho with changed answers, which isn't surprising since this was a compiler update: And there are seven unexpected fails that I need to look into: |
|
The first two fails are an error in the SparseMatrixMultiply for CN matrix with errors that look like this: cesm.log: For these tests: |
|
The five mpi-serial tests that fail: Do so because of a divide by zero that looks like this: |
…he namelist move for ESCOMP#3391 the move_nml_parameters branch
…he paramfile" Temporarily revert the change to fix the coupled model parameter file, so that it comes in on an answer changing tag. This reverts commit 580708d.
…-issues' into ccs_config_1077_derecho_intel_needed_updates
|
@slevis-lmwg and I discussing this. On the CN fails. In #3360 we note that resubmitted enabled the test to pass. So I'll do the same here. Also there is a matrixcn testlist and I should run it, to check more CN matrix tests. Actually looking again it looks like all the matrixcn tests are in aux_clm. Maybe we should delete the matrixcn testlist then? Actually let's keep it because the list is 50 tests and this makes it easier to run just that set of things. |
…use a trim for the strings so it formats better, this seems to be needed in order to be able to change the setting for ch4finundatedmapalgo with the updated intel compiler
…e intel/2025.3.2 issue ESCOMP#3798
slevis-lmwg
left a comment
There was a problem hiding this comment.
@ekluzek thanks! Looks simple enough, and I have one "nit" request:
Seems the fix consists of four lines rather than one, so could you change the title in ChangeLog, ChangeSum, as well as the PR, and anywhere else? UNLESS three of the four updated lines were not required to fix the problem.
I did not look one-by-one at the ExpectedTestFails changes, but I like that mostly there are fewer tests failing.
Description of changes
Update to ccs_config1.0.77 with derecho_intel updates to intel/2024.3.2. And a two line fix on a continuation line in the middle of a string, so that it will build.
Improve the logging about the ch4finundation stream. Some of this seemed to be important in order to change that ch4finundationmapalo.
Remove a bunch of tests in the expected fail files that have been working now. Add a few more newly failing tests.
Specific notes
Contributors other than yourself, if any: @jedwards4b
CTSM Issues Fixed (include github issue #):
Fixes #3791
Fixes #3795
Are answers expected to change (and if so in what way)? Yes, some
Any User Interface Changes (namelist or namelist defaults changes)? No
Does this create a need to change or add documentation? Did you do so? No No
Testing performed, if any: will run regular
So far ran:
SMS.f45_f45_mg37.I2000Clm60FatesSpRsGs.derecho_nvhpc.clm-FatesColdSatPhen
SMS_D.f10_f10_mg37.I1850Clm60BgcCrop.derecho_intel.clm-ciso_soil_matrixcn_only
ERS_D_Ld5.f45_f45_mg37.I2000Clm60Fates.derecho_intel.clm-FatesColdCamLndTuningMode
which all pass and compare identically to the ctsm5.4.022 baseline
After running regular tests 73 compare different to the baseline.