correct benchmark diag_tables for coupled model configurations; move bm_ic directory out of inputdata directory; use aws ec2 for CI test; auto-rt fixes

[DeniseWorthen]

Description

• Adds configuration variable DIAG_TABLE to allow different diag_tables to be set for the coupled model depending on configuration: non-benchmark, benchmark and benchmark-v16. The benchmark and benchmark-v16 diag tables are identical except for the removal of the grid_spec, atmos_4xdaily and atmos_static files. The benchmark-v16 diag table differs from the diag_table_gfsv16 in the input data directory FV3_input_data384 by the ocean output fields.

• adds configuration variable PSM_BC for the coupled model. Default value is 0. In Benchmark v16 tests set to 1

• sets DT_ATMOS=225s for benchmark v16 tests, DDMP=0.2 and FSICL=0

• implements a date-tagged BM_IC directory (set using INPUTDATA_ROOT_BMIC) in cpl_bmark_run.IN for the MOM6 and CICE6 ICs when running the benchmark configuration tests. For WW3, only the 35d tests use ICs in this directory. This directory is a copy of the BM_IC directory in the inputdata-20210212 directory.

• Adds cpld_bmarkfrac_v16_35d (not used in RT testing but was left out in prior commit of v16 tests).

Combined with PRs:

#431
#422
#452

Issue(s) addressed

fixes #412
fixes #430
fixes #307
fixes #447
fixes #449

Testing

The diag_table changes were tested prior to the other v16 test changes. Those tests on hera.intel showed that benchmark coupled baselines on Hera change only because of the difference in the fields written to the phyf and dynf files. Restart files for all components are b4b with current baselines for benchmark baselines. Non-benchmark baselines for the coupled model reproduce existing baselines.

Changes added later to this PR (DT_ATMOS, PSM_BC) for the v16 test will also change baselines for v16 tests.

No changes are due to PRS #431 and #422, #452

RT Log Checks:

• hera.intel
• hera.gnu (auto)
• orion.intel (auto)
• cheyenne.intel
• cheyenne.gnu (auto)
• gaea.intel (auto)
• jet.intel
• wcoss_cray
• wcoss-dell-p3

Co-author: @MinsukJi-NOAA
Co-author: @BrianCurtis-NOAA

[climbfuji]

Question regarding #430 - BM_IC doesn't have any date tag on it. How do we deal with updates to this directory once it sits outside of input-data-YYYYMMDD?

[DeniseWorthen]

The MOM6 and CICE6 IC have not changed since they were generated especially for the prototype project. The FV3 input for the prototypes (which has been changing) is carried in the input data directory FV3_input_frac/BM_IC

[climbfuji]

The MOM6 and CICE6 IC have not changed since they were generated especially for the prototype project. The FV3 input for the prototypes (which has been changing) is carried in the input data directory FV3_input_frac/BM_IC

So we do not expect any changes at all, whatsoever, to the directory that has been moved out input-data-YYYYMMDD? Because if we do, even if it is rare, this would be the time to add its own date tag to it.

Another comment regarding the changes in this PR (DIAG_TABLE). It seems to me that the uncoupled tests could also make use of this capability in the future? Currently we have different diag_table files sitting in the various FV3_input_data* directories.

[BrianCurtis-NOAA]

I'm looking into something, i'll run those two tests in a little while.

[DeniseWorthen]

I ran cheyenne.intel manually and committed the log. I suspected something is wrong on that one too for the auto-rt.

I had just finished running the baseline creation on hera.intel, which is why I added the label.

[BrianCurtis-NOAA]

I ran cheyenne.intel manually and committed the log. I suspected something is wrong on that one too for the auto-rt.

I have zero access to that, so its using the older code. Have to wait for Dom to get back from his PTO.

[DeniseWorthen]

I'm running the hera-intel now manually since the auto-rt didn't seem to run.