update to latest cmeps master

[DeniseWorthen]

Resolves Issue #17 Resolves Issue #23

Description of changes

• Fixes incorrect mapping of ifrac for both nems_orig and nems_frac modes. The mapping of the ifrac to the atm in med_fraction_mod was done with either conservative fraction (nems_frac) or nstod conservative fraction (nems_orig). It should have been conservative destination or nstod conservative destination. For nems_orig, this replicates the mapping that was done in the nems mediator.

• Adds diagnostic budgets (cesm)

• Fixes bug for land runoff (cesm)

• Adds field packing for better performance.

Baseline Change

• Answers change for all baselines in coupled model. These changes are solely due to fix of ifrac mapping. Other changes made for CESM have no impact on UFS

• No impact on NEMS Datm since no mapping is done back to the Atm

• Performance update to CMEPS (field packing) has no impact on results. Verified by running all ufs-weather s2s and datm regression tests with and without the addition of the performance features. All baselines were b4b.

[DeniseWorthen]

@junwang-noaa @binli2337 Thanks for the reviews. As it turns out, the performance update to cmeps (Issue #23) is also now ready. I've done a test comparing the cpld_control<> tests in this branch with a branch including the performance update. Comparing the coupler restart files between the two, I get b4b results. So the performance update has no impact on answers.

I also did a 5d test run using the cpld_control_c384 test case. Comparing just the wall clock time, the perf run was about 3min faster. For a 35d run, that would be ~20min faster. So the question is whether we go ahead and include the perf update at this time.

[junwang-noaa]

Thanks for doing the tests, I'd suggest to include the performance updates in this PR.

On Fri, Oct 30, 2020 at 8:09 AM Denise Worthen notifications@github.com wrote:

@junwang-noaa https://github.com/junwang-noaa @binli2337 https://github.com/binli2337 Thanks for the reviews. As it turns out, the performance update to cmeps (Issue #23 https://github.com/NOAA-EMC/CMEPS/issues/23) is also now ready. I've done a test comparing the cpld_control<> tests in this branch with a branch including the performance update. Comparing the coupler restart files between the two, I get b4b results. So the performance update has no impact on answers.

I also did a 5d test run using the cpld_control_c384 test case. Comparing just the wall clock time, the perf run was about 3min faster. For a 35d run, that would be ~20min faster. So the question is whether we go ahead and include the perf update at this time.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/NOAA-EMC/CMEPS/pull/22#issuecomment-719516151, or unsubscribe https://github.com/notifications/unsubscribe-auth/AI7D6TOJLIOEUWZCWHYQRL3SNKUIRANCNFSM4TBP4K3Q .

[DeniseWorthen]

OK thanks. As soon as they merge the perf branch into master I will update my branch and merge to emc/develop.