search

ESMCI / cime

Common Infrastructure for Modeling the Earth
http://esmci.github.io/cime
Other
161 stars 206 forks source link

Outline and practice steps for bringing new CIME to ACME #228

Closed rljacob closed 7 years ago

rljacob commented 8 years ago

I've put together wiki page to start listing the steps needed to bring the new CIME in to ACME. https://github.com/ACME-Climate/cime/wiki/Converting-ACME-from-cime2-to-ESMCI-cime

This issue can be used for discussion and also for reporting on merge attempts.

rljacob commented 7 years ago

My attempts at doing no 2, reverting to ACME's cime2 coupler, isn't working. I can get it to compile and build, but there are still diffs after 1 day of an F-case.

SUMMARY of cprnc: A total number of 389 fields were compared of which 199 had non-zero differences and 0 had differences in fill patterns A total number of 0 fields could not be analyzed A total number of 0 fields on file 1 were not found on file2. diff_test: the two files seem to be DIFFERENT

We're just going to have to upgrade to CIME5 with bfb diffs.

jedwards4b commented 7 years ago

Have you checked that the same compiler versions and flags are being used? Check that the compset is the same. Check that the CAM_CONFIG_OPTS and CLM_CONFIG_OPTS are the same. Check that namelists have only expected changes.

rljacob commented 7 years ago

CAM_CONFIG and CLM_CONFIG are the same. The compset is the same. I checked one or 2 of the bldlog files and the flags were the same. Do I need to check all of them? I diffed the 2 CaseDocs directors and the only diffs are in path names for some vars and pio options (some don't exist in the old cime). Also "drv_threading" is different but I don't think this is built with threading.

jedwards4b commented 7 years ago

I guess the next thing I would try is to reduce the test from 1 day to 1 timestep - often that will reduce the number of differences from 199 to a few or even one and allow you to isolate the problem.

rljacob commented 7 years ago

Trying an X-case first. 1 day. Only 29 fields different.

RMS x2a_Faxx_taux 2.1516E+16 NORMALIZED 1.0400E+00 RMS x2a_Faxx_tauy 2.1516E+16 NORMALIZED 9.5824E-01 RMS x2a_Faxx_lat 2.1516E+16 NORMALIZED 8.8836E-01 RMS x2a_Faxx_sen 2.1516E+16 NORMALIZED 8.4923E-01 RMS x2a_Faxx_lwup 2.1516E+16 NORMALIZED 8.1340E-01 RMS x2a_Faxx_evap 2.1516E+16 NORMALIZED 7.8047E-01 RMS x2a_Fall_flxdst1 4.2421E+16 NORMALIZED 9.2174E-01 RMS x2a_Fall_flxdst2 4.2421E+16 NORMALIZED 8.9204E-01 RMS x2a_Fall_flxdst3 4.2421E+16 NORMALIZED 8.6419E-01 RMS x2a_Fall_flxdst4 4.2421E+16 NORMALIZED 8.3803E-01 RMS l2x_Fall_swnet 4.4431E+16 NORMALIZED 1.1713E+00 RMS l2x_Fall_taux 4.4431E+16 NORMALIZED 1.1226E+00 RMS l2x_Fall_tauy 4.4431E+16 NORMALIZED 1.0778E+00 RMS l2x_Fall_lat 4.4431E+16 NORMALIZED 1.0364E+00 RMS l2x_Fall_sen 4.4431E+16 NORMALIZED 9.9802E-01 RMS l2x_Fall_lwup 4.4431E+16 NORMALIZED 9.6241E-01 RMS l2x_Fall_evap 4.4431E+16 NORMALIZED 9.2926E-01 RMS l2x_Fall_flxdst1 4.4431E+16 NORMALIZED 8.9831E-01 RMS l2x_Fall_flxdst2 4.4431E+16 NORMALIZED 8.6935E-01 RMS l2x_Fall_flxdst3 4.4431E+16 NORMALIZED 8.4221E-01 RMS l2x_Fall_flxdst4 4.4431E+16 NORMALIZED 8.1670E-01 RMS l2x_Flrl_rofsur 4.4431E+16 NORMALIZED 7.9270E-01 RMS l2x_Flrl_rofgwl 4.4431E+16 NORMALIZED 7.7006E-01 RMS l2x_Flrl_rofsub 4.4431E+16 NORMALIZED 7.4868E-01 RMS l2x_Flrl_rofi 4.4431E+16 NORMALIZED 7.2846E-01 RMS x2r_Flrl_rofsur 2.1504E+16 NORMALIZED 1.3471E+00 RMS x2r_Flrl_rofgwl 2.1504E+16 NORMALIZED 1.3086E+00 RMS x2r_Flrl_rofsub 2.1504E+16 NORMALIZED 1.2722E+00 RMS x2r_Flrl_rofi 2.1504E+16 NORMALIZED 1.2377E+00

rljacob commented 7 years ago

The above was entirely due to glc_nec being "10" in the old cime vs 0 in new cime. No diffs when both set to 0. But they're both 0 in the F-case.

rljacob commented 7 years ago

A 1-step F-case had fewer differences. All in the a2x fields:

RMS a2x_Sa_z 1.0093E-05 NORMALIZED 1.6035E-07 RMS a2x_Sa_u 1.1808E-04 NORMALIZED 2.4652E-05 RMS a2x_Sa_v 2.7711E-05 NORMALIZED 8.2672E-06 RMS a2x_Sa_tbot 4.6655E-05 NORMALIZED 1.6367E-07 RMS a2x_Sa_ptem 4.6754E-05 NORMALIZED 1.6367E-07 RMS a2x_Sa_shum 1.0003E-08 NORMALIZED 1.0767E-06 RMS a2x_Sa_pbot 2.7216E-05 NORMALIZED 2.7814E-10 RMS a2x_Sa_dens 2.0675E-07 NORMALIZED 1.7260E-07 RMS a2x_Sa_pslv 4.2512E-04 NORMALIZED 4.2005E-09 RMS a2x_Faxa_rainc 1.0878E-11 NORMALIZED 5.1502E-07 RMS a2x_Faxa_rainl 6.1514E-09 NORMALIZED 2.3283E-03 RMS a2x_Faxa_snowc 4.4108E-13 NORMALIZED 2.0851E-06 RMS a2x_Faxa_snowl 1.8591E-10 NORMALIZED 6.5354E-04 RMS a2x_Faxa_bcphidry 5.7006E-18 NORMALIZED 3.1405E-02 RMS a2x_Faxa_bcphodry 2.6738E-16 NORMALIZED 3.3257E-03 RMS a2x_Faxa_bcphiwet 5.3080E-17 NORMALIZED 1.9593E-03 RMS a2x_Faxa_ocphidry 1.1813E-17 NORMALIZED 9.4429E-03 RMS a2x_Faxa_ocphodry 8.5546E-16 NORMALIZED 2.1344E-03 RMS a2x_Faxa_ocphiwet 3.7915E-16 NORMALIZED 2.7591E-03 RMS a2x_Faxa_dstwet1 8.0593E-16 NORMALIZED 3.7028E-03 RMS a2x_Faxa_dstwet2 1.1030E-14 NORMALIZED 9.6611E-03 RMS a2x_Faxa_dstwet3 1.9332E-14 NORMALIZED 1.0219E-02 RMS a2x_Faxa_dstwet4 1.2580E-14 NORMALIZED 1.0245E-02 RMS a2x_Faxa_dstdry1 8.2436E-17 NORMALIZED 1.3257E-05 RMS a2x_Faxa_dstdry2 1.6117E-15 NORMALIZED 1.8823E-05 RMS a2x_Faxa_dstdry3 3.0055E-15 NORMALIZED 2.0065E-05 RMS a2x_Faxa_dstdry4 1.9614E-15 NORMALIZED 2.0125E-05

jedwards4b commented 7 years ago

Carefully compare atm logs and initial conditions.

rljacob commented 7 years ago

All the initial conditions are in atm_in? Because those are identical.

I diffed the logs and they're identical until we get to some QNEG3 warning messages. This one:

QNEG3 from TPHYSBCb:m= 4 lat/lchnk= 1537 Min. mixing ratio violated at 5 points. Reset to 0.0E+00 Worst =-1.6E- 12 at i,k= 5 25

Appears in ACME with cime2 but not with cime5.

rljacob commented 7 years ago

Found it. Depends.Intel is different, with different files from the atmosphere, in cime2 and cime5. Copying over the cime2 one and recompling makes the files match for 1 step and 1 day.

jedwards4b commented 7 years ago

good news.