Add two extra I-Cases into acme test suite

[daliwang]

Add an I1850CLM45CN case on f09_g16 grid and
an I1850CRUCLM45CN case on hcru_hcru grid

more information can be found on "Add two extra I_Cases
into acem test suite" on confluence

We also changed the default machine size to 64 for testing purpose

modified: scripts/acme/update_acme_tests.py
modified: scripts/ccsm_utils/Machines/config_pes.xml
modified: scripts/ccsm_utils/Testlistxml/testlist.xml

[Non-BFB]

LG-9 and SEG-67

[jgfouca]

@daliwang , how come these tests were not entered into scripts/acme/update_acme_tests.py?

[daliwang]

I changed the scripts/ccsm_utils/Testlistxml/testlist.xml directly. Should I use the update_acme_test.py?

[bishtgautam]

@jgfouca : Are you suggesting that apart from changes made to scripts/ccsm_utils/Testlistxml/testlist.xml, one should additionally update the scripts/acme/update_acme_tests.py?

[douglasjacobsen]

When adding tests to the ACME test suites, one should never modify testlist.xml directly. Instead, they should add the tests to update_acme_tests.py and re-run the script to generate testlist.xml.

As is listed at the top of this page: https://acme-climate.atlassian.net/wiki/display/Docs/Modifying+the+tests

[daliwang]

@jgfouca and @mrnorman .

I have problem to run update_acme_test.py on Titan, any suggestion?

FAIL: Command: './manage_xml_entries -removetests -category acme_developer' failed with error 'Can't locate object method "nonBlankChildNodes" via package "XML::LibXML::Element" at ./manage_xml_entries line 583.

[douglasjacobsen]

That looks like a perl issue. Maybe you need to load a perl module before running it?

[jgfouca]

As @douglasjacobsen said, the update_acme_tests is used to manage ACME test suites. If you update testlist.xml directly, then the next time someone updates tests the right way your changes will be lost.

[jgfouca]

@daliwang , for what it's worth, if you're having problem on titan, the update_acme_tests can be run from any machine.

[daliwang]

@jgfouca and @douglasjacobsen. It is the perl problem. I have updated the scripts/acme/update_acme_tests.py and regenerated the testlist.xml. I have also tested these two cases again. (I have already tested whole test suites on Monday). New baselines need to be generated. Thanks.

[bishtgautam]

@douglasjacobsen : Thanks for the link to the webpage on adding tests. I earlier missed it.

[jgfouca]

Looks like you guys updated update_acme_tests. Is this now ready to go?

[daliwang]

@jgfouca. Yes. it is ready to go. Thanks.

[jgfouca]

@daliwang , I had to amend your commit. You added your tests to both acme_developer and acme_integration in update_acme_tests.py. This is not valid because acme_integration is a superset of acme_developer, so the tests only needed to be added to acme_developer. Also, you forgot to update the acme_integration test suite in testlist.xml.

[jgfouca]

@daliwang , it looks like these tests require a number of new input files that are not available in the acme input repository. Is that correct?

[jgfouca]

@jnjohnsonlbl , I do not have the permissions necessary to get the input files I need on edison. Please run:
./create_test -g SMS.f09_g16.I1850CLM45CN SMS.hcru_hcru.I1850CRUCLM45CN
on the next branch on edison.

[jgfouca]

@daliwang , even with the CESG input repo, I still see a ton of these errors:
File status unknown: /home/jgfouca/acme/ptclmdata/atm_forcing.datm7.cruncep_qianFill.0.5d.V4.c130305/Solar6Hrly/clmforc.cruncep.V4.c2011.0.5d.Solr.1901-01.nc

[jgfouca]

@ACME-Climate/developers , any idea why blues is happily downloading dozens of input files while every other platform is just copping out with this message:

File status unknown: /glade/proj2/cgd/tss/atm_forcing.datm7.cruncep_qianFill.0.5d.V4.c130305/Solar6Hrly/clmforc.cruncep.V4.c2011.0.5d.Solr.1901-01.nc
...

[jgfouca]

Nevermind, DIN_LOC_ROOT_CLMFORC was not defined correctly for a number of our machines.

[daliwang]

@jgfouca. Sorry for the late reply. The CLMFORC is new dataset generated by ORNL, along with PR 268 (https://acme-climate.atlassian.net/wiki/display/LND/Generating+new+surface+data+including+soil+order+for+phosphorus+cycle). Thanks.

[jgfouca]

@daliwang , I cannot obtain this file /home/ccsm-data/inputdata/atm/datm7/atm_forcing.datm7.cruncep_qianFill.0.5d.V4.c130305/TPHWL6Hrly/clmforc.cruncep.V4.c2011.0.5d.TPQWL.1907-08.nc

From either the ACME or SVN input data repositories.

[jgfouca]

@daliwang , do these tests really use 100+ years of data?

[daliwang]

@jgfouca. Sorry for the late reply.

Those cases check the data availability (1850-2005) before the simulation. The SMS tests themselves only need 5 days or 1 month. We will need those data for other follow up (long) test cases.

I am working to get the /home/ccsm-data/inputdata/atm/datm7/atm_forcing.datm7.cruncep_qianFill.0.5d.V4.c130305/TPHWL6Hrly/clmforc.cruncep.V4.c2011.0.5d.TPQWL.1907-08.nc into acme-svn2. Is this the only file missing?

[jgfouca]

@daliwang , yes I think that's the only missing file.

[daliwang]

@jgfouca. That file is in ACME repo now. I tried to upload the whole CRU forcing dataset into acme data repo today without success. It will try to get it sorted out late today.

[jgfouca]

@daliwang , thanks, I'm getting that file now. Another thing, I'm seeing this diff:

17198,17199c17198,17199 < 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.0468376233638459, 0.135802914565793, < 0.0981012885553448, 0.0854576027744043, 0.0245657950765278, 0, 0, 0, 0, --- > 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.0487456139386837, 0.146150304607452,
FAIL
hist file comparison is FAIL

For test ERS.f09_g16.B1850C5, is that expected?

[bishtgautam]

@jgfouca : The difference might be because of changes made in processor layout in scripts/ccsm_utils/Machines/config_pes.xml. Aren't there differences in env_mach_pes.xml?

[jgfouca]

Yes, I'm sure that must be it.

[daliwang]

@jgfouca. The default machine size is 384 for the f09_g16. I have changed it to 64 for the testing purpose. Thanks.

[jgfouca]

Integration tests look good and I'm finally done downloading input files on all the test platforms, merging. @jnjohnsonlbl , did you get a chance to download all those input files on edison?

[jeff-cohere]

Retrying. I encountered problems trying to do it earlier with

./create_test -g SMS.f09_g16.I1850CLM45CN SMS.hcru_hcru.I1850CRUCLM45CN -xml_mach edison -xml_category acme_integration -xml_compiler intel

but had to move onto something else, and I don't recall what they were.

[jgfouca]

That create_test command doesn't look right. Is that our script or theirs?

[jeff-cohere]

It's theirs. I'll try the ACME one with your instructions.

[jgfouca]

I don't theirs supports "-g" or listing test cases like that. For ours:
scripts/acme/create_test -g [test1] [test2]
It will use the "next" baselines if that's your current branch.

[jeff-cohere]

Ok, I've run the script and fetched the input files. You're good to go from my end, @jgfouca.

[jgfouca]

Thanks

[singhbalwinder]

@daliwang : Why were these tests implemented using FV grid and not the SE grid?

[jgfouca]

@singhbalwinder , @daliwang i was just about to merge this to master, should I hold off?

[singhbalwinder]

I thought they are already in master. They showed up today when I ran acme developer tests on master

[jgfouca]

Heh, I'm clearly juggling too many things. Yes, I forgot that I already merged this.