Jgfouca/branch for acme split 2019 06 10 #3139

jgfouca · 2019-06-10T21:32:34Z

Changes:

Fix bug in PET timestamp (need to use build as a library, not a subprocess) (Fix bug in PET timestamp E3SM-Project/E3SM#2949)
Fix bug in generating baselines (Need to use the base case id when parsing test opts.) (Fix bug in generating baselines E3SM-Project/E3SM#2950)
Change FFLAGS_NOPT to be an append to the normal set of flags. This reduces duplication with FFLAGS (Make atm build using CMake! E3SM-Project/E3SM#2922)
Add ability to ask for CMake args (get_standard_cmake_args) (Make atm build using CMake! E3SM-Project/E3SM#2922)
CMake macros should NOT be looking at environment, just CMake vars (Make atm build using CMake! E3SM-Project/E3SM#2922)
case_setup will make both macro types (Make and CMake) (Make atm build using CMake! E3SM-Project/E3SM#2922)
Import a FindPetsc CMake module (Make atm build using CMake! E3SM-Project/E3SM#2922)
Implement a more-advanced way of sharing builds within test suites (Implements a concept of build groups within test_scheduler.) (Implement a more-advanced way of sharing builds within test suites E3SM-Project/E3SM#2961)

Test suite: scripts_regression_tests
Test baseline:
Test namelist changes:
Test status: [bit for bit, roundoff, climate changing]

Fixes [CIME Github issue #]

User interface changes?:

Update gh-pages html (Y/N)?:

Code review:

and cam is to just put all the sources into cam

[BFB] - Bit-For-Bit See confluence for a more detailed description about these tags.

Added logic for compy to provenance.py, specified the location of the performance archive in the compy entry in config_machines.xml (enabled for all projects) and added the job progress monitoring script syslog.compy. Also removed performance archiving support for edison, now that it has been decommissioned. [BFB] - Bit-For-Bit

* origin/master: (159 commits) Update CIME to ESMCI cime5.8.2-3 (PR #2938) Update namelist scripts to reflect changes in mpas-ocean Registry files Update mpas source. bless_test_results: Be able to handle build-only tests Fix machine file entry for C++ compiler on melvin. Add x2g_*_from_lnd lists for use in prep_glc LND routines Update config_machines.xml Fix optics allocation Cleanup formatting in compute_melt_fluxes() Remove no longer-used cime_model variable Fix unallocated cloud optics Fix unallocated band midpoints COMPOSE: Fix the scope of the share_kokkos_ut standalone unit tests. Update CIME to ESMCI cime5.8.2-2 (#2908) Minor cleanup to previous commit Add default to PIO_CLOBBER if mode unset Test with changing logic for ocn-glc coupling Fix formatting and whitespace Remove `mode_in` in calls to `cam_pio_createfile` Add subroutine call to get PIO format ...

Need to use build as a library, not a subprocess. [BFB]

Need to use the base case id when parsing test opts. [BFB]

Adding an FC test with theta SL Adding a short test for coupled run with theta-l and SL. [BFB] * oksanaguba/eam/fc-test-thetasl: addign the test

Add coupled dycore tests using shared execs Add tests for coupling mechanisms in EAM and for SL transport with theta dycore for coupled atmosphere. Tests re-use executables; only 2 builds are required, one for preqx dycore and another one for theta. [Non-BFB] because new baselines are required. * origin/oksanaguba/eam/tests-ftypes2: reducing dtime for all theta + ne4 tests, still need to try it on skybridge moving tests to integration change hires settings to lowres moving thetasl test to shared execs set start over with tests

* esmci_remote_for_split/master: (1734 commits) Update for cime5.8.3 update esmf nuopc build on cheyenne fix nuopc build, update for stamepede esmf lib fixes for nuopc scripts_regressions_tests to work do not move mediator build Add logic to control activation of glcshelf_c2_ice fix py3 issue fix nuopc K_TestCimeCase rework nuopc build updates for nuopc on stampede fix build issues for nuopc, add dir to .gitignore Use idmap_ignore for optional mapping files revert change fix indentation Move MAP xml variables into shared config_component.xml make test optional Cleanup whitespace / indentation move Externals.cfg to ../Externals_cime.cfg (UFSCOMP) Minor refactor of kokkos stuff in build.py addition uf usermods documentation ... # Conflicts: # scripts/lib/CIME/build.py # src/components/data_comps/datm/cime_config/namelist_definition_datm.xml # src/drivers/mct/main/cime_comp_mod.F90

jgfouca · 2019-06-10T21:33:25Z

config/cesm/machines/config_compilers.xml

@@ -511,6 +503,38 @@ using a fortran linker.
  <TRILINOS_PATH MPILIB="mpich2">$ENV{TRILINOS_PATH}</TRILINOS_PATH>
 </compiler>

+<compiler MACH="athena">


These additions are suspect.

seem to be a repeat of lines starting at 474

OK, i will remove

…-threads-for-mpi-only-runs" This reverts commit 89fad14, reversing changes made to 5c43a15.

So test drivers need to be updated to pass Cmake data as arguments.

billsacks · 2019-06-11T20:06:59Z

@jgfouca with this change

Change FFLAGS_NOPT to be an append to the normal set of flags. This reduces duplication with FFLAGS

What is the purpose of FFLAGS_NOOPT as compared with <append DEBUG="TRUE">? It seems that the two are now accomplishing the same thing, so it seems confusing to have both of them.

Also, have you or someone else verified that the resulting flags for a debug build of cesm are unchanged for all compilers for which we have FFLAGS_NOOPT defined? If not yet, I'd like to request that that comparison be done before this PR is merged to master, to avoid any surprises.

billsacks · 2019-06-11T20:08:32Z

(Adding myself as a reviewer so that I can mark when I feel my above comment has been addressed.)

jgfouca · 2019-06-11T20:22:37Z

What is the purpose of FFLAGS_NOOPT as compared with ? It seems that the two are now accomplishing the same thing, so it seems confusing to have both of them.

@billsacks , that's a good point, they are very similar now. I think the only difference is that FFLAGS_NOOPT guarantees that the flags come later in the compile line than the normal FFLAGS. EDIT: they are actually different concepts: DEBUG is set to true when user requests a debug build, but we have certain special files that always disable optimizations because they would take too long to compile.

I did confirm that the builds we as expected for us in cam (using CMake). I used the bld_diff tool to validate.

jgfouca · 2019-06-11T20:24:58Z

@billsacks , as an example of a NOOPT file:

rrtmg_lw_k_g.o: rrtmg_lw_k_g.f90
        $(FC) -c $(FPPFLAGS) $(INCLDIR) $(INCS) $(FREEFLAGS) $(FFLAGS) $(FFLAGS_NOOPT) $<

rrtmg_sw_k_g.o: rrtmg_sw_k_g.f90
        $(FC) -c $(FPPFLAGS) $(INCLDIR) $(INCS) $(FREEFLAGS) $(FLLAGS) $(FFLAGS_NOOPT) $<

billsacks · 2019-06-11T20:32:00Z

@jgfouca Thanks a lot for the bulleted summary of the major changes in here. In the future, for cime splits like this that include a lot of different changes that potentially impact CESM as well as E3SM (i.e., a lot more than just e3sm-specific config changes), it would help if you could also include pointers to the E3SM PR associated with each bulleted item, to make it easier for us to review particular changes separately. I have done that this time by editing your initial comment (hope you don't mind).

billsacks

@jgfouca thanks for your clarification about FFLAGS_NOOPT. However, now I'm confused by what did and didn't change here. Given that you changed the definition of FFLAGS_NOOPT in config_compilers.xml, I expected to see all references to $(FFLAGS_NOOPT) changed to $(FFLAGS) $(FFLAGS_NOOPT). I see this in the Makefile, but not in the various Depends files in cime/config/machines/Depends.*. Am I missing something?

Also, it seems like this change will change behavior for CESM with intel: files compiled with FFLAGS_NOOPT will now have all of -qno-opt-dynamic-align -convert big_endian -assume byterecl -ftz -traceback -assume realloc_lhs -fp-model source in the compile line, followed by -O0. @jedwards4b is this acceptable / desirable?

Also see one inline comment about a typo.

That said, I do like this cleanup in general.

billsacks · 2019-06-11T20:39:39Z

scripts/Tools/Makefile


 rrtmg_sw_k_g.o: rrtmg_sw_k_g.f90
-	$(FC) -c $(FPPFLAGS) $(INCLDIR) $(INCS) $(FREEFLAGS) $(FFLAGS_NOOPT) $<
+	$(FC) -c $(FPPFLAGS) $(INCLDIR) $(INCS) $(FREEFLAGS) $(FLLAGS) $(FFLAGS_NOOPT) $<


Typo: FLLAGS should be FFLAGS

Thanks! Not sure how that made it through my verification...

Oh, I wasn't using the Makefile to build those (they are in cam, which switched to CMake). And the baselines were generated with master.

jgfouca · 2019-06-11T22:49:34Z

@billsacks , you're right, I neglected to update some of the Make Depends files. Doing that now.

jgfouca · 2019-06-11T22:53:15Z

OK, all tests are passing for me now, so I'll merge once you guys approve.

billsacks · 2019-06-12T16:35:36Z

@jgfouca thanks for these fixes. This looks fine to me, now, but I'd still like @jedwards4b to confirm the change I called out in #3139 (review)

jedwards4b · 2019-06-12T16:38:27Z

@billsacks I don't think that it will be a problem, overall I think that the cleanup is worth the risk.

jgfouca · 2019-06-12T18:02:27Z

Thanks, @billsacks !

jgfouca and others added 30 commits May 2, 2019 17:47

Progress on cam CMakeLists file

638ff00

Add FindPETSc

1546442

done with raw porting, now comes debugging

37e95b1

Fixes

337d52d

Tool for validating transition to CMake

2426d5d

Change FFLAGS_NOOPT to be an append

1bcc269

Progress changing cam/buildib to python

642c941

Remove Depends files for unsupported platforms

0be6d8d

Port Depends to cmake

9d7622b

Get cosp integration working

a808912

bld_diff can run with any python

328bd4f

Homme needs to use GMAKE_J and should do a VERBOSE build

d6205d7

bld_diff should not fail so easily

ebe3766

The best way to deal with the circular deps between cosp

586f317

and cam is to just put all the sources into cam

Update cori's cmake version

229d622

Update anvil cmake

404eb56

Update Livermore Computing configs

5c0c773

[BFB] - Bit-For-Bit See confluence for a more detailed description about these tags.

Fixes for anvil

513c2cb

Update theta cmake

ad1441d

Fix for bebop

8cfdc8f

addign the test

c425be2

Fix configure to work with cmake Depends files

26b4b2a

Fix PNETCDF for some platforms

733e188

Another try for pnetcdf

928a202

Fix bug in PET timestamp

0855336

Need to use build as a library, not a subprocess. [BFB]

Fix bug in generating baselines

481075c

Need to use the base case id when parsing test opts. [BFB]

Better handling of kokkos

c2b8aa6

Merge branch 'oksanaguba/eam/fc-test-thetasl' into master (PR #2942)

3145712

Adding an FC test with theta SL Adding a short test for coupled run with theta-l and SL. [BFB] * oksanaguba/eam/fc-test-thetasl: addign the test

oksanaguba and others added 3 commits June 6, 2019 17:03

moving tests to integration

2b9dc42

jgfouca requested a review from jedwards4b June 10, 2019 21:32

jgfouca self-assigned this Jun 10, 2019

jgfouca commented Jun 10, 2019

View reviewed changes

Remove duped code

b0a787b

jedwards4b approved these changes Jun 10, 2019

View reviewed changes

jgfouca added 5 commits June 10, 2019 16:17

Revert "Merge pull request #3085 from ESMCI/azamat/threading/dont-set…

1878414

…-threads-for-mpi-only-runs" This reverts commit 89fad14, reversing changes made to 5c43a15.

Fix pylint error

180ae0d

Cmake macro no longer relies on environment

54cb3fb

So test drivers need to be updated to pass Cmake data as arguments.

Better fix for cmake macros

1f8d04a

Typo fix

40361f0

billsacks self-requested a review June 11, 2019 20:08

billsacks requested changes Jun 11, 2019

View reviewed changes

Another fix

ae6ee47

Fix some occurances of FFLAGS_NOOPT that now need FFLAGS

7366718

billsacks approved these changes Jun 12, 2019

View reviewed changes

jgfouca merged commit a3d36d3 into master Jun 12, 2019

jgfouca deleted the jgfouca/branch-for-acme-split-2019-06-10 branch June 12, 2019 18:03

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Jgfouca/branch for acme split 2019 06 10 #3139

Jgfouca/branch for acme split 2019 06 10 #3139

jgfouca commented Jun 10, 2019 •

edited by billsacks

Loading

jgfouca Jun 10, 2019

jedwards4b Jun 10, 2019

jgfouca Jun 10, 2019

billsacks commented Jun 11, 2019

billsacks commented Jun 11, 2019

jgfouca commented Jun 11, 2019 •

edited

Loading

jgfouca commented Jun 11, 2019

billsacks commented Jun 11, 2019

billsacks left a comment

billsacks Jun 11, 2019

jgfouca Jun 11, 2019

jgfouca Jun 11, 2019

jgfouca commented Jun 11, 2019

jgfouca commented Jun 11, 2019

billsacks commented Jun 12, 2019

jedwards4b commented Jun 12, 2019

jgfouca commented Jun 12, 2019

Jgfouca/branch for acme split 2019 06 10 #3139

Jgfouca/branch for acme split 2019 06 10 #3139

Conversation

jgfouca commented Jun 10, 2019 • edited by billsacks Loading

jgfouca Jun 10, 2019

Choose a reason for hiding this comment

jedwards4b Jun 10, 2019

Choose a reason for hiding this comment

jgfouca Jun 10, 2019

Choose a reason for hiding this comment

billsacks commented Jun 11, 2019

billsacks commented Jun 11, 2019

jgfouca commented Jun 11, 2019 • edited Loading

jgfouca commented Jun 11, 2019

billsacks commented Jun 11, 2019

billsacks left a comment

Choose a reason for hiding this comment

billsacks Jun 11, 2019

Choose a reason for hiding this comment

jgfouca Jun 11, 2019

Choose a reason for hiding this comment

jgfouca Jun 11, 2019

Choose a reason for hiding this comment

jgfouca commented Jun 11, 2019

jgfouca commented Jun 11, 2019

billsacks commented Jun 12, 2019

jedwards4b commented Jun 12, 2019

jgfouca commented Jun 12, 2019

jgfouca commented Jun 10, 2019 •

edited by billsacks

Loading

jgfouca commented Jun 11, 2019 •

edited

Loading