release/public-v1: update and cleanup of build configurations #192

climbfuji · 2020-08-27T22:42:09Z

Description

(1) Update modulefiles for hera.intel and jet.intel in accordance with NCEPLIBS-external/NCEPLIBS update. It was necessary to build our own netCDF libraries as part of NCEPLIBS-external, because using the default modules led to shared library issues with libsz in the ncep_post executable.

(2) Remove modulefiles, gnumake and cmake configurations for all platforms except level-1 (preconfigured). This is to avoid confusion, because the modulefiles for level 2/3 platforms by definition cannot contain libraries in shared spaces (libraries for the UFS MRW App are installed by the user in his/her own space).

Testing

Manual testing on pre-configured, configurable and limited testing platforms

ufs-weather-model compiles using build.sh and compile_cmake.sh on the following pre-configured platforms: hera.intel, cheyenne.intel, cheyenne.gnu, gaea.intel, jet.intel
ufs-weather-model compiles on tier-2 platforms (orion, stampede) using build.sh
ufs-weather-model compiles on macOS with clang+gfortran and gcc+gfortran using build.sh
ufs-weather-model compiles on Ubuntu 18 and RedHat 8 Amazon cloud instances using build.sh

Regression testing on cheyenne.intel, cheyenne.gnu, hera.intel

Regression tests against existing baselines:

all tests pass on cheyenne.gnu and cheyenne.intel
rt_cheyenne_gnu.log
rt_cheyenne_intel.log
on hera.intel, all tests fail due to netCDF/HDF5 metadata mismatches (caused by the HDF5+netCDF library update) in atmos_4xdaily*.nc and RESTART/*.nc files; however, for all of the tests, the phyf*.nc and dynf*.nc files are b4b identical
rt_hera_intel.log
rt_hera_intel_fail_test.log
log_hera.intel.tar.gz

Creating new baselines ufs-public-release-20200728:

for cheyenne.{intel,gnu}, I simply copied ufs-public-release-20200618 to ufs-public-release-20200728
on hera.intel: done

Final regression testing against new baselines:

on cheyenne.{intel,gnu}, all tests pass; logs updated in the PR
rt_cheyenne_gnu.log
rt_cheyenne_intel.log
on hera.intel, all tests pass; logs updated in the PR
rt_hera_intel.log

Dependencies

waiting on release/public-v1: update submodule pointer for ccpp/physics 2020/08/27 NOAA-EMC/fv3atm#163

…s except level-1 platforms (preconfigured platforms)

…nd testing

uturuncoglu · 2020-08-28T15:55:11Z

@climbfuji @ligiabernardet this PR along with #191 #193 seems changing the build system. It seems NCEPLIBS_DIR is no longer used. If so, this will affect the CIME build that requires additional work and we need to discuss about it. Are we missing something in here? If you could clarify it that would be great.

climbfuji · 2020-08-28T15:56:37Z

@climbfuji @ligiabernardet this PR along with #191 #193 seems changing the build system. It seems NCEPLIBS_DIR is no longer used. If so, this will affect the CIME build that requires additional work and we need to discuss about it. Are we missing something in here? If you could clarify it that would be great.

NCEPLIBS_DIR is still to be used for the tier-2 and tier-3 platforms, just not for the tier-1 platforms.

jedwards4b · 2020-08-28T15:57:35Z

So this will break the build on tier-1 systems.

climbfuji · 2020-08-28T16:01:49Z

So this will break the build on tier-1 systems.

No, why? It was decided to load the tier-1 modulefiles from ufs-weather-model instead of generating it from CIME. These modulefiles still set NCEPLIBS_DIR and should work out of the box. They (or, more precisely, the standalone ufs-weather-model build) don't rely on NCEPLIBS_DIR.

uturuncoglu · 2020-08-28T16:10:06Z

@climbfuji as I remember, we did not decide to add that capability or use it, we just discussed about its possibility. We did not even test it and we don't know it is working or not for this particular case. It also creates additional problem with other UFS applications such as HAFS. Currently, we are using same machine file for all UFS apps and the changes need to be test with all of them to be sure it is not breaking anything in other apps. So, that changes might require additional changes in the CIME side and it could have side effects. As I know, we need to keep the changes minimal in this release compared to v1.0. We already make couple of changes in the CIME interface.

climbfuji · 2020-08-28T16:14:24Z

@climbfuji as I remember, we did not decide to add that capability or use it, we just discussed about its possibility. We did not even test it and we don't know it is working or not for this particular case. It also creates additional problem with other UFS applications such as HAFS. Currently, we are using same machine file for all UFS apps and the changes need to be test with all of them to be sure it is not breaking anything in other apps. So, that changes might require additional changes in the CIME side and it could have side effects. As I know, we need to keep the changes minimal in this release compared to v1.0. We already make couple of changes in the CIME interface.

I have no preference on how CIME is using or generating the modulefiles. If you think it is easier to keep the MRW v1.0 way of generating them, then that is ok with me as long as they are updated and match the ones in ufs-weather-model for the tier-1 platforms. I will also update the doc/README_*.txt files in the release/public-v1 branch of NCEPLIBS-external so that they provide the correct information for all (tier 1, tier 2, tier 3) systems. The two files that need to change, as far as I am aware of, are doc/README_hera_intel.txt and doc/README_jet_intel.txt. I will tag you in the NCEPLIBS-external PR.

uturuncoglu · 2020-08-28T16:16:06Z

@climbfuji Thanks. Yes, I think it is better and easier to be close as much as possible to v1.0 structure to prevent any potential error. Anyway, we could discuss it in the call.

climbfuji · 2020-08-28T16:18:33Z

@junwang-noaa @DusanJovic-NOAA this PR and the fv3atm PR NOAA-EMC/fv3atm#163 are ready to merge.

…ysics

climbfuji

The release/public-v1 branch needs modulefiles only for pre-configured platforms for the release. These modulefiles have to match what is stored in the CIME machine config. The module files for the other machines (wcoss, ...) were outdated because not used and not tested. Cheyenne is kept, by the way (for intel and gnu).

netCDF is provided by the NCEPLIBS-external install as described in the PR above, to fix linker issues on the NOAA RDHPC platforms after the UPP (ncep_post) updates were made.

junwang-noaa · 2020-08-28T17:34:45Z

cmake/configure_wcoss_cray.cmake

@@ -1,34 +0,0 @@
-message("")


I am curious why wcoss platforms configuration files are removed, aren't they level-1 platforms?

No, they are not supported at all for UFS MRW v1.0 or v1.1: https://github.com/ufs-community/ufs/wiki/Supported-Platforms-and-Compilers

* Fix to allow quilting with non-factors for layout (ufs-community#244) * Remove the inline comments

Mods to update real-time data on Jet and update pre-defined Alaska domains/GFDLgrid option for GSD 3-km CONUS domain

climbfuji added 5 commits August 27, 2020 15:40

Remove modulefiles, gnumake and cmake configurations for all platform…

26c0f57

…s except level-1 platforms (preconfigured platforms)

Update modulefiles for hera.intel and jet.intel

6a686d7

Update .gitmodules and submodule pointer for fv3atm for code review a…

a36d819

…nd testing

tests/rt.sh: update regression test baseline date tag

7e9e181

Regression test logs for cheyenne.intel and cheyenne.gnu

cf7618e

climbfuji marked this pull request as ready for review August 28, 2020 13:14

climbfuji requested review from junwang-noaa and DusanJovic-NOAA August 28, 2020 13:14

climbfuji mentioned this pull request Aug 28, 2020

Error building CCPP file radiation_aerosols.f with cime. ufs-community/ufs-mrweather-app#169

Closed

Regression test logs for hera.intel

43df5f7

climbfuji mentioned this pull request Aug 28, 2020

release/public-v1: update README for hera and jet NOAA-EMC/NCEPLIBS-external#67

Merged

Revert change to .gitmodules and update submodule pointer for ccpp-ph…

4a751e1

…ysics

DusanJovic-NOAA approved these changes Aug 28, 2020

View reviewed changes

climbfuji commented Aug 28, 2020

View reviewed changes

junwang-noaa reviewed Aug 28, 2020

View reviewed changes

junwang-noaa approved these changes Aug 28, 2020

View reviewed changes

DusanJovic-NOAA merged commit 409c5a9 into ufs-community:release/public-v1 Aug 28, 2020

climbfuji mentioned this pull request Aug 28, 2020

doxygen updates for v4.1.0 release NCAR/ccpp-physics#492

Merged

junwang-noaa mentioned this pull request Apr 19, 2021

Bring in flake updates from ccpp-physics. Compile issue with esmf 810bs47 debug on WCOSS_Cray. #532

Merged

13 tasks

pjpegion pushed a commit to NOAA-PSL/ufs-weather-model.p7b that referenced this pull request Jul 20, 2021

Fix to allow quilting with non-factors for layout (ufs-community#192)

b6957d8

* Fix to allow quilting with non-factors for layout (ufs-community#244) * Remove the inline comments

epic-cicd-jenkins pushed a commit that referenced this pull request Apr 17, 2023

Merge pull request #192 from JeffBeck-NOAA/community_develop

776d4e9

Mods to update real-time data on Jet and update pre-defined Alaska domains/GFDLgrid option for GSD 3-km CONUS domain

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

release/public-v1: update and cleanup of build configurations #192

release/public-v1: update and cleanup of build configurations #192

climbfuji commented Aug 27, 2020 •

edited

Loading

uturuncoglu commented Aug 28, 2020

climbfuji commented Aug 28, 2020

jedwards4b commented Aug 28, 2020

climbfuji commented Aug 28, 2020 •

edited

Loading

uturuncoglu commented Aug 28, 2020

climbfuji commented Aug 28, 2020

uturuncoglu commented Aug 28, 2020

climbfuji commented Aug 28, 2020

climbfuji left a comment

junwang-noaa Aug 28, 2020 •

edited

Loading

climbfuji Aug 28, 2020

release/public-v1: update and cleanup of build configurations #192

release/public-v1: update and cleanup of build configurations #192

Conversation

climbfuji commented Aug 27, 2020 • edited Loading

Description

Testing

Manual testing on pre-configured, configurable and limited testing platforms

Regression testing on cheyenne.intel, cheyenne.gnu, hera.intel

Dependencies

uturuncoglu commented Aug 28, 2020

climbfuji commented Aug 28, 2020

jedwards4b commented Aug 28, 2020

climbfuji commented Aug 28, 2020 • edited Loading

uturuncoglu commented Aug 28, 2020

climbfuji commented Aug 28, 2020

uturuncoglu commented Aug 28, 2020

climbfuji commented Aug 28, 2020

climbfuji left a comment

Choose a reason for hiding this comment

junwang-noaa Aug 28, 2020 • edited Loading

Choose a reason for hiding this comment

climbfuji Aug 28, 2020

Choose a reason for hiding this comment

climbfuji commented Aug 27, 2020 •

edited

Loading

climbfuji commented Aug 28, 2020 •

edited

Loading

junwang-noaa Aug 28, 2020 •

edited

Loading