Build nceplibs, use external libraries, and build different configurations in parallel

[laurenchilutti]

Description

• BUILDlibfms: This script will build libfms and takes an argument (either intel or gnu) with the default being intel.
• COMPILE now calls BUILDlibfms and BUILDnceplibs to build libFMS and/or nceplibs if it does not exist
• BUILDnceplibs: This script will build the necessary nceplibs and takes an argument (either intel or gnu) with the default being intel.
• When you want to store the nceplibs and libFMS outside of SHiELD_build/Build you can define the environment variable EXTERNAL_LIBS=<path to nceplibs and libFMS> and this scripts (COMPILE, BUILDnceplibs, and BUILDlibfms) will build them there.
• When you want to compile SHiELD with nceplibs and libFMS that are stored externally the environment variable EXTERNAL_LIBS=<path to nceplibs and libFMS> can be set, and the COMPILE script will use these libs to compile with
• Users will see no change if they do not define EXTERNAL_LIBS
• The exec/ directory that is created during the compile now has a sub-directory specific to each unique build configuration so that we can build multiple configurations in parallel. This resulted in changes to some of the SHiELD_build/Build/mk_scripts
• The SHiELD_build/Build/nceplibs prebuilt libs directory is no longer in the repository and nceplibs is now included in the .gitignore file

Fixes # (issue)

How Has This Been Tested?

I have tested this build on Gaea with gnu and intel. I have tested with the CI tests on Parallelworks to see that answers do reproduce

Checklist:

Please check all whether they apply or not

• [x] My code follows the style guidelines of this project
• [x] I have performed a self-review of my own code
• [x] I have commented my code, particularly in hard-to-understand areas
• [x] I have made corresponding changes to the documentation
• [x] My changes generate no new warnings
• [x] Any dependent changes have been merged and published in downstream modules

[lharris4]

This is a good idea. Thanks.

Lucas

On Thu, Sep 8, 2022 at 4:04 PM Rusty Benson @.***> wrote:

@.**** approved this pull request.

In Build/COMPILE https://github.com/NOAA-GFDL/SHiELD_build/pull/15#discussion_r966368566:

#

check to make sure libFMS exists

if [ -d libFMS/${compiler} ] && [ -e libFMS/${compiler}/32bit/libFMS.a ] && [ -e libFMS/${compiler}/64bit/libFMS.a ] ; then echo " pre-built libFMS/${compiler} exists" else

• echo " libFMS/${compiler} does not exist - building libFMS/${compiler}"
• MAKE_libFMS ${compiler} >> buildlibFMS${compiler}.out 2>&1 # build 32bit and 64bit versions of libFMS
• #

• test and report on libFMS build success

• if [ $? -ne 0 ] ; then
• echo ">>> libFMS $compiler build failed"
• exit 2
• fi
• echo " libFMS build successful"
• echo " libFMS/${compiler} does not exist"
• ./BUILDlibfms ${compiler}

I think you need to check the exit code from BUILDlibfms to ensure it completed correctly before continuing.

In Build/mk_scripts/mk_make https://github.com/NOAA-GFDL/SHiELD_build/pull/15#discussion_r966373144:

@@ -77,12 +84,12 @@ $FC --version NCEPLIBS="../../libFMS/${COMPILER}/${bit}/libFMS.a" if [ ${CONFIG} = 'shield' ] ; then NCEPLIBS+=" ./libgfs.a"

• NCEPLIBS+=" -L../../nceplibs/${COMPILER}/ -lbacio_4 -lsp_v2.0.2_d -lw3emc_d -lw3nco_d"
• NCEPLIBS+=" -L../../nceplibs/${COMPILER}/ -lbacio -lsp_d -lw3emc_d -lw3nco_d"

You could set a library path variable in Build/COMPILE that would point to internal or external locations and simply refer to it here. It would save you from managing the various links.

— Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/SHiELD_build/pull/15#pullrequestreview-1101340594, or unsubscribe https://github.com/notifications/unsubscribe-auth/AMUQRVGRUVISXJ6FZENFYXTV5JBDJANCNFSM6AAAAAAQIBULFA . You are receiving this because you are subscribed to this thread.Message ID: @.***>

[laurenchilutti]

I would be supportive of removing the prebuild nceplibs and I can do that in this PR. Would we want to do that in this PR or wait to get input from Lucas and his team?

[lharris4]

I would be fine with removing the pre-built NCEPLIBS as long as they are only built when absolutely necessary, as what is done with FRE.

Linjiong, Kai, Joseph: do you have any opinions on this?

Thanks, Lucas

On Fri, Sep 9, 2022 at 10:23 AM laurenchilutti @.***> wrote:

I would be supportive of removing the prebuild nceplibs and I can do that in this PR. Would we want to do that in this PR or wait to get input from Lucas and his team?

— Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/SHiELD_build/pull/15#issuecomment-1242038269, or unsubscribe https://github.com/notifications/unsubscribe-auth/AMUQRVGYBYRGWZ7PLRFVYL3V5NB5FANCNFSM6AAAAAAQIBULFA . You are receiving this because you commented.Message ID: @.***>

[laurenchilutti]

I would be fine with removing the pre-built NCEPLIBS as long as they are only built when absolutely necessary, as what is done with FRE. Linjiong, Kai, Joseph: do you have any opinions on this? Thanks, Lucas

@JosephMouallem @kaiyuan-cheng @linjiongzhou Would you be able to give your input on this? I can modify the COMPILE script so that the nceplibs gets built similarly to libFMS (when it does not exist or when you use the cleanall argment)

[kaiyuan-cheng]

I think building the NCEPlibs from the source code is neater than the prebuilt.

On Fri, Sep 9, 2022 at 1:21 PM laurenchilutti @.***> wrote:

I would be fine with removing the pre-built NCEPLIBS as long as they are only built when absolutely necessary, as what is done with FRE. Linjiong, Kai, Joseph: do you have any opinions on this? Thanks, Lucas

@JosephMouallem https://github.com/JosephMouallem @kaiyuan-cheng https://github.com/kaiyuan-cheng @linjiongzhou https://github.com/linjiongzhou Would you be able to give your input on this? I can modify the COMPILE script so that the nceplibs gets built similarly to libFMS (when it does not exist or when you use the cleanall argment)

— Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/SHiELD_build/pull/15#issuecomment-1242253214, or unsubscribe https://github.com/notifications/unsubscribe-auth/AR2VX67RECPTTLVMPCEUEJDV5NW2VANCNFSM6AAAAAAQIBULFA . You are receiving this because you were mentioned.Message ID: @.***>

[laurenchilutti]

This should be my final commit for this PR - I have removed the prebuilt nceplibs. If you are compiling for the first time the nceplibs and libfms will be built. If you choose to compile with cleanall libfms, ncepplibs, and the build area (exec) will be deleted and rebuilt.

[laurenchilutti]

I had to make one more round of changes - on Parallelworks in testing I found that the nceplibs weren't building and adding the cmake flag -DCMAKE_C_STANDARD=99 fixes this issue. I also moved the module load cmake to the site/environment..sh files as different systems have different versions of cmake. I have tested on Gaea and Parallelworks and everything works

[JosephMouallem]

looks good and the compilation works on both gaea and orion

[linjiongzhou]

I am totally fine with that. Building it the same way as FMS would be great.

Linjiong

On Fri, Sep 9, 2022 at 1:21 PM laurenchilutti @.***> wrote:

I would be fine with removing the pre-built NCEPLIBS as long as they are only built when absolutely necessary, as what is done with FRE. Linjiong, Kai, Joseph: do you have any opinions on this? Thanks, Lucas

@JosephMouallem https://github.com/JosephMouallem @kaiyuan-cheng https://github.com/kaiyuan-cheng @linjiongzhou https://github.com/linjiongzhou Would you be able to give your input on this? I can modify the COMPILE script so that the nceplibs gets built similarly to libFMS (when it does not exist or when you use the cleanall argment)

— Reply to this email directly, view it on GitHub https://github.com/NOAA-GFDL/SHiELD_build/pull/15#issuecomment-1242253214, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACVIY3SOVQWCD6AIZ57U2S3V5NW2VANCNFSM6AAAAAAQIBULFA . You are receiving this because you were mentioned.Message ID: @.***>