EarthWorks v2.1.001 Branch

[gdicker1]

Incorporate PRs and fixes in externals

[gdicker1]

See also: #34

Features:

• Add RRTMGP by incorporating upstream changes to EarthWorksOrg/CAM

Issues:

• Fixes fully coupled model does not compile or run due to mpas-framework issues. #31
• Fixes ice_boundary.F90 internal compiler error under nvhpc 23.11 compiler  #23 by creating a new mode MPASSI%PRES for compsets to use
• Fixes CAM no longer builds with intel-oneapi compilers #35 by fixing perl logic for Intel-OneAPI compilers

Relevant PRs in externals:

• Modify mpas framework to match mpasa in EW v2.1 mpas-framework#6
• Prescribed seaice mode mpas-seaice#8
• Incorporate upstream changes for RRTMGP CAM#14
• Fix intel-oneapi builds of CAM that include changes from cam6_3_147 or newer CAM#15

[dazlich]

I tested the 32L 120km fully-coupled model. Builds and runs for intel, gnu, and nvhpc compilers on cpu.

Has a build issue with intel-oneapi:
ERROR: Command: '/glade/u/home/dazlich/ewv2.1/post_relv21/components/cam/bld/configure -s -fc_type intel-oneapi -dyn mpas -hgrid mpasa120 -cpl nuopc -usr_src /glade/derecho/scratch/dazlich/frametest.intel-oneapi/SourceMods/src.cam -spmd -nosmp -ocn mpaso -dyn mpas -nlev=32 -mpas_libdir /glade/derecho/scratch/dazlich/frametest.intel-oneapi/bld/atm/obj/mpas' failed with error 'ERROR: intel-oneapi is not a valid value for parameter fc_type: valid values are cray,pgi,intel,gnu,pathscale,ibm,nag,nvhpc,pgi-gpu,nvhpc-gpu' from dir '/glade/derecho/scratch/dazlich/frametest.intel-oneapi/Buildconf/camconf'

[gdicker1]

I tested the 32L 120km fully-coupled model. Builds and runs for intel, gnu, and nvhpc compilers on cpu.

Has a build issue with intel-oneapi: ERROR: Command: ...

Thanks for testing, @dazlich!

It looks like the issue comes from the CAM side in the bld/config_files
/definition.xml file: Lines 224-227 which doesn't include intel-oneapi. We've been able to build/run with this compiler before, right?

<entry id="fc_type" valid_values="cray,pgi,intel,gnu,pathscale,ibm,nag,nvhpc,pgi-gpu,nvhpc-gpu" value="">
Type of Fortran compiler.  Used when -fc specifies a generic wrapper script
such as mpif90 or ftn.
</entry>

[dazlich]

I’ve had successful tests with oneapi as recently as February 20 with the v2.1 release. On Mar 6, 2024, at 2:37 PM, G. Dylan Dickerson ***@***.***> wrote: ** Caution: EXTERNAL Sender ** I tested the 32L 120km fully-coupled model. Builds and runs for intel, gnu, and nvhpc compilers on cpu. Has a build issue with intel-oneapi: ERROR: Command: ... Thanks for testing, @dazlich<https://github.com/dazlich>! It looks like the issue comes from the CAM side in the bld/config_files /definition.xml file: Lines 224-227<https://github.com/EarthWorksOrg/CAM/blob/04bbb089b65aa452a2ae3344cf7ac7c69254e228/bld/config_files/definition.xml#L224-L227> which doesn't include intel-oneapi. We've been able to build/run with this compiler before, right? <entry id="fc_type" valid_values="cray,pgi,intel,gnu,pathscale,ibm,nag,nvhpc,pgi-gpu,nvhpc-gpu" value=""> Type of Fortran compiler. Used when -fc specifies a generic wrapper script such as mpif90 or ftn. </entry> — Reply to this email directly, view it on GitHub<#32 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADS3XWCGFJCNGFKILAWYKCTYW6EBPAVCNFSM6AAAAABDTVQ652VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSOBRHAZTCNRRG4>. You are receiving this because you were mentioned.Message ID: ***@***.***>

[gdicker1]

I would guess it's this change below in CAM/bld/configure

Previously:

if ($fc_type == "oneapi") {$fc_type = 'intel'; }

Changed to in EWOrg/CAM PR#14. Link to line in PR#14

if ($fc_type eq "oneapi") {$fc_type = 'intel'; }

[gdicker1]

Edited this comment above to change the MPAS-SI PR from EWOrg/MPAS-SI #7 to EWOrg/MPAS-SI #8

[gdicker1]

NOTE: this PR now needs to add #35 to the issues and EWOrg/CAM #15 to the PRs

[gdicker1]

Edited this comment above to add #35 and EWOrg/CAM #15

[gdicker1]

...

• Fixes EarthWorks porting to Perlmutter #33 by modifying the perlmutter_ew_debug machine entry in EarthWorksOrg/ccs_configs_cesm
  ...
• Fix case.setup on Perlmutter for perlmutter_ew_debug ccs_config_cesm#17

Dropping references to Perlmutter from this comment according to discussion and commit 8558c58