In a few places, we trigger the same behavior as for the CAM macro.

EAM's functions are just wrapping share_file_mod's ones, without adding
any logic. This way, we can use the module even if EAM is not used
(e.g., for SCREAM builds).

Also, fixed indentation of some if blocks in namelist_mod.F90

This can happen for SCREAM, where dt is not known at the place
where the initialization happen. It will instead be known
only when timestepping is performed.

To be safe, add a check at the beginning of prim_run_subcycle,
to ensure nsplit has been set to a value>=1.

… a noinline clause to the FFT init to fix it.

Add -fallow-argument-mismatch for all GNU builds (including gnugpu).

Only add -fallow-invalid-boz for 2 specific files for gnu/gnugpu builds.

This PR adds Depends.gnugpu.cmake which is similar to Depends.gnu.cmake

[bfb]

Change the implementation of glacier factor

Fix the soil texture type bug

1) change the equation of snow cover effect on soil erosion; 2) use Manning's coefficient from MOSART

revert to the vegetation-based Manning coefficient method

Add ftillage_tc and flitho into the calculation of Tc

Add the reference to ELM-Erosion v2

…SM-Project#4822)

Add flags for GNU builds to allow using GNU v10 and higher versions

…(PR E3SM-Project#4818)

Update module versions on Perlmutter (including to gnu v10)

…roject#4788)

I added some ifdefs handling the SCREAM macro, mostly in namelist_mod (plus a couple occurrences in prim_[cxx]_driver_base.F90 init routines). These are needed in order for SCREAM to use the readnl subroutine in Homme (rather than dangerously rolling its own namelist parsing, which is what it was doing before).

In addition, there are two cleanups:

    remove some warnings (unused vars, initialization order, inconsistent tabs);
    removed a few namelist options that were never used, either in components/homme or components/eam. By that I mean that no F90 module or C/CXX file was ever using, save for their appearance in control_mod and namelist_mod (which was just parsing them from nl file, and setting them in control_mod).

Notes:

    The ifdef logics should not change anything for both EAM and standalone Homme.
    @oksanaguba @mt5555 I split the removal of the namelist options into separate commits, so that if one of them is indeed to be kept, I can simply revert that single commit.
    The PR changes 190 files, but most of them are homme namelists, where the purged namelist options have been removed.
    I added the 'HOMME' label (rather than 'HOMME standalone'), since this PR also touches the eam/bld folder, in order to remove a couple of the purged namelist options from namelist_definition.xml.

Results from homme and hommebfb tests run by @oksanaguba:
[ac.onguba@chrlogin2 ~]$ ~/ess/cs.status.20220228_121931_1jups0 | grep -i fail
  REP_Ln5.ne4_oQU240.F2010.chrysalis_intel (Overall: FAIL) details:
    FAIL REP_Ln5.ne4_oQU240.F2010.chrysalis_intel RUN time=1948
[ac.onguba@chrlogin2 ~]$ ~/ess/cs.status.20220228_153258_5qqo8a | grep -i fail
    FAIL ERS.f19_g16.I1850GSWCNPECACNTBC.chrysalis_intel.elm-eca_f19_g16_I1850GSWCNPECACNTBC TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL ERS_Ln9.ne4pg2_oQU480.F20TR-MMFXX.chrysalis_intel.eam-mmf_use_VT TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL PET.f19_g16.X.chrysalis_intel.allactive-mach-pet MEMCOMP Error: Memory usage increase >5% from baseline's 1529.550000 to 1529.550000
    FAIL PET.f19_g16.X.chrysalis_intel.allactive-mach-pet TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL PET_Ln5.ne4_oQU240.F2010.chrysalis_intel.allactive-mach-pet TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL SMS.f09_g16_a.IGELM_MLI.chrysalis_intel TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL SMS_Ld20.f45_f45.IELMFATES.chrysalis_intel.elm-fates_rd TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL SMS.MOS_USRDAT.RMOSGPCC.chrysalis_intel.mosart-unstructure TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL SMS.ne30_f19_g16_rx1.A.chrysalis_intel TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline
    FAIL SMS.ne4_oQU240.F2010.chrysalis_intel.eam-cosplite TPUTCOMP Error: TPUTCOMP: Computation time increase > 5% from baseline

[BFB]

Reduce ELM test build times with shared executables.
On Chrysalis,
* before: 74 mins
* after: 17 mins (4x faster)

[BFB]

…nd PGI compilers, but more testing is needed before merging to next again

…oves the need for usin gYAKL_DEVICE_LAMBDA

Update OpenMPI on Chrysalis.

Addresses E3SM-Project#4819

[BFB]

see MPAS-Dev/MPAS-Model#959

This PR fixes non-standard declarations of 8-byte integers in ESMF_BaseMod.F90.

The ESMF_BaseMod module contains several non-standard declarations of 8-byte
integer variables. These were identified using the GNU compiler:

   ESMF_BaseMod.F90:75:20:

      75 |           integer*8 :: ptr
         |                    1
   Error: GNU Extension: Nonstandard type declaration INTEGER*8 at (1)

and also using the NAG compiler:

   Error: ESMF_BaseMod.F90, line 75: KIND value (8) does not specify a valid representation method
   Error: ESMF_BaseMod.F90, line 163: KIND value (8) does not specify a valid representation method

Since the ESMF_BaseMod module already defines a parameter for selecting 8-byte
integers, namely, ESMF_KIND_I8, all that is needed is to move the definition of
this (and related) parameters to appear earlier in the module and to then use

   integer(kind=ESMF_KIND_I8)

rather than

   integer(kind=8)

The new features added in this PR have been described and well validated in Tan et al. (2022). The new features include
the impact of irrigation, conservation agriculture, crop residue, glacier and lithology on soil erosion and sediment yield. As
the current ELM-Erosion, the model can be turned on by `use_erosion = .true.` and the erosion-bgc feedback can be
turned on by `ero_ccycle = .true.`. I will create a new test to cover the new features later.

Tan, Z., Leung, L. R., Li, H.-Y., & Cohen, S. (2022). Representing global soil erosion and sediment flux in Earth System
Models. Journal of Advances in Modeling Earth Systems, 14, e2021MS002756. https://doi.org/10.1029/2021MS002756

[BFB]

Adding new scalar type to pointer retrievals so that messy code does
   not need to be added for OpenACC modifications
Removing unneeded use module statements in EOS routines

…enACC"

This reverts commit c77d2a8.

Upgrading to the latest YAKL. Includes performance improvements for CUDA and HIP as well as a bug (memory leak) fix for the pool allocator. The new YAKL also has a fully functioning SYCL backend for Intel GPUs.

CUDA has what appears to be a compiler bug in the FFT routines. These have been worked around in the new YAKL. The FFT API changed by no longer requiring a parameter to init(). This has been reflected in pressure.cpp and crm_variance_transport.cpp

An issue with the standalone samxx has been fixed, and new machine files have been added.

This PR passes ./create_test SMS_Ld10_P6x7.ne4pg2_ne4pg2.F-MMFXX.summit_gnugpu.eam-rrtmgpxx

[BFB]

This is a combination of 17 commits.

    OpenACC changes for time integration

    Changing pgi-summit compile flags for OpenACC
    Adding scalar get_config routines to simplify OpenACC changes
    Adding GPU vector reconstruct routines
    Adding varinp_ scalar names as alternative to config_ input variable pointer names
    Cleaning up unused use statements in EOS
    Changing input coeff in EOS from pointer to scalar

    Starting to pull openacc directives out of diagnostics into time integration routines

    Changing diagnostics variables to create rather than copyin
    Moving some data movement out of diagnostics

Moving more openacc data movement directives out of diagnositics routine

Moving more OpenACC data movement directives out of diagnostics

Moving rest of OpenACC copyins out of diagnostics

Moving OpenACC data copyout movement directives out of diagnostic routine

Moving more OpenACC data movement out of diagnostics

Moving more OpenACC data movement directives out of diagnostic solve

Moving more OpenACC data movement directives out of diagnostic solve

Moving more OpenACC data movement directives out of diagnostic solve

Adding use of varinp scalar variable in diagnostics routine to simplify OpenACC mods

Extending out OpenACC region in time integration

Extending OpenACC region in time integration routines

Extending OpenACC region to include normal Barotropic/Baroclinic Velocity calcs

Extending OpenACC region above Freq calcs

Moving more OpenACC data directives out of diagnostics

Moving remaining copyouts/updates out of diagnositic solve

…ject#4825)

Fix non-standard declarations of 8-byte integers in ESMF in MPAS-Framework

This PR fixes non-standard declarations of 8-byte integers in
ESMF_BaseMod.F90. See the original PR for MPAS at
MPAS-Dev/MPAS-Model#959. The ESMF_BaseMod module contains several non-
standard declarations of 8-byte integer variables. These were identified
using the GNU and NAG compilers.

[BFB]

Changes:
* add crusher node in config_machines.xml, config_batch.xml
* create gnu_crusher.cmake
* for testing amdclang and crayclang, cmake configurations for the compilers are also added
* add CPRAMD macro in shr_infnan_mod.F90.in to use ieee arithmetic

Add gnu compiler support on Crusher and AMD.

Changes:

    add crusher configuration in config_machines.xml, config_batch.xml
    create gnu_crusher.cmake
    for testing amdclang and crayclang, cmake configurations for the compilers are also added
    add CPRAMD macro in shr_infnan_mod.F90.in to use ieee arithmetic

[BFB]

Extending OpenACC diagnostics up to time integration

Pulling OpenACC data movement directives up into top-level time
integration routines and extending OpenACC region by:
* Changing diagnostics variables to create rather than copyin
* Changing pgi-summit compile flags for OpenACC
* Adding scalar get_config routines to simplify OpenACC changes
* Adding GPU vector reconstruct routines
* Adding varinp_ scalar names as alternative to config_ input variable
  pointer names
* Cleaning up unused use statements in EOS
* Changing input coeff in EOS from pointer to scalar

[BFB]

…variables