LMDZ Release notes

March 8th, 2024: svn tag LMDZ-testing-20240308, svn revision 4848

There is one additional variable in the physical restart file. Otherwise, results of tests are identical to those of the previous tag, except as noted below.

Change of results since previous tag:

Results for configuration with rad=ecrad have changed. This is expected.

Resolved issues:

On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi, execution of gcm does not crash any longer.

Remaining issues:

Change in running environment (modipsl/libIGCM) needed for ICOLMDZOR_LAM experiments (they are presently too slow).
ECRAD is not yet operational.
With Gnu compilers, for LMDZ configured with -strataer true, we do not have 1+1=2. On jean-zay, with Intel compilers, for LMDZ configured with -strataer true, a restarted run crashes.

October 22nd, 2023: svn tag LMDZ-testing-20231022, svn revision 4729

Trac ticket 152

Change of results since previous tag:

Results of 1D cases have changed because of bugfix 4650
Results for configuration with rad=ecrad have changed. This is expected.

Resolved issues:

With Gnu compilers version 11, for LMDZ configured with rad=ecrad, results with 2 MPI × 2 OpenMP now are identical to results with 1 MPI (no OpenMP).
On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi_omp, we now have 1+1=2.

Remaining issues:

On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi, execution of gcm crashes in radiation_general_cloud_optics_data.F90, line 189, with message:
```
  0: forrtl: severe (408): fort: (8):
  Attempt to fetch from allocatable variable MASS_EXT when it is
  not allocated
```

June 26th, 2023: svn tag LMDZ-testing-20230626, svn revision 4586

On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi_omp, results with LMDZ-testing-20230626 are different than results with LMDZ-testing-202304. This is expected.

Except for the case above, results of runs with LMDZ-testing-20230626 are exactly the same as with LMDZ-testing-202304.

Except for issues described below, we have 1 + 1 = 2 and 2 MPI × 2 OpenMP = 1 MPI.

Issues:

ECRAD
- On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi, execution of gcm crashes in radiation_general_cloud_optics_data.F90, line 189, with message:
```
0: forrtl: severe (408): fort: (8):
Attempt to fetch from allocatable variable MASS_EXT when it is
not allocated
```
- On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi_omp, we do not have 1+1=2.
- With Gnu compilers version 11, for LMDZ configured with rad=ecrad, results with 2 MPI × 2 OpenMP are different than results with 1 MPI (no OpenMP).
ICOLMDZ configurations compile and run but do not converge with previous testing versions. To Be Investigated

April 17th, 2023: svn tag LMDZ-testing-202304, svn revision 4507

This version includes the first merge of the ECRAD aerosols development branch.

Issues:

On jean-zay, with Intel compilers, for LMDZ configured with rad=ecrad, parallel=mpi, execution of gcm crashes in radiation_general_cloud_optics_data.F90, line 189, with message:
```
  0: forrtl: severe (408): fort: (8):
  Attempt to fetch from allocatable variable MASS_EXT when it is
  not allocated
```

On a PC, with Gnu compilers version 11, for LMDZ configured with rad=ecrad, parallel=mpi_omp, compilation fails for radiation_cloud_generator.f90, with message:

  554 |     !$omp declare simd(sample_from_pdf_simd) uniform(this) &
  Error: Symbol ‘sample_from_pdf_simd’ at (1) has already been
  host associated fcm_internal compile failed (256)

This version will compile and execute a LMDZ-DYNAMICO configuration provided:
- the ICOSA-LMDZ interface is downloaded from the LMDZ svn server rather than the HEAT server
- modifications are made to the arch.path modipsl configuration (in ARCH/) to include linking details to the NetCDF95 library

March 21st, 2023: modipsl.20230321.trunk.tar.gz, svn revision 4471

First version of testing using a full battery of tests.

Issues:

Interface between DYNAMICO and LMDZ needs to be changed due to a modification in the tracers parser.