mnh_expand expansion write acc loop independant collapse on its own
+ bugfix do concurrent order of indices

Quentin 27/06/2024: add kernels for sources_neg_correct. Simple merge with 55X-dev-openACC branch + add mnh_expand_where and kenrels for LIMA, KHKO, C2R2 cases (not tested)

Quentin 25/06/2024: add kernels for turb_hor_dyn_corr + apply addStack after shumanFUNCtoCALL for PHYEX_SHUMAN list

Quentin 29/03/2024: add MNH_MEM_GET interfaces for PHYEX and pyft application with gfortran and nvhpc. Remove the default maximum caracters by line to allow pyft transformation that may adds long line

Quentin 28/03/2024: fix optionnal arguments not allocated and add MERGE(key) condition to have the right size for this array
This is mandatory for applying --addMPPDB_CHECK pyft function

Juan 23/02/2023:modeln.f90 , misplaced TPBAKFILE=> NULL() , inside declaration statement for CRAY compiler

Juan 20/02/2024:modeln.f90 , Pb with WRITE_STATPROF_n TSTATIONS/TPROFILERS not initialized  -> add associated test

Juan 20/02/2024:dyn_sources.f90, Pb with Nvhpc compiler, expand some array syntaxe with "!$acc loop seq"

Quentin 05/02/2024: bf merge for merge  MNH-55X-dev-OPENACC-FFT (now compile on ifort, nvhpc and PACK_KTESTS ok on ifort)

Quentin 05/02/2023: bf for merge MNH-55X-dev-OPENACC-FFT into 56-branch : need of modifs of gradient_m for computation outside of physics

(cherry picked from commit b20d4c81)

(some missing data problems with OpenACC disappear maybe because less array copies)

Juan 11/12/2024:condsamp.f90, PB Bitrepro -> replace SUM be SUM_DD_R2_ll & average on all interiors pts for NFINDTOP = 0 ou 2

Philippe 11/12/2023: OpenACC: contrav: add condition for data present for some variables (available or not depending on the condition)

(cherry picked from commit 4795ceda)

(cherry picked from commit e83d72ab)

Juan 22/11/2023:modeln.f90, Bypass Bug with nvhpc/22.11 <-> cuMemcpyDtoHAsync returned error 1 => move ( NSV /= 0 ) test out of kernels

Juan 14/11/2023:init_mnh.f90, add initialisation of GPU halo with INIT_HALO_D , to avoid some potentiel null pointer a first GET_HALO_D call

Juan 14/11/2023: rain_ice_fast_ri.f90, Bug/data not present with CCE/16.0.1 around where -> add mnh_expand_where to bypass the bug (other where have already been transformed this way)

Juan 27/10/2023:prep_ideal_case.f90, for OPENACC add initialisation of MNH_ALLOC_ZWORK+INIT_HALO_D => PREP_IDEAL work with VER_USER=ZSOLVER version

Juan 24/10/2023:MNH/*.f90 , for cce/16.0.0 , add new present_cr directive to avoid FALSE RECURRENCE on GPU