Juan 14/04/2022:MNH/gradient_m.f90/shuman_device.f90, Bypass Cray GPU  bug/opti: use present_cr + do concurrent + acc_nv loop where needed

Juan 14/04/2022:MNH/turb*.f90, Bypass Cray GPU  bug/opti: use present_cr + do concurrent + acc_nv loop where needed

Juan 14/04/2022:MNH/get_halo.f90, Bypass Cray GPU bug , For GPU direct ,acc declare PSRC not working -> replace acc data present(PSRC)

Juan 11/04/2022:dyn_sources.f90,Optimization, put call MPPDB_CHECK under if MPPDB_INITIALIZED -> 25% cpu time save !!!

Juan 08/04/2022:rain_ice_fast_rg.f90, Bug Nvhpc22.2 seg-fault, array syntax -> do conc & some acc -> acc_nv for Cray Opt

Philippe 28/03/2022: OpenACC: use MNH_MEM_GET family calls in several subroutines + necessary modifications to keep performance and successful compilation with NVHPC (22.2)

Juan 25/03/2022:MNH/mode_mnh_zwork.f90, Cray Gpu Opt: add kernel present , make array-syntax serial -> parallel !!!

Juan 24/03/2022:MNH , Bug nvhpc/22.2 doesnt accepte kernel present added for Cray -> add #ifdef MNH_COMPILER_CCE

Juan 15/03/2022:tools.f90, Cray OpenACC PB , for atomic to work ,replace acc kernels -> acc parallel

Juan 03/03/2022:ice_adjust.f90, nvhpc22.2 optimisation -> replace WHERE+ARRAY SYNTAX -> DO CONCURRENT

Juan 23/02/2022:condensation.f90, Cray Optimisation, convert temporair scalar variable to Pointer + DO CONCURRENT  / ARRAY SYNTAX to avoid serialization on GPU

Juan 23/02/2022:tools.f90, convert to "modi" to avoid very long compilation time when change only implementation

Juan 10/02/2022: OpenACC: workaround for Cray CCE compiler: add an intermediate variable for VSIGQSAT

Naima + Philippe 10/02/2022: move data present directives for optional dummy arguments inside if-present protected regions

Naima 10/02/2022: OpenACC: bugfix: update XRUS, XRVS and XRWS and not copy them because already present

Naima + Philippe 09/02/2022: OpenACC: workaround CCE compiler problem: problems with character arrays in kernels regions

Philippe 08/02/2022: OpenACC: replace MEM_GET_ZT3D family calls by MNH_MEM_ALLOCATE family calls (not yet finished)

Philippe 01/02/2022: OpenACC: use MNH_ALLOCATE_FLAT and MNH_RELEASE_FLAT instead of old allocate/release procedures