Juan 19/05/2021: add loop independent collapse , for all do conccurent & BR_ routine , or the do loop seq !!!

Juan 04/01/2021: rain_ice.f90 , for OpenACC , use ZT1D & array routines for allocate/manage 1D array

Juan 04/01/2021: mode_mnh_zwork.f90 , add new ZT1D & array routines for allocate/manage 1D array on rain_ice & openacc

Juan 04/01/2021: MNH/ini_rain_ice.f90, for OpenACC , add acc enter data create for X... module variable

Juan 17/11/2020: Rules.LXpgi2004.mk, for MANAGEDO2 compile spll_modd_halo_d.o without managed memory for real GPUDIRECT usage + minor modif

Juan 17/11/2020: Rules.LXnvhpc2005.mk, for MANAGEDO2 compile spll_modd_halo_d.o without managed memory for real GPUDIRECT usage + minor modif

Juan 17/11/2020: mesonh.f90 OpenACC , add acc init before mpi_init for finding correctly context in Cuda AWARE/GPUDIRECT mpi

Juan 16/10/2020: tridiag_thermo/tke/wind.f90 , OpenACC opti -> add loop seq + loop gang,  vector collapse(2)

Juan 16/10/2020: Makefile , pass option CPP=cpp , for netcdf compilation , for some problem with pgi compiler/configure

Juan 16/10/2020:modd_precision.f90 , non standard MPI_LOGICAL4/MPI_LOGICAL8  -> MPI_LOGICAL , to be done correctly for integer*8 compilation

Juan 18/09/2020: Opimisation OpenACC , BR_POW parallelized by adding temp 1D array ZLBEXI(:) = XLBEXI ,
DO CONCURRENT not bit-reporductible <-> NVHPC 20.7 Bug ?

Juan 14/09/2020: Rules.LXpgi2004.mk , add support of arch CPU type = TP & MANAGEDO2/OPENACCO2 optlevel <=> without "-g"

Juan 14/09/2020: Rules.LXnvhpc2005.mk , add support of arch CPU type = TP & MANAGEDO2/OPENACCO2 optlevel <=> without "-g"

Juan 14/09/2020: Rules.LXifort.mk , for reproductibility add MNH_BITREP compilation + -fp-model consistent

Juan 14/09/2020 : tridiag_wind.f90 , OpenACC optimisation -> DO CONCURRENT +  loop gang vector collapse