LIMA refactoring and cleaning

[SebastienRietteMTO]

Several modifications have been identified:

1. Setup:

   • setup doesn't work in simple precision (eg on NEC: "ABOR1 [PROC=1,THRD=1] : PA argument is too large or ITMAX is too small, the incomplete GAMMA_INC function cannot be evaluated correctly by the continuous fraction method")
   • setup doesn't work with NVFORTRAN (deactivated in lima and lima_adjust testprogs with OPENACC)

2. lima and lima_adjust common issues

   • IN/OUT/INOUT
   • use 2 dimensions instead of 3
   • use structures (instead of using values from MODD_NSV, MODD_PARAM_LIMA… for lima; NEBn, TURBn… to be passed through the interface for lima_adjust)
   • add DrHook
   • use explicit shapes in interfaces
   • conformity to Doctor norm (especially P for dummy args, Z for local variables)
   • code must be in upper cases
   • would it be possible to use interp_micro.func.h instead of ZVEC/IVEC?
   • ALLOCATABLE arrays must be removed
   • reactivate lima and lima_adjust calls with OPENACC in the testprogs

3. issues specific to lima

   • PEVAP must be initialised with 0
   • useless arguments NIFN, NCCN, NIMM must be suppressed

4. issues specific to lima_adjust

   • CFw is not reset to 0 after the control « remove cloud droplets if there are few ». Would it be a solution to call lima_compute_fractions here?
   • code still uses 3 time steps whereas the third time step has been removed from other parametrisations
   • there is an asymmetry between C et I (PCIT/PCIS vs PCCT/PCCS), is it OK?
   • NBUMOD and KMI must be suppressed
   • PSIGQSAT must be a 2D array