Custom Query (986 matches)

This update includes several minor corrections that do not affect the computations:

• The syntax for the ampersand (&) location is now strict to facilitate alternate compiling using CMAKE, which will be used to interact with the JEDI data assimilation project. Many thanks to Dave Robertson for writing the PERL script to determine the ampersand location that triggers the free-form compiling rules. After using our fancy GMAKE and makefile in ROMS for years, we feel that are retrograding to the stone age with CMAKE. We are appalled and disgusted with its minimal capabilities.

• WARNING: Changed the metadata entry for obs_scale in varinfo.dat. Now, we have a precise long_name;

  'obs_scale'                                        ! Output
    'observation screening and quality control scale'
    'nondimensional'
    'nulvar'
    'datum'
    'idObsS'
    'nulvar'
    1.0d0
  

• Introduced new C-preprocessing option TRACING to track the sequence of ROMS routines calls during debugging. It will help track the location where the model is hanging-up in MPI applications in some computer environments.

• Corrected logic in the Coupled 4D-Var algorithms using the split scheme overwrites several output variables in the MODNAME NetCDF file. Notice that Coupled 4D-Var algorithms have not been released yet.

• Added additional global attributes to the MODNAME NetCDF file to facilitate post-processing:

  // global attributes:
                  :type = "ROMS/TOMS 4D-Var output observation processing file" ;
                  :Algorithm = "RBL4DVAR" ;
                  :str_day = 13008. ;
                  :end_day = 13012. ;
                  :str_date = "2004-01-03 00:00:00.00" ;
                  :end_date = "2004-01-07 00:00:00.00" ;
  

  showing the 4D-Var cycle starting and ending times and dates.

To clear any confusion the modules ROMS_export_mod (Master/roms_export.F) and ROMS_import_mod (Master/roms_import.F) were removed from the repository. These modules were never used. It were intended for coupling many years ago.

Updated memory.F to avoid creating a temporary vector for the fifth argument in the call to mp_collect during Fortran execution. Now, we have:

#ifdef DISTRIBUTE
      IF (.not.allocated(Bwrk)) THEN
        allocate ( Bwrk(Ntiles+1) )
      END IF
      IF (.not.allocated(Dwrk)) THEN
        allocate ( Dwrk(Ntiles+1) )
      END IF
#endif
      ...

#ifdef DISTRIBUTE
      Bwrk=spv
      Dwrk=spv
      DO ng=1,Ngrids
        Bwrk(MyRank+1)=BmemMax(ng)*1.0E-6_r8        ! already in bytes
        Dwrk(MyRank+1)=megabytefac*Dmem(ng)
        CALL mp_collect (ng, iNLM, numthreads, spv, Bwrk)
        CALL mp_collect (ng, iNLM, numthreads, spv, Dwrk)
        Bsize(MyRank,ng)=Bwrk(MyRank+1)
        Dsize(MyRank,ng)=Dwrk(MyRank+1)
        Bwrk=spv
        Dwrk=spv
      END DO
#else

where Bwrk and Dwrk are temporary vectors.