Check experiments results

Our openMP implementation is going to produce changes in the results. We need to ensure that these changes are "good" and not "bad", where with bad we mean changes that accumulate over time or results totally wrong.

Actually we expect just one of the parallelization to produce different results, since in the other two the operations are always done in the same order as in the original code.

The protocol decided for this is the following:

• Run the unmodified version of NEMO4.2 for 10 Years (58400 ts) with ORCA1 configuration
• Using produced restart files run the original NEMO for other 5 years (I think it is enough)
• Run the OpenMP parallelization for the same duration
• Compare the results just subtracting the two NetCDF files produced buy the different simulation: this will produce a nice simulation to show during the presentation.

For the other two parallelization we could do the same or show just some scalar variable result depending on how much time is left for this

Thing to keep in mind for output/restarts:

• In order to rebuild restart into a single file

    cd PATH/nemo/tools
    ./maketools -n REBUILD_NEMO -m mn4 # your arch file
    cd -
    cp ${sources}/tools/REBUILD_NEMO/rebuild_nemo* your_run_folder
    # Enter spinup folder and get restart names
    cd ${spinup_folder}
    restart_name=( $(ls ORCA*_????????_restart* ) )
    restart_name=${restart_name[0]/_0000.nc/}
    # Merge restarts
    ${rebuild_nemo} ${restart_name} ${spinup_cores}

• In the namrun namelist set the ln_rstart parameter to .true.
• Activate Heat and Salt content, i.e. to put

  !----------------------------------------------------------------------- &namhsb ! Heat and salt budgets (default: OFF) !----------------------------------------------------------------------- ln_diahsb = .true. ! output the heat and salt budgets (T) or not (F)