Required modules on JURECA:

    module load Intel ParaStationMPI SciPy-Stack petsc4py mpi4py FFTW
    module load Scalasca CubeLib CubeGUI Vampir PAPI
    module unload Score-P


Build bindings (unless fixed in the official release/installed version):

    cd $PROJECT/tools/scorep_binding_python-master
    export PATH=/p/project/ccstma/scorep/6.0-trunk-mrobefix_intel-impi-papi/bin:$PATH
    pip install --user .


In your jobscript:

export PATH=/p/project/ccstma/scorep/6.0-trunk-mrobefix_intel-parastation-papi/bin:$PATH
export SCOREP_PROFILING_MAX_CALLPATH_DEPTH=90
export SCOREP_ENABLE_TRACING=1
export SCOREP_TOTAL_MEMORY=256M
export SCOREP_EXPERIMENT_DIRECTORY=data/mydir

srun python -m scorep --mpi run_simple_forcing_benchmark.py -n N
srun scout.mpi --time-correct $SCOREP_EXPERIMENT_DIRECTORY/traces.otf2


Workflow:
* Activate scorep-controller in the main Python file
* Run job without tracing, only do profiling (export SCOREP_ENABLE_TRACING=0)
* scorep-score -c N $SCOREP_EXPERIMENT_DIRECTORY/profile.cubex
* Check memory requirements, adapt in job script (export SCOREP_TOTAL_MEMORY=XXX) and restart job with tracing (export SCOREP_ENABLE_TRACING=1)
* square $SCOREP_EXPERIMENT_DIRECTORY

