The code used in the first example was compiled using the CCPP static build option, which has the best timing/memory performance. However, this means that you will need to recompile to run a different suite.
In this exercise, you will compile and run the code using the GFDL microphysics instead of the Zhao-Carr microphysics used in the first example.
Compile
Repeat the steps in the Compile section of the first example, this time using the following SDF:
suite_FV3_GFS_2017_updated_gfdlmp.xml
- For bash shell
./compile.sh $PWD/../FV3 theia.intel "CCPP=Y HYBRID=N STATIC=Y SUITE=suite_FV3_GFS_2017_updated_gfdlmp.xml" ccpp2 YES NO 2>&1 | tee compile2.log
- For t/csh shell
./compile.sh $PWD/../FV3 theia.intel "CCPP=Y HYBRID=N STATIC=Y SUITE=suite_FV3_GFS_2017_updated_gfdlmp.xml" ccpp2 YES NO |& tee compile2.log
Compare this SDF with the one used in the previous exercise - you will see a change in the physics parameterizations called in the suite.
Run
Copy a new run directory, fv3_ccpp_gfdlmp, and follow the same steps.
cp -r /scratch4/BMC/gmtb/emc_training_march_2019/handson_day1/fv3_ccpp_gfdlmp .
cd fv3_ccpp_gfdlmp
Edit the batch script to use your own project account: run_ccpp.sh
Link (or copy) your executable into this run directory:
ln -s ${YOUR_WORK_DIR}/NEMSfv3gfs/tests/fv3_ccpp2.exe fv3.exe
Submit the run script to the batch queue:
qsub run_ccpp.sh
NOTE: the only difference between these two cases, are the SDF and the namelist input.nml.
