1 | HOW TO DO REGRESSION TESTS OF THE FORTRAN CODE |
---|
2 | |
---|
3 | 1. Go to flex_extract/Testing/Regression/FortranEtadot |
---|
4 | 2. Download the tarball containing the input files and reference outputs |
---|
5 | 2. Untar the tarball |
---|
6 | 3. Create a wirking directory: mkdir Work |
---|
7 | 4. Compile the unmodified Fortran code with makefile_fast and makefile_debug |
---|
8 | (is in flex_extract/Source/Fortran) |
---|
9 | 5. Run a regression test to see whether the current Fortran code gives |
---|
10 | output consistent with the reference output. |
---|
11 | If not, carefully check why (machine-dependent small deviation?) |
---|
12 | The output from the regression run is in 'Outputs' (automatically created). |
---|
13 | If you need a new reference, you could remove or rename 'Outputs_ref', |
---|
14 | and then run |
---|
15 | ./mk_outputdirs.sh |
---|
16 | ./run_ref.sh |
---|
17 | to create a new reference version. |
---|
18 | 6. Work on the code and use the 'run_regrtest.sh' script to test your results. |
---|
19 | |
---|
20 | Note 1: The regression tests except those with "high" in their name will |
---|
21 | altogether run in about 1 minute. The "high" tests (hemispherical data) |
---|
22 | can take many minutes and also require up to ca. 20 GB of memory. |
---|
23 | Therefore, the script can be invoked as |
---|
24 | ./run_ref.sh omithigh |
---|
25 | to omit the "high" tests. For single development steps this should be |
---|
26 | sufficient. When you are satisfied, run the "high" tests at the end. |
---|
27 | Note 2: The test scripts contain |
---|
28 | export OMP_NUM_THREADS=4 # you may want to change this |
---|
29 | export OMP_PLACES=cores |
---|
30 | You should set OMP_NUM_THREADS to the number of physical cores of your |
---|
31 | test machine or less. |
---|
32 | OMP environment variables are explained on |
---|
33 | https://gcc.gnu.org/onlinedocs/libgomp/#toc-OpenMP-Environment-Variables |
---|