Testing codes

At minimum, please test your modified codes with Cameo basin testCase. Ideally please test compiling the modified code with two compilers - intel-oneapi (recent version of intel - ifx) and gcc. If intel compiler is not available, testing with gcc. This testCase uses small memory footprint so it should be run on a laptop by executing the compiled executable without mpirun.

We plan to provide global test case for those who access to larger computing resources (HPC or cluster) for MPI multi-cores run test.

There are two types of code changes - no-answer-change or and answer-change.

For the former, the developer must verify the simulations with the modified codes are bit-to-bit identical to the original code. The verification can be done using nccmp utility. The bit-to-bit comparison cannot be verified by just plotting like comparing time series plots. Note that using different optimization levels of compiler as well as different compilers generate bit-to-bit differences, so the consistent compilers and optimization flag should be for the original and modified codes.

For the latter, please document the expected changes in the results. For this, some plots like time series or spatial map would be useful.