• C++17 compiler (gcc >= 7 or clang >= 5)
• CMake (version >= 3.11) or Bazel
• SuiteSparse
• Gurobi optimizer

Before installing SuiteSparse, make sure you have a blas and lapack implementation installed. For convenience a suitesparse installer is available in the scripts/ directory. Make sure you are in the root Stuka directory and run the following command.

./scripts/install_suitesparse

If the scripts fails due to an inability to find blas or lapack you may need to manually specify the location of your implementation by editing the scripts/install_suitesparse.sh file with the following

make library BLAS=${BLAS_LOCATION}

Stuka is built using the CMake build system. To create the library simply run the following commands from the root Stuka directory.

mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
cmake --build . --target stuka

This should create a shared library in the build/lib directory.

Building the python library is as simple running

python setup.py install

This should work assuming there are no issues in the previous step.

In order to test the code, a test script is available to be built. Assuming you have build the library as described in the Installation section, run the following.

cd build
cmake --build . --target tester

This will create a tester executable in the build/bin directory. Running the tester will run into an error, but this is expected. In the final example, Bender's Decomposition is unable to determine the optimal solution due to degeneracy of the problem.

This will be added at a later date.