Install SU2 with docker

1. cd into the folder where Dockerfile is.

2. Create a docker ubuntu system image for installing SU2: docker build -t su2 . (-t to specify image name as su2, and . means using the Dockerfile in this folder)

3. Create a docker image for installing SU2 by executing ./create_su2_container.sh

4. Create a docker container for installing SU2 by executing ./start_su2.sh

5. Now, you should be in docker container, and the current path in docker container should correspond to test folder in your host system. Run ls, you can see two files: Miniconda3-latest-Linux-x86_64.sh and su2code-SU2-v7.3.1-0-g328a1b7.tar.gz.

6. Install Miniconda3 by bash Miniconda3-latest-Linux-x86_64.sh. Set yes to all install options.

7. Put export OMPI_MCA_btl_vader_single_copy_mechanism=none into .bashrc (See this issue)

8. Create a vistual environment for installing SU2: conda create -n su2 numpy python=3.7 (numpy is needed by parallel_computation.py)

9. Install mpi4py by python -m pip install mpi4py --user. Warning: DO NOT use conda install mpi4py. See this discussion

10. Uncompress su2code-SU2-v7.3.1-0-g328a1b7.tar.gz by tar zxvf su2code-SU2-v7.3.1-0-g328a1b7.tar.gz. Rename the code folder to su2code, and then cd su2code.

11. Important: change the static: true argument to static: false in externals/cgns/hdf5/meson.build. See this issue

12. Install SU2 following the guidance:

    Firstly, executing (remember change username to your own system)

    ./meson.py build -Denable-autodiff=true -Denable-directdiff=true -Denable-pywrapper=true -Dwith-mpi=enabled --prefix=/home/username/SU2
    

    Then put the suggested path (see the terminal outputs) into your .bashrc.

    Then executing

    ./ninja -C build install
    

SU2 should be installed successfully now.

Note: Consider using Tsinghua mirrors for anaconda and pip if you are in China.

For anaconda, see https://mirrors.tuna.tsinghua.edu.cn/help/anaconda/

For pip, see https://mirrors.tuna.tsinghua.edu.cn/help/pypi/

For the FSI tutorial, do some extra steps (make sure you are now in the created su2 python virtual environment)

13. Install blas and lapack libraries by sudo apt-get install -y libblas-dev liblapack-dev

14. Install Cython and scipy by python -m pip install Cython scipy

15. Uncompress petsc-with-docs-3.16.1.tar.gz, and rename it to petsc, then cd petsc

16. Config petsc by ./configure -with-petsc4py=1, then following the terminal outputs to install PETSc (make PETSC_DIR=/home/username/petsc PETSC_ARCH=arch-linux-c-debug all and then make PETSC_DIR=/home/username/petsc PETSC_ARCH=arch-linux-c-debug check)

17. Put the following settings into .bashrc (again, change username according to your own host system)

export PYTHONPATH=$PYTHONPATH:/home/username/petsc/arch-linux-c-debug/lib
export PETSC_DIR=/home/username/petsc
export PETSC_ARCH=arch-linux-c-debug
export PATH=$PATH:$PETSC_DIR/$PETSC_ARCH/bin

18. Install spatialindex:

tar zxvf spatialindex-src-1.9.2.tar.gz
mv spatialindex-src-1.9.2 spatialindex-src
cd spatialindex-src
cmake . # (do not forget the .)
make
sudo make install

19. Install rtree by python -m pip install rtree.

Things should work now.

Test the dynamic FSI python tutorial by mpirun -n 2 python -m mpi4py $SU2_RUN/fsi_computation.py --parallel -f fsi.cfg > log 2>&1 (remember put mesh into subfolder)