- Install virtual environment [See https://packaging.python.org/en/latest/guides/installing-using-pip-and-virtual-environments/]
python3 -m pip install --user virtualenv - Create the virtual environment for a specific python version .and. say yes to the prompt question
python3 -m venv <pathtosave>/stl2sdf - Install the required libraries
pip install -r requirements.txt
To deactivate the special python environment use [Note: dont do this if you wish to run the code...]
deactivate
- Running in parallel [using concurrent.futures]
python generateSDF.py 'input-file' 'output-array-location' nprocs
input-file: String with the location and name of the input file
output-array-location: String with the location to save the numpy arrays with SDF
<nprocs>: Number of processors
- Running in parallel [Truly MPI]
mpirun -np <nprocs> python -m mpi4py generateSDFmpi.py
The generateSDFmpi.py file is setup to run an example with a spehere in the domain. Additionally, the multi_sphere.stl example can also be used to understand the workflow. To run the example using concurrent futures style multithreading use the runall.sh script. To run the example using truly MPI based parallelisation use the runall_mpi.sh script.
options:
-h, --help show this help message and exit
--infile INFILE Input STL/OBJ file | Default 'assets/sphere.stl'
--filename FILENAME Target mesh folder | Default 'assets'
--nsamples NSAMPLES Number of sampling points used to compute the SDF | Default '1000000'
--writeData WRITEDATA Flag to write SDF to files | Default 'True'
--clipSDF CLIPSDF Flag to clip the SDF computation at zmax | Default 'False'.
You can find the documentation here.
NOTE: The code decomposes in the streamwise direction. Forced to use nprocs such that Nx%nprocs == 0
Check docs/scaling.md for full details of the scaling.
For a case with 2 Billion grid points, the expected peak memory usage is around 150 GB [approx.]. These tests were carried out a rather large stl file. We have observed that for stl files larger than 2 GB, the memory requirements can increase quite rapidly. Thus, we recommend using around 32 processors for these cases such that the total memory requirement stays bounded below 160 GB [relatively realistic limit!]
Signed distance field generated using a typical Large-Eddy-Simulation grid using the windAroundBuildings geometry supplied as part of the OpenFOAM tutorials.
While the code is setup to work consistently with CaNS [and staggered grid layout], porting the code to adapt to any other solver is straightforward. Follow the steps below:
- In
generateSDF.py, comment load grid portion of the code and load the x,y, and, z grid arrays as numpy arrays. This can be done a forced typecasting such as np.array(,dtype=np.float64). - Additionally, if you would like to output the SDF in a different format, you can edit the numpy2CaNSarray.py file. Beware: Python is not a strongly-typed language so you will need to force the precision of write!
We welcome contributions from everyone!
Making Changes: Please fork the repository, create a branch off of main, and submit a pull request with your changes for review.
Found a Bug or Error? If you discover any bugs in the code or errors in the documentation, open a new issue with as much information as possible, especially steps to reproduce the bug.
Questions or Suggestions: For any queries or ideas, don't hesitate to open an issue.
This code was written while listening to Leprous. Recommended albums Aphelion [2021], Coal [2013], and Bilateral [2011].