GenSDF - Signed-Distance-Field Generator

MPI + Fortran based software to compute the signed-distance-field for OBJ file format geometry over cartesian grid. The code assumes regular grid along the streamwise and spanwise directions, while a non-uniform grid can be used in the vertical ($z$) direction.

How to compile the code

For small problems - serial CPU version of the code can be used

cd src/serial
make

For large problems - MPI version of the code can be used

cd src/mpi
make

How to change parameters

Follow the instructions in parameters.in to change the input parameters as per requirements

How to run the code

Edit parameters.in

# Run the serial code
./gensdf_serial

# Run the MPI code as
mpirun -np 8 ./gensdf_mpi

Example output for MPI based job. Note: The elapsed time on the progress bar is indicative and not actual as some MPI ranks may experience a higher computational load compared to the rank that prints the job. MPI rank nprocs/2 prints the progress bar.

  ░▒▓██████▓▒░░▒▓████████▓▒░▒▓███████▓▒░ ░▒▓███████▓▒░▒▓███████▓▒░░▒▓████████▓▒░ 
 ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░      ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░      ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░        
 ░▒▓█▓▒░      ░▒▓█▓▒░      ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░      ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░        
 ░▒▓█▓▒▒▓███▓▒░▒▓██████▓▒░ ░▒▓█▓▒░░▒▓█▓▒░░▒▓██████▓▒░░▒▓█▓▒░░▒▓█▓▒░▒▓██████▓▒░   
 ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░      ░▒▓█▓▒░░▒▓█▓▒░      ░▒▓█▓▒░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░        
 ░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░      ░▒▓█▓▒░░▒▓█▓▒░      ░▒▓█▓▒░▒▓█▓▒░░▒▓█▓▒░▒▓█▓▒░        
  ░▒▓██████▓▒░░▒▓████████▓▒░▒▓█▓▒░░▒▓█▓▒░▒▓███████▓▒░░▒▓███████▓▒░░▒▓█▓▒░        
  *** Starting with            8 MPI ranks ***
 *** Input file sucessfully read ***
 *** Successfully read the CaNS grid ***
 *** Sucessfully finished setting up the grid spacing ***
 Successfully read OBJ file: data/lucy.obj
 Number of vertices:        49987
 Number of normals:        49987
 Number of faces:        99970
 Geometry is bounded by (minimum)  0.32529841999999998       0.39974423999999997       0.20000000000000001     
 Geometry is bounded by (maximum)  0.67470158000000002       0.60025576000000003       0.80000000000000004     
 *** Min-Max Index-Value pair ***
 Min-Max x:         156  0.30468750000000000      |         355  0.69335937500000000     
 Min-Max y:         195  0.37988281250000000      |         317  0.61816406250000000     
 Min-Max z:          92  0.17871093750000000      |         420  0.81933593750000000     
 -- Finished pre-processing geometry in   0.27099799999999996      seconds...
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
 *** Calculating the signed-distance-field | u-faces ***
||||||||||||||||||||||100.00%    99970/   99970 Elapsed:     3.56s Remaining:     0.00s
 *** Writing output data to file ***
 -- Done with file write in   0.23250000000000082      seconds... | u-faces 
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
 *** Calculating the signed-distance-field | v-faces ***
||||||||||||||||||||||100.00%    99970/   99970 Elapsed:     3.57s Remaining:     0.00s
 *** Writing output data to file ***
 -- Done with file write in   0.23231200000000030      seconds... | v-faces 
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
 *** Calculating the signed-distance-field | w-faces ***
||||||||||||||||||||||100.00%    99970/   99970 Elapsed:     3.56s Remaining:     0.00s
 *** Writing output data to file ***
 -- Done with file write in   0.23273299999999963      seconds... | w-faces 
 - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 
 *** Calculating the signed-distance-field | Cell-Center ***
||||||||||||||||||||||100.00%    99970/   99970 Elapsed:     3.57s Remaining:     0.00s
 *** Writing output data to file ***
 -- Done with file write in   0.23209999999999908      seconds... | Cell-Center 
 *** Calculation for SDF completed in    23.869882000000000      seconds ***

Figure: Comparison of the calculated SDF around the Stanford Armadillo geometry scaled to a smaller size

How to visualise the results

1. Use python/numpy2vtk.py to convert the mask.bin binary file to VTK file to read the data in Paraview/VisIT or other post-processing software
2. Use python/readmask.py to directly read the mask.bin binary file and plot slices
3. Alternatively, you can also directly parse the binary file in Paraview or VisIT using the .xmf file format reader both for constant and variable grid spacing

Contributors

Akshay & Udhaya [https://github.com/udhaya-chandiran]