gpuFoam - Tools and extensions for GPU-accelerated OpenFOAM simulations

Overview

• gpu_chemistry - Gpu accelerated detailed chemistry model

For more details check the README section in each of the subfolders listed above.

Requirements

• C++17 supporting host compiler
• Either Nvidia cuda compiler (nvcc) version 10+
• Or AMD hip compiler (hipcc) version 5.6+
• Latest OpenFOAM foundation development release

Compilation

Everything in this folder should compile with the Allwmake script. Before running it, ensure that you have a valid OpenFOAM installation and compilers available.

• For Nvidia

  To check that you have valid compilers for the Nvidia backend

  foamRun -help
  g++ --version
  nvcc --version
  

  Then check the name of your graphics card (for example with nvidia-smi) on and go to this page to see which compute capability it has. Modify the file ./nvcpp accordingly if necessary.

  Finally, run the commands

  export GPUFOAM_BACKEND_NVIDIA=1
  ./Allwmake
  

• For AMD

  To check that you have valid compilers for the AMD backend

  foamRun -help
  g++ --version
  hipcc --version
  

  Then check the name of your graphics card and go to this page to see the matching architecture keyword. Modify the file ./hipcc if necessary.

  Finally, run the commands

  export GPUFOAM_BACKEND_AMD=1
  ./Allwmake
  

Credits

This project contains source code taken from the following projects

• mdspan
• variant

The licenses of the external projects can be found from their respective folders.

Citing

To cite gpuFoam use the below reference

@software{peltonen_gpufoam,
 author = {Peltonen, Petteri and Niemi, Timo and Celik, Hasan and Peltola, Juho},
 title = {gpuFoam: Tools and extensions for GPU-accelerated OpenFOAM simulations},
 organization = {VTT Technical Research Centre of Finland},
 url = {https://github.com/vttresearch/gpuFoam},
 year = {2024}
}