[Lukasz]

We have developed a library of differential operators operating on the whole mesh implemented fully on GPU. This way the user has a flexibility to build the solvers that are much faster! After the mesh is uploaded to the GPU memory no data are exchanged between CPU and GPU. Therefore, we were able to achieve up to x3 acceleration comparing to OpenFOAM running on Intel Xeon at full speed. See the report with the analysis of the laminar versions of SIMPLE and PISO solvers.

Now, we are working on the next milestone and need your assistance.

Could you please fill in this short questionnaire so that the OpenFOAM community could benefit from GPU acceleration?