aeroFluidX - Background

Printer-friendly version

aeroFluidX ist based on Culises

Finite-Volume flow simulations involve setting up and solving large systems of equations. With the Culises library, FluiDyna GmbH already offers a solution for accelerating the latter: porting the solution procedure of the linear system to the GPU allows to reduce the time spent on this task.


More GPU-Power with AFX

aeroFluidX completes this approach by executing the matrix assembly on GPU as well – this way, the accelaration enabled by the use of the GPU is extended to the complete run, allowing overall speed-ups of 2-3x (GPU vs. full CPU socket).



Speed-up vs. OpenFOAM®.


How does aeroFluidX work?

aeroFluidX currently solves the steady-state, incompressible Navier-Stokes equations using a SIMPLE procedure. The solver is designed for working with unstructured 3D grids.

On-going development work covers the implementation of turbulence models (RANS), which will be incorporated in the form of the k-omega-SST model combined with wall functions.


Validation case: flow around NACA 0012 airfoil.

Work flow integration

aeroFluidX has an interface to the open source CFD platform OpenFOAM®, i.e., the solver can read mesh and run information in OpenFOAM® format and return solutions in OpenFOAM® format. This allows to integrate the solver into existing work flows, combining the accelerated CFD run with OpenFOAM® compatible mesh generators and a variety of post-processing tools (e.g., ParaView).


Images from:

Landmann, B., Wieczorek, K., Bachschuster, S.: GPU Acceleration of Computational Fluid Dynamics (CFD) in Industrial Applications using Culises and aeroFluidX. GPU Technology Conference, 2014, San Jose, CA, USA