GPGPU implementation and benchmarking of the unsteady vortex lattice method

Abstract

The unsteady vortex lattice method is widely known for its robustness in modeling inviscid flow fields and aerodynamic loading. This method can be used to model a variety of configurations including multiple flapping wings in unsteady flows. One of the key computational algorithms used in this model can be expressed as nested loops or as a linear algebra based matrix vector product. Both forms are compared, and several different implementations including Matlab, C, and a GPGPU version are compared. The GPGPU based implementation is faster than the Matlab implementation by almost three orders of magnitude, and faster than the C implementation by almost two orders of magnitude. Details of the implementation are provided. The linear algebra implementation is determined to consume excess memory without a corresponding increase in computation performance. The fastest GPGPU implementation is applied to the study of ground effect, and separately, to the roll up of a vortex filament. Each of these cases would be intractable to compute without the acceleration offered by the GPGPU.

Publication
51th AIAA Aerospace Sciences Meeting