A GPU-Accelerated Filtered Density Function Simulator of Turbulent Reacting Flows

Abstract

A new computational methodology is developed for large eddy simulation (LES) of turbulent reacting flows using graphic processing units (GPUs). The LES is based on the filtered density function (FDF) of the scalar-composition in conjunction with a discontinuous Galerkin (DG) discretisation scheme on a structured rectangular mesh. This hybrid solver is developed in a manner suitable for GPU computing. The simulator, as devised, is shown to be of the order of 200 times faster than the serial CPU-based calculations; facilitating the use of FDF for practical applications. The consistency and the accuracy of the methodology are demonstrated by simulations of a temporally developing mixing layer, under both non-reacting and reacting conditions.