Computational Fluid Dynamics using OpenCL – a Practical Introduction

The main aim of the Computational Fluid Dynamics (CFD) simulations is to reconstruct the reality of fluid motion and behaviour as accurately as possible in order to better understand the natural phenomena under specified conditions. Ideally, general solutions can also cover different scales and geometric configurations. Unfortunately, due to expensive algorithms, classic CFD codes most often require long computational times to satisfy the convergence criteria. With the advent of high-performance GPUs with massively-parallel multi-threaded architectures, basic CFD algorithms can now be implemented to give results in near real-time. The current work will briefly review our existing explicit solver based on finite difference methods, the derivation and discretisation of the mathematical model and equations, through to GPU algorithm implementation. During presentation, several case studies computed using CSIRO's CPU/GPU supercomputer cluster will be described and compared against well known analytical and experimental solutions, i.e. natural convection, driven cavity, scaling analysis, magneto-thermal convection, etc.