Investigation into Computational Performance of a Multi-Modal Turbulent Combustion Model

Abstract

High Performance Computing is a rapidly evolving research field with a wide range of applications. Faster computers enable complex problems and equations to be solved within a reasonable time scale. This thesis applies High Performance Computing concepts to increase computational performance of a multi-modal turbulent combustion model. This thesis investigates different methods to improve performance including optimizing code within the model, retrofitting the model for offloading onto a Graphical Processing Unit (GPU), and exploring hybrid programming models. While the GPU implementation led to significantly slower execution times, CPU optimizations led to a roughly 40% speedup for a simple case and even higher speedups for more complex cases. Future work should apply the principles explored in this thesis to other models to improve computational performance.