查看文章

The present work gives a contribution to the investigation of optimization strategies for a parallel CFD solver for compressible flows. The focus of this study is the analysis and improved performance of an existing scientific application for CFD simulations by means of reducing MPI communication time. This is achieved by redesigning domain decomposition and communication patterns. Tests were carried out on multiple hardware configurations with various problem sizes. Results are illustrated by comparison with the original application, showing good improvement in performance when suitable optimization strategies for specific simulations are applied. The study concludes with a methodology for choosing the appropriate optimization strategy depending on the available hardware and the problem size.