An efficient communication strategy for massively parallel computation in CFD

With the development of high-performance computers, it is necessary to develop efficient parallel algorithms in field computational fluid dynamics (CFD). In this study, a novel communication strategy based on asynchronous and packaged proposed. The implements an aggregated process, which requires only one each iteration step, significantly reducing number communications. correctness convergence are demonstrated from both theoretical experimental perspectives. And real vehicle CHN-T model with 140 million meshes, detailed performance comparison analysis performed for traditional strategy, showing that has significant advantages terms scalability. Finally, strong scalability weak tests carried out separately model. scaling efficiency can reach 74% 10.5 billion meshes 256,000 cores. 90% 10 179,000 This research work laid important foundation fast design aircraft cutting-edge numerical methods.