LBM parallel optimization method based on multilayer grids

Abstract

The invention relates to a heterogeneous many-core-based LBM parallel optimization method. According to the inventive concept, the invention provides the LBM parallel optimization method for multi-layer grids, which comprises the following steps of: 1, changing a calculation process under the multi-layer grids according to LBM parallel operation characteristics; 2, adopting a two-dimensional division strategy for process task division, and meanwhile designing a corresponding communication strategy according to the division condition; and 3, designing the size of a communication buffer area according to LBM parallel operation characteristics and arranging required data for communication. The method has the specific advantages that the number of times of communication synchronization is reduced according to the characteristics of LBM parallel operation, so that the proportion of CPU vacancy is reduced; the two-dimensional division strategy is adopted in task process division, so that the computational domain boundary can be flexibly adjusted according to the calculated amount on the basis of ensuring correctness, and neat correspondence is not needed; correct calculation can be ensured only through partial data communication, and the communication data volume is reduced. The final effect is the improvement of the speed-up ratio and the speed-up efficiency.

Description

technical field [0001] The invention belongs to the field of computational fluid dynamics, and specifically relates to an LBM algorithm and parallel computing. Background technique [0002] With the application of CFD to different fields, the corresponding physical process and geometric model become more and more complex, and the grid scale of calculation becomes larger, which poses higher challenges to the computer. As a very important computational fluid dynamics method, the LBM method requires a large amount of computing resources, so the improvement of parallel performance is very important. [0003] At present, LBM parallel computing includes but is not limited to the following problems: [0004] (1) During the LBM parallel computing process, the greater the communication volume, the lower the parallel performance; [0005] (2) In the process of parallel computing, the synchronization of communication will cause some CPUs to be vacant, wasting computing resources. C...

AI Technical Summary

Problems solved by technology

[0004] (1) During the LBM parallel computing process, the greater the communication volume, the lower the parallel performance;

[0005] (2) In the process of parallel computing, the synchronization of communication will cause some CPUs to be vacant, wasting computing resources

Images