The Parallel Solving Optimization Of OpenFOAM Based On PETSc

In last decades,CFD simulation has won a continued attention in both industry and academia.It is cheap and efficient to solve simple physical and chemical problems with CFD.However,when it comes to some more complex problems,or the problems require higher accuracy,the CFD simulation often suffers from a large amount of calculation,which results in a terrible efficiency.Thus,it is practically significant to study how to improve the simulation efficiency of CFD.Based on the typical CFD software framework OpenFOAM,this paper launched a research of the key technologies of performance optimization in CFD simulation from the aspect of numerical solution and matrix assembly.First of all,we have improved the OpenFOAM framework based on PETSc.A newly designed framework and simulation process are given in this paper after analyzing the characteristics of both the OpenFOAM and PETSc framework.By introducing a third-party package,the numerical solve module of OpenFOAM has been expended and improved,which has a practical significance on improving the overall simulation efficiency of OpenFOAM.Then,we have analyzed and optimized the numerical solution process based on the improved OpenFOAM framework.Take the characteristics of the matrix format of FVM discrete into account,this paper focuses on the selection and configuration of the numerical algorithms to analyze and optimize the numerical solution process.Experimental results show that the performance of OpenFOAM has been significantly improved.Finally,we have also analyzed and optimized the matrix assembly process based on the improved OpenFOAM.In this paper,the matrix assembly process of OpenFOAM framework has been improved and optimized utilizing the symmetry of matrices and the idea of memcpy.Experimental results show that these two optimization approaches can accelerate the matrix assembly process to some extent.