Development of an effective high-performance computation for solving the Poisson equation

Abstract

The article discusses the development of efficient high-performance computing for solving the Poisson equation. The article proposes a parallel programming approach for the numerical solution of the two-dimensional Poisson equation using an explicit difference scheme for one computer. The partial differential equation has been replaced by its finite difference counterpart. As technology for parallelization was chosen from the software-hardware architecture (CUDA). Numerical experiments were carried out to test and compare the effectiveness of the proposed parallelization approach. The results compared with the results obtained based on simulations using central processing units (CPU) and on a graphical processing unit (GPU), which, based on a comparative analysis, showed the effectiveness of a parallel numerical algorithm. The performance of the algorithms was illustrated on the NVIDIA GTX 1050 architectures and with the sequential implementation of the Core (TM) i5-7500H.