静场Poisson方程的CUDA并行计算
Poisson Equation in Static Field Problems Based on CUDA
本文研究了静场问题中采用CUDA编程进行并行数值计算，即静场Poisson方程中用有限差分法形成的代数方程组，将求解方程组的Jacobi迭代计算移植到基于CUDA编程框架，在GPU中进行计算。提出了一种在Jacobi迭代计算中高效并行计算方式，最后通过数值实验，比较基于传统CPU和GPU两种不同计算体系下不同大小计算问题的计算效率和精度，得到CUDA编程的Jacobi迭代算法在GPU中运行具有非常高加速比。
In this paper, electrostatic field problems are parallel numerical computed by Compute Unified Device Architecture (CUDA) programming. That is, Jacobi iterative calculation for algebra equation which is derived from Poisson equation by finite difference numerical method is migrated to CUDA programming architecture and calculated by Graphics Processing Unit (GPU). Highly parallel implementation of the new parallel architecture in Jacobi iterative computation is propose and discuss in detail. At last, by the numerical experiment, the computation efficiency and accuracy of the CPU and GPU computing system are compared by the different dimension problems, and Jacobi iterative computation in GPU has highly acceleration ration.
赖生建、周振华
电工基础理论计算技术、计算机技术
电子科学与技术UDAJacobi迭代算法并行计算有限差分法
Electronic science and technologyCUDAJacobi iterative algorithmparallel computationfinite difference method
赖生建,周振华.静场Poisson方程的CUDA并行计算[EB/OL].(2011-03-25)[2025-08-03].http://www.paper.edu.cn/releasepaper/content/201103-973.点此复制
评论