作者: Chunsheng Feng , Shi Shu , Jinchao Xu , Chen-Song Zhang
关键词:
摘要: The geometric multigrid method (GMG) is one of the most efficient solving techniques for discrete algebraic systems arising from elliptic partial differential equations. GMG utilizes a hierarchy grids or discretizations and reduces error at number frequencies simultaneously. Graphics processing units (GPUs) have recently burst onto scientific computing scene as technology that has yielded substantial performance energy-efficiency improvements. A central challenge in implementing on GPUs, though, computational work coarse levels cannot fully utilize capacity GPU. In this work, we perform numerical studies CPU–GPU heterogeneous computers. Furthermore, compare our implementation with an CPU popular fast Poisson solver, Fast Fourier Transform, cuFFT library developed by NVIDIA. AMS subject classifications: 65M10, 78A48