期刊文献+

基于通用计算图形处理器的电磁场有限元计算加速方法探讨

Study on Acceleration Methods for Parallel Computation of Electromagnetic Finite Element Method Based on General Purpose Graphics Processing Units
原文传递
导出
摘要 利用电磁场数值方法计算集成电路互联线寄生参数提取是电子设计自动化中重要的课题,随着电路集成度的增加与特征尺寸的减小,更遇到规模与精度的双重挑战,进行全芯片级的快速提取成为必要。并行计算是最为直接的解决途径,而通用计算图形处理器因其高度线程并发性以及低成本低功耗得到迅速应用。本文基于电磁场有限元计算方法,一方面针对通用图形处理器的特点采用自适应不连续伽辽金有限元法,在提高计算精度的同时不显著提高问题的自由度且充分利用处理器计算能力,另一方面发掘稀疏矩阵计算的并行潜力,采用区域分解算法进行分块离散、对节点重排及分组便于矩阵划分、改进稀疏矩阵线性运算在图形处理器中的实现效率,最终实现基于消息传递接口(MPI)与计算统一设备架构下(CUDA)的大规模并行计算目标。 Utilizing numerical methods in electromagnetic fields for extraction of interconnects parasitic in IC(integrated circuit) is a crucial issue in EDA(electronic design automation). With increasing of circuit density and shrinking of feature size of IC, EDA is facing the dual challenges of scale problem and calculation accuracy even more, hence, the fast full-chip extraction becomes necessary. Parallel computing is a direct and effective solution to the problem. Furthermore, the general-purpose computing on graphics processing units are expected to instantly applied because GPU has the high density of concurrent threads with low direct cost and low power consumption. In this paper we commence the study on GPU-accelerated extraction of interconnect parasitic based on CUDA and MPI to achieve a scalable high-performance large- scale parallel computation in FEM(finite element method) in electromagnetic field. There are two main aspects of the study: 1) Improving adaptive nodal discontinuous Galerkin finite element method(DG-FEM) which is appropriate to computation characteristics of GPU. DG-FEM combines the merit of both h-type and p-type adaptive approaches in FEM, which leads to higher calculation accuracy without remarkable increasing of problems' degree of freedom and makes considerable use of GPU's processing capacity. 2) Parallelism of solving of sparse matrix involved in the implementation of FEM is taken fully into account. Domain decomposition method is used for blocking/partitioning meshing, graph method is used to permute nodes, and linear algebra programs are revised for sparse matrix calculation in GPU.
出处 《科研信息化技术与应用》 2010年第4期29-36,共8页 E-science Technology & Application
基金 自然科学基金项目(50977085 随机有限元法在集成电路互连线寄生参数提取中的应用研究)的资助
关键词 电磁场并行计算 图形处理器(GPU) 统一计算设备架构(CUDA) 集成电路参数提取 不连续伽辽金有 Electromagnetic field parallel computation Graphics processing unit(GPU) Compute unified device architecture(CUDA) Integrated circuit parameter extraction Discontinuous Galerkin finite element method(DG-FEM
  • 相关文献

参考文献10

  • 1A. Kl?ckner,T. Warburton,J. Bridge,J.S. Hesthaven.Nodal discontinuous Galerkin methods on graphics processors[J].Journal of Computational Physics.2009(21)
  • 2Samuel Williams,Leonid Oliker,Richard Vuduc,John Shalf,Katherine Yelick,James Demmel.Optimization of sparse matrix–vector multiplication on emerging multicore platforms[J].Parallel Computing.2008(3)
  • 3J. Hesthaven,T. Warburton.Nodal Discontinuous Galerkin Methods[].Journal of Women s Health.2008
  • 4Cristopher Cecka,Adrian Lew,Eric Darve.Introduction to assembly of finite element methods on graphics processors[].IOP Conference Series: Materials Science and Engineering.2010
  • 5Bell, N,M Garland.Implementing sparse matrix-vector multiplication on throughput-oriented processors[].Proceedings of the Conference on High Performance Computing Networking Storage and Analysis.2009
  • 6M. M. Baskaran,R. Bordawekar.Optimizing sparse matrix-vectormultiplication on GPUs[].IBM Research Report RC.2009
  • 7Loula Fezoui,St\’ephane Lanteri,St\’ephanie Lohrengel,Serge Piperno.Convergence and stability of a discontinuous Galerkin time-domain method for the 3D heterogeneous Maxwell equations on unstructured meshes[].ESAIM Math Model Numer Anal.2005
  • 8GPGPU.org.General-Purpose computation on Graphics Processing Units. http://gpgpu.org .
  • 9N.Godel,T.Warburton,M.Clemens.GPU accelerated discontinuous Galerkin FEM for electromagnetic radio frequency problems[].Antennas and Propagation Society International SymposiumIEEE.2009
  • 10.METIS[]..

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部