期刊文献+

GPU异构平台上的第一性原理计算 被引量:1

First Principle Calculation on Multi-GPU Machines
原文传递
导出
摘要 平面波赝势密度泛函方法是材料科学计算中最常用的方法,但是其CPU代码计算时间长并且扩展性差,限制了其在超级计算机上的绝对速度。本文综述了使用图形处理器加速第一性原理计算软件的发展,讨论了VASP(Vienna Ab-initio Simulation Package),Quantum Espresso,PEtot使用的不同的加速策略,提出了众核平台上平面波赝势方法并行机制的改进,最后展望了使用图形处理器求解第一性原理计算的挑战和前景。 Plane Wave Pseudo-potential Density Functional Theory (PWP-DFT) calculation is the most important in scientific simulations. However, its absolute speed is limited by the inscalability of CPU PWP-DFT codes. This paper presents the development of first principle calculation on GPU platform and discusses different speedup strategies that applied in VASP, Quantum Espresso and PEtot. Based on the analysis of the new algorithm and parallel scheme, the paper gives the challenges and prospect of the ab initio calculation on many-core systems.
出处 《科研信息化技术与应用》 2012年第5期69-75,共7页 E-science Technology & Application
关键词 第一性原理计算 GPU 量子力学 电子结构 密度泛函理论 平面波赝势 First principle calculation GPU Quantum mechanics Electronic structure Density functional theory Plane wave pseudo-potential
  • 相关文献

参考文献20

  • 1Carter, E.A.. 2008, Challenges in Modeling Materials Properties Without Experimental Input, Science, 321(Aug.2008), 800-803.
  • 2Rapaport, D.C..The Art of Molecular Dynamics Simula- tion (Cambridge University Press, 2004, 2nd Ed).
  • 3Wang, L.W.. 2005, A survey of codes and algorithms used in NERSC material science allocations, LBNL technical report, LBNL-61051. (http://escholarship.org/uc/ item/9kh2q0cd).
  • 4VASP: http://cms.mpi.univie.ac.at/vasp/.
  • 5Qbox: http://eslab.ucdavis.edu/software/qbox/http://eslab. ucdavis.edu/software/qbox/index.htm.
  • 6Wang, L., Jia, W.L., Chi, X.B., Wu,Y., Gao W.G., Wang, L.W.. 2011 Large scale plane wave pseudopotential density function theory calculations on GPU clusters. The proceeding of SC 11 (ACM, 2011).
  • 7http://www.nvidia.com/content/PDF/fermi_whitegapers/ NVIDIA Ferrni_Compute_Architecture_Whitepaper.pdf.
  • 8CUBLAS: http://developer.nvidia.com/eublasCUFFT http://developer.nvidia.com/eufft CUDPP: http://gpgpu org/developer/cudpp.
  • 9MAGMA: http://icl.cs.utk.edu/magma/.
  • 10CUDA: http://en.wikipedia.org/wiki/CUDA.

引证文献1

二级引证文献2

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