NLSEmagic: Nonlinear Schrödinger equation multi-dimensional Matlab-based GPU-accelerated integrators using compact high-order schemes
作者： R.M. Caplan
DOI: 10.1016/J.CPC.2012.12.010
关键词:
摘要: Abstract We present a simple to use, yet powerful code package called NLSEmagic numerically integrate the nonlinear Schrodinger equation in one, two, and three dimensions. is high-order finite-difference which utilizes graphic processing unit (GPU) parallel architectures. The codes running on GPU are many times faster than their serial counterparts, much cheaper run standard clusters. developed with usability portability mind, therefore written interface MATLAB utilizing custom GPU-enabled C MEX-compiler interface. packages freely distributed, including user manuals set-up files. Program summary title: Catalogue identifier: AEOJ_v1_0 URL: http://cpc.cs.qub.ac.uk/summaries/AEOJ_v1_0.html obtainable from: CPC Library, Queen’s University, Belfast, N. Ireland Licensing provisions: Standard licence, http://cpc.cs.qub.ac.uk/licence/licence.html No. of lines distributed program, test data, etc.: 124453 bytes 4728604 Distribution format: tar.gz Programming language: C, CUDA, MATLAB. Computer: PC, MAC. Operating system: Windows, MacOS, Linux. Has been vectorized or parallelized?: Yes. Number processors used: Single CPU, number dependent chosen card (max currently 3072 cores GeForce GTX 690). Supplementary material: Setup guide, Installation guide. RAM: Highly dimensionality grid size. For typical medium–large problem size dimensions, 4GB sufficient. Keywords: Nonlinear Schroodinger Equation, GPU, finite difference, Bose-Einstien condensates. Classification: 4.3, 7.7. Nature problem: Integrate solutions time-dependent one-, two-, three-dimensional cubic equation. Solution method: integrators utilize fully-explicit fourth-order Runge–Kutta scheme time both second- differencing space. NVIDIA GPUs interfaced built-in visualization analysis tools. Restrictions: main restriction for amount RAM as only designed single GPU. Unusual features: Ability visualize real-time simulations through interaction compiled integrators. Additional comments: guide provided. has dedicated web site at www.nlsemagic.com . Running time: A dimension 87×87×203 3360 steps (100 non-dimensional units) takes about one half minutes 580 card.
harvard.edu
arxiv.org
arxiv-vanity.com参考文章(38)
C. T. Kelley, Solving Nonlinear Equations with Newton's Method ,(1987)
Ricardo Carretero-González, Dimitri J. Frantzeskakis, Panayotis G. Kevrekidis, Emergent nonlinear phenomena in Bose-Einstein condensates : theory and experiment Springer. ,(2008)
John A. Stratton, Sam S. Stone, Wen-mei W. Hwu, MCUDA: An Efficient Implementation of CUDA Kernels for Multi-core CPUs languages and compilers for parallel computing. pp. 16- 30 ,(2008) , 10.1007/978-3-540-89740-8_2
Firas Hamze, Kamran Karimi, Neil G. Dickson, A Performance Comparison of CUDA and OpenCL arXiv: Performance. ,(2010)
Jeffrey S. Vetter, Richard Glassbrook, Jack Dongarra, Karsten Schwan, Bruce Loftis, Stephen McNally, Jeremy Meredith, James Rogers, Philip Roth, Kyle Spafford, Sudhakar Yalamanchili, Keeneland: Bringing Heterogeneous GPU Computing to the Computational Science Community Computing in Science and Engineering. ,vol. 13, pp. 90- 95 ,(2011) , 10.1109/MCSE.2011.83
Weizhong Dai, An unconditionally stable three-level explicit difference scheme for the Schro¨dinger equation with a variable coefficient SIAM Journal on Numerical Analysis. ,vol. 29, pp. 174- 181 ,(1992) , 10.1137/0729011
Stanley J. Farlow, Stephen F. Becker, Partial differential equations for scientists and engineers American Journal of Physics. ,vol. 53, pp. 702- 702 ,(1985) , 10.1119/1.14292
John E. Stone, David Gohara, Guochun Shi, OpenCL: A Parallel Programming Standard for Heterogeneous Computing Systems computational science and engineering. ,vol. 12, pp. 66- 73 ,(2010) , 10.1109/MCSE.2010.69
Douglass E. Post, Lawrence G. Votta, Computational Science Demands a New Paradigm Physics Today. ,vol. 58, pp. 35- 41 ,(2005) , 10.1063/1.1881898