文档下载 文档下载导航
设为首页 | 加入收藏
搜索 请输入内容:  
 导航当前位置: 文档下载 > 所有分类 > Scalable Parallel Sparse Factorization with Left-Right Looking Strategy on Shared Memory Mu
侵权投诉

Scalable Parallel Sparse Factorization with Left-Right Looking Strategy on Shared Memory Mu

An efficient sparse LU factorization algorithm on popular shared memory multiprocessors is presented. Interprocess communication is critically important on these architectures- the algorithm introduces O(n) synchronization events only. No global barrier is

Scalable Parallel Sparse Factorization with Left-Right Looking Strategy on Shared Memory Multiprocessors Olaf Schenk? 1, Klaus Gartner2, and Wolfgang Fichtner1 1

Integrated Systems Laboratory, Swiss Federal Institute of Technology Zurich, ETH Zurich, 8092 Zurich, Switzerland

foschenk, fwg@iis.ee.ethz.ch doc.guandang.net gaertner@wias-berlin.de

2

Weierstrass Institute for Applied Analysis and Stochastics, Mohrenstr. 39, 10117 Berlin, Germany

Abstract. An e cient sparse LU factorization algorithm on popular

shared memory multiprocessors is presented. Interprocess communication is critically important on these architectures - the algorithm introduces O(n) synchronization events only. No global barrier is used and a completely asynchronous scheduling scheme is one central point of the implementation. The algorithm aims at optimizing the single node performance and minimizing the communication overhead. It has been successfully tested on SUN Enterprise, DEC AlphaServer, SGI Origin 2000, Cray T90, J90, and NEC SX-4 parallel computers, delivering up to 2.3 GFlop/s on an eight processor DEC AlphaServer for medium-size semiconductor device simulations and structural engineering problems.

1 Introduction: Semiconductor Device Simulation and Algorithms Numerical semiconductor device and process simulation is based on the solution of a coupled system of non-linear partial di erential equations (PDE's) that can be either stationary, time dep

endent, or even complex depending on the problem considered. The nonlinear nature of the semiconductor transport equations, with exponential relations between variables, leads to discretized equations and the sparse linear systems are typically ill-conditioned. These systems are solved by iterative methods or direct methods. Since robust transient 2-D and 3-D simulations necessitate large computing resources, the choice of architectures, algorithms and their implementations becomes of utmost importance. Sparse direct methods are the most robust methods ?

The work of O. Schenk was supported by a grant from the Cray Research and Development Grant Program and the Swiss Commission of Technology and Innovation under contract number 3975.1.

上一页第2页

你可能喜欢

热门文档

相关文档

站点地图 | 文档上传 | 侵权投诉 | 手机版
本站所有资源均来自互联网,本站只负责收集和整理,均不承担任何法律责任,如有侵权等其它行为请联系我们.
文档下载 Copyright 2013 doc.guandang.net All Rights Reserved.  -  email