Nano Science and Technology Institute - NSTI  
Nano Science and Technology Institute   Home | Subscribe | Site Map  
  ABOUT | COURSES | EVENTS | PUBLICATIONS | LEADERSHIP | OUTREACH | NEWS | PRESS | JOBS | Nanotechnology Solutions
px
px fade_top
Publications
Nanotech 2008 CDROM
Nanotech 2007 CDROM
Nanotech 2006 CDROM
Nanotech 2005 CDROM
Nanotech 2004 CDROM
3 CDROM Special Offer
Nanotech 2008 Vol. 1
Nanotech 2008 Vol. 2
Nanotech 2008 Vol. 3
Nanotech 2007 Vol. 1
Nanotech 2007 Vol. 2
Nanotech 2007 Vol. 3
Nanotech 2007 Vol. 4
Nanotech 2006 Vol. 1
Nanotech 2006 Vol. 2
Nanotech 2006 Vol. 3
Nanotech 2005 Vol. 1
Nanotech 2005 Vol. 2
Nanotech 2005 Vol. 3
WCM 2005
Nanotech 2004 Vol. 1
Nanotech 2004 Vol. 2
Nanotech 2004 Vol. 3
Nanotech 2003 Vol. 1
Nanotech 2003 Vol. 2
Nanotech 2003 Vol. 3
Nanotech 2002 Vol. 1
Nanotech 2002 Vol. 2
Nanotech 2001 Vol. 1
Nanotech 2001 Vol. 2
MSM 2000
MSM 99
MSM 98
Index of Authors
Index of Keywords
Index of Affiliations
Library Request Form
Shopping Cart
Order Form
 
Publications Publications
Nanotech 2004 Vol. 2
p
 
Technical Proceedings of the 2004 NSTI Nanotechnology Conference and Trade Show, Volume 2
Nanotech 2004 Vol. 2
Technical Proceedings of the 2004 NSTI Nanotechnology Conference and Trade Show, Volume 2
 
Chapter 8: Computational Methods and Numerics
 

Parallelization of BEM Electrostatic Solver Using a PC Cluster

Authors:H-K Tang and Y-J Yang
Affilation:National Taiwan University, TW
Pages:482 - 485
Keywords:pre-conditioner, PC cluster, BEM, fast multipole method, parallelization, MPI
Abstract:In this paper, the application of parallel-processing techniques for electrostatic calculation is presented. Evaluating the pre-conditioner used for inverting the system matrix of a boundary-element-method (BEM) electrostatic solver (Fastlap) is one of the most time consuming steps during the whole computation process. Therefore, we apply parallelization techniques to accelerate this step using a PC cluster, with minimum modification on the original source code. The basic idea is to parallelize the inversions of the sub-matrices which are extracted from the original system matrices. Then the inverted sub-matrices will be assembled into the pre-conditioner. The efficiency of the parallelization is also compared between different methods of distributing sub-matrices for inverting. The preliminary results for a testing sphere structure show that the computing time can be effectively reduced using the parallel processing without any compromise in accuracy. The speed-up is close to 4 by using a four-node cluster.
ISBN:0-9728422-8-4
Pages:519
Hardcopy:$150.00
 
Order:Mail/Fax Form
Special:3 CD Set — 15% off with Free Shipping
Up
nanoPRwire™
nanoPRwire
News Headlines
nano World news
 
 
 
 
px
© Nano Science and Technology Institute     About NSTI | Terms of Use | Privacy Policy | Contact