An odyssey into local refinement and multilevel preconditioning III: Implementation and numerical experiments

被引:20
作者
Aksoylu, B [1 ]
Bond, S
Holst, M
机构
[1] CALTECH, Dept Comp Sci, Pasadena, CA 91125 USA
[2] Univ Calif San Diego, Dept Chem & Biochem, La Jolla, CA 92093 USA
[3] Univ Calif San Diego, Dept Math, La Jolla, CA 92093 USA
关键词
finite element methods; local mesh refinement; multilevel preconditioning; BPX; red and red-green refinement; 2D and 3D; data structures;
D O I
10.1137/S1064827502407676
中图分类号
O29 [应用数学];
学科分类号
070104 ;
摘要
In this paper, we examine a number of additive and multiplicative multilevel iterative methods and preconditioners in the setting of two-dimensional local mesh refinement. While standard multilevel methods are effective for uniform refinement-based discretizations of elliptic equations, they tend to be less effective for algebraic systems, which arise from discretizations on locally refined meshes, losing their optimal behavior in both storage and computational complexity. Our primary focus here is on Bramble, Pasciak, and Xu (BPX)-style additive and multiplicative multilevel preconditioners, and on various stabilizations of the additive and multiplicative hierarchical basis (HB) method, and their use in the local mesh refinement setting. In parts I and II of this trilogy, it was shown that both BPX and wavelet stabilizations of HB have uniformly bounded condition numbers on several classes of locally refined two- and three-dimensional meshes based on fairly standard (and easily implementable) red and red-green mesh refinement algorithms. In this third part of the trilogy, we describe in detail the implementation of these types of algorithms, including detailed discussions of the data structures and traversal algorithms we employ for obtaining optimal storage and computational complexity in our implementations. We show how each of the algorithms can be implemented using standard data types, available in languages such as C and FORTRAN, so that the resulting algorithms have optimal ( linear) storage requirements, and so that the resulting multilevel method or preconditioner can be applied with optimal ( linear) computational costs. We have successfully used these data structure ideas for both MATLAB and C implementations using the FEtk, an open source finite element software package. We finish the paper with a sequence of numerical experiments illustrating the effectiveness of a number of BPX and stabilized HB variants for several examples requiring local refinement.
引用
收藏
页码:478 / 498
页数:21
相关论文
共 21 条
[1]  
AKSOYLU B, UNPUB SIAM J NUMER A
[2]  
AKSOYLU B, 2001, THESIS U CALIFORNIA
[3]   THE HIERARCHICAL BASIS MULTIGRID METHOD [J].
BANK, RE ;
DUPONT, TF ;
YSERENTANT, H .
NUMERISCHE MATHEMATIK, 1988, 52 (04) :427-458
[4]  
BANK RE, 1980, CNA159 U TEX AUST CT
[5]   A BASIC NORM EQUIVALENCE FOR THE THEORY OF MULTILEVEL METHODS [J].
BORNEMANN, F ;
YSERENTANT, H .
NUMERISCHE MATHEMATIK, 1993, 64 (04) :455-476
[6]   NEW ESTIMATES FOR MULTILEVEL ALGORITHMS INCLUDING THE V-CYCLE [J].
BRAMBLE, JH ;
PASCIAK, JE .
MATHEMATICS OF COMPUTATION, 1993, 60 (202) :447-471
[7]  
BRAMBLE JH, 1990, MATH COMPUT, V55, P1, DOI 10.1090/S0025-5718-1990-1023042-6
[8]   MULTILEVEL PRECONDITIONING [J].
DAHMEN, W ;
KUNOTH, A .
NUMERISCHE MATHEMATIK, 1992, 63 (03) :315-344
[9]   A supernodal approach to sparse partial pivoting [J].
Demmel, JW ;
Eisenstat, SC ;
Gilbert, JR ;
Li, XYS ;
Liu, JWH .
SIAM JOURNAL ON MATRIX ANALYSIS AND APPLICATIONS, 1999, 20 (03) :720-755
[10]   SPARSE-MATRIX TEST PROBLEMS [J].
DUFF, IS ;
GRIMES, RG ;
LEWIS, JG .
ACM TRANSACTIONS ON MATHEMATICAL SOFTWARE, 1989, 15 (01) :1-14