Testing for effects of asymmetry and instability on preconditioned iterations of conjugate gradient type |
| |
Authors: | STMPSON R. B. |
| |
Affiliation: | Department of Computer Science, University of Waterloo Ontario Canada N2L 3G1 |
| |
Abstract: | ![]() We develop a parametrized family of matrices and use them totest the performance of some preconditioned iterative methodsas we vary the asymmetry and stability of the test matrices.The test matrices are based on a simple discretization of adynamic, two-species, contant coefficient, reaction-diffusionsystem of partial differential equations. The reaction coefficientsprovide natural parameters for varying the properties of thetest matrices, which are typical of modelling applications.These matrices are reducible via a red-black ordering, and itis shown that the reduced matrices are M-matrices for a largerrange of parameters than the unreduced test matrices. The iterative methods tested are of conjugate gradient type,using incomplete factorization preconditioning. The componentsof the methods tested are: the acceleration technique (conjugategradient squared, stabilized biconjugate gradi ent, orthomin),the level of fill-in of the incomplete factorization preconditioner,the use of the reduced system, and the effect of time-step sizereduction (for dynamic simulations). The tests are carried Outby extensive sampling in regions of the parameter space. The results appear to confirm observations of other studiesusing diffusion-convection based tests, and, in particular,show that in these instances the performance of the methodsis essentially unaffected by asymmetry, but is strongly affectedby instability. |
| |
Keywords: | |
本文献已被 Oxford 等数据库收录! |
|