Home About us Contact
|
Home About us Contact | ||
|
|
||
Linear Solver (linear + solver)
Selected AbstractsA parallel multigrid solver for high-frequency electromagnetic field analyses with small-scale PC clusterELECTRONICS & COMMUNICATIONS IN JAPAN, Issue 9 2008Kuniaki Yosui Abstract Finite element analyses of electromagnetic fields are commonly used for designing various electronic devices. The scale of the analyses becomes larger and larger, therefore, a fast linear solver is needed to solve linear equations arising from the finite element method. Since a multigrid solver is the fastest linear solver for these problems, parallelization of a multigrid solver is quite a useful approach. From the viewpoint of industrial applications, an effective usage of a small-scale PC cluster is important due to initial cost for introducing parallel computers. In this paper, a distributed parallel multigrid solver for a small-scale PC cluster is developed. In high-frequency electromagnetic analyses, a special block Gauss, Seidel smoother is used for the multigrid solver instead of general smoothers such as a Gauss, Seidel or Jacobi smoother in order to improve the convergence rate. The block multicolor ordering technique is applied to parallelize the smoother. A numerical example shows that a 3.7-fold speed-up in computational time and a 3.0-fold increase in the scale of the analysis were attained when the number of CPUs was increased from one to five. © 2009 Wiley Periodicals, Inc. Electron Comm Jpn, 91(9): 28, 36, 2008; Published online in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/ecj.10160 [source] Large-scale topology optimization using preconditioned Krylov subspace methods with recyclingINTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING, Issue 12 2007Shun Wang Abstract The computational bottleneck of topology optimization is the solution of a large number of linear systems arising in the finite element analysis. We propose fast iterative solvers for large three-dimensional topology optimization problems to address this problem. Since the linear systems in the sequence of optimization steps change slowly from one step to the next, we can significantly reduce the number of iterations and the runtime of the linear solver by recycling selected search spaces from previous linear systems. In addition, we introduce a MINRES (minimum residual method) version with recycling (and a short-term recurrence) to make recycling more efficient for symmetric problems. Furthermore, we discuss preconditioning to ensure fast convergence. We show that a proper rescaling of the linear systems reduces the huge condition numbers that typically occur in topology optimization to roughly those arising for a problem with constant density. We demonstrate the effectiveness of our solvers by solving a topology optimization problem with more than a million unknowns on a fast PC. Copyright © 2006 John Wiley & Sons, Ltd. [source] Parallel coarse-grid selectionNUMERICAL LINEAR ALGEBRA WITH APPLICATIONS, Issue 8 2007David M. Alber Abstract Algebraic multigrid (AMG) is a powerful linear solver with attractive parallel properties. A parallel AMG method depends on efficient, parallel implementations of the coarse-grid selection algorithms and the restriction and prolongation operator construction algorithms. In the effort to effectively and quickly select the coarse grid, a number of parallel coarsening algorithms have been developed. This paper examines the behaviour of these algorithms in depth by studying the results of several numerical experiments. In addition, new parallel coarse-grid selection algorithms are introduced and tested. Copyright © 2007 John Wiley & Sons, Ltd. [source] Algebraic multilevel preconditioning of finite element matrices using local Schur complementsNUMERICAL LINEAR ALGEBRA WITH APPLICATIONS, Issue 1 2006J. K. Kraus Abstract We consider an algebraic multilevel preconditioning technique for SPD matrices arising from finite element discretization of elliptic PDEs. In particular, we address the case of non-M matrices. The method is based on element agglomeration and assumes access to the individual element matrices. The left upper block of the considered multiplicative two-level preconditioner is approximated using incomplete factorization techniques. The coarse-grid element matrices are simply Schur complements computed from local neighbourhood matrices, i.e. small collections of element matrices. Assembling these local Schur complements results in a global Schur complement approximation that can be analysed by regarding (local) macro elements. These components, when combined in the framework of an algebraic multilevel iteration, yield a robust and efficient linear solver. The presented numerical experiments include also the Lamé differential equation for the displacements in the two-dimensional plane-stress elasticity problem. Copyright © 2005 John Wiley & Sons, Ltd. [source] Multigrid approaches to non-linear diffusion problems on unstructured meshesNUMERICAL LINEAR ALGEBRA WITH APPLICATIONS, Issue 8 2001Dimitri J. Mavriplis Abstract The efficiency of three multigrid methods for solving highly non-linear diffusion problems on two-dimensional unstructured meshes is examined. The three multigrid methods differ mainly in the manner in which the non-linearities of the governing equations are handled. These comprise a non-linear full approximation storage (FAS) multigrid method which is used to solve the non-linear equations directly, a linear multigrid method which is used to solve the linear system arising from a Newton linearization of the non-linear system, and a hybrid scheme which is based on a non-linear FAS multigrid scheme, but employs a linear solver on each level as a smoother. Results indicate that, in the asymptotic convergence region, all methods are equally effective at converging the non-linear residual in a given number of multigrid cycles, but that the linear solver is more efficient in cpu time due to the lower cost of linear versus non-linear grid sweeps. Copyright © 2001 John Wiley & Sons, Ltd. [source] Switching thresholds in MTJ using SPICE model , Effects of spin and Ampere torquesPHYSICA STATUS SOLIDI (A) APPLICATIONS AND MATERIALS SCIENCE, Issue 8 2008M. Malathi Abstract Spin torque due to spin polarized tunneling current can be used to switch the free layer in a magnetic tunnel junction (MTJ). This current also gives rise to an Ampere torque, which influences the switching threshold of the MTJ. We modified the Landau,Lifschitz,Gilbert equation (LLGE) to include an Ampere torque term and solved for the magnetization dynamics under the single domain approximation using a linear solver in SPICE. We also extend the model to a square array of MTJs to study the effect of nearest neighbour interactions in addition to effects like demagnetization and magnetostatic interactions with the pinned layer. The interlayer exchange field between the free and pinned layers of a MTJ and the spin torque are competing factors that decide the threshold current density for switching the MTJ. We used a two current model to study the effects of barrier height and barrier thickness on spin torque and exchange energy. We observe that both the spin torque and exchange energy decrease with an increase in barrier height (for ferromagnetic coupling) and barrier thickness. We find that the inclusion of Ampere torque causes a reduction in the switching current. Varying the thickness of MgO and Al2O3 barriers allows us to minimize the switching threshold voltage. (© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim) [source] Voronoi cell finite difference method for the diffusion operator on arbitrary unstructured gridsINTERNATIONAL JOURNAL FOR NUMERICAL METHODS IN ENGINEERING, Issue 1 2003N. SukumarArticle first published online: 11 MAR 200 Abstract Voronoi cells and the notion of natural neighbours are used to develop a finite difference method for the diffusion operator on arbitrary unstructured grids. Natural neighbours are based on the Voronoi diagram, which partitions space into closest-point regions. The Sibson and the Laplace (non-Sibsonian) interpolants which are based on natural neighbours have shown promise within a Galerkin framework for the solution of partial differential equations. In this paper, we focus on the Laplace interpolant with a two-fold objective: first, to unify the previous developments related to the Laplace interpolant and to indicate its ties to some well-known numerical methods; and secondly to propose a Voronoi cell finite difference scheme for the diffusion operator on arbitrary unstructured grids. A conservation law in integral form is discretized on Voronoi cells to derive a finite difference scheme for the diffusion operator on irregular grids. The proposed scheme can also be viewed as a point collocation technique. A detailed study on consistency is conducted, and the satisfaction of the discrete maximum principle (stability) is established. Owing to symmetry of the Laplace weight, a symmetric positive-definite stiffness matrix is realized which permits the use of efficient linear solvers. On a regular (rectangular or hexagonal) grid, the difference scheme reduces to the classical finite difference method. Numerical examples for the Poisson equation with Dirichlet boundary conditions are presented to demonstrate the accuracy and convergence of the finite difference scheme. Copyright © 2003 John Wiley & Sons, Ltd. [source] | ||||||||||