Send mail to:             for the digests or bakeoff
            for comments or help
 Current editor:  Craig Douglas       
Anonymous ftp repository: (

World Wide Web: or

Today's editor:  Craig Douglas (

Volume 9, Number 3 (approximately March 31, 1999)

Today's topics:

     Important Dates
     Virtual Proceedings (Bonn and Copper Mountain)
     Paper by Beck
     Ph.D. Dissertation of Feng Wang
     The future of HPC - DIRECT Questionnaire 
     Algebraic Multigrid (AMG) Pre-print Available from GMD
     Copper Mountain Conference on Multigrid Methods Schedule


Date: Wed, 31 Mar 1999 11:23:22 +0500
From: Craig C. Douglas 
Subject: Important Dates

April 19    Title and abstracts (roughly 500 words) are due for the European
            Multigrid Conference (September 27-30, 1999 at the Universiteit
            Gent, Belgium).
            See for more details.

April 30    Title and abstracts (one page) are due for the International
            Workshop on Computational Physics (August 2-6, 1999 in Beijing).
            See for
            more details.


Date: Wed, 31 Mar 1999 11:24:22 +0500
From: Craig C. Douglas 
Subject: Virtual Proceedings (Bonn and Copper Mountain)

There are two virtual proceedings being worked on at the moment.  One is for
the Tenth GAMM Workshop on Parallel Multigrid Methods (held in Bonn, Germany
during October, 1998).  The second is for the Ninth Copper Mountain Conference
on Multigrid Methods (held during April, 1999).  Anyone who participated in
these conferences are welcome to post a paper for the respective virtual
proceedings.  This does not preclude your publishing your paper elsewhere.  If
you are interested, contact me by e-mail or see the directions in

The Copper pre-proceedings is receiving papers while I am working on this
issue.  You can see what is there by looking at the web page  Papers that come in this week will be
installed within 24 hours of receipt.


Date: Fri, 26 Mar 1999 16:51:04 -0500
From: "Thomas L. Beck" 
Subject: Paper by Beck

               Multigrid High Order Mesh Refinement Techniques
                for Composite Grid Electrostatics Calculations

                                Thomas L. Beck
                           Department of Chemistry
                           University of Cincinnati
                          Cincinnati, OH 45221-0172


A new method for performing high order mesh refinement multigrid computations
in real space is presented.  The method allows for accurate linear scaling
electrostatics calculations over composite domains with local nested fine

The Full Approximation Scheme (FAS) multigrid technique is utilized for a
sequence of refinement patches of increasing resolution.  Conservation forms
are generated on coarse scales by additional defect correction terms which
counter the local excess fluxes at the boundaries.  Formulas are given for
arbitrary order, extending the existing technique of Bai and Brandt.  Test
calculations are presented for a singular source in three dimensions which
illustrate the multigrid convergence properties, numerical accuracy, and
correct order of the approach.

The paper can be accessed at under the lind 'research'.
The files are or ce.pdf

Thomas L. Beck |
Department of Chemistry |
University of Cincinnati  | phone: 513-556-4886
Cincinnati, OH  45221-0172 | fax:  513-556-9239

    Editor's Note: See or access it at


Date: Mon, 8 Mar 1999 18:38:31 -0800 (PST)
Subject: Ph.D. Dissertation of Feng Wang

          Efficient Block Iterative Methods for Convection-Dominated
                Problems and Multigrid for Long-Thin Elements

                                  Feng Wang

                               Ph.D.; May 1998

                      The Pennsylvania State University

                          Thesis Advisor: Jinchao Xu

The purpose of this research is to study efficient iterative methods
(especially domain decomposition and multigrid methods) for convection-
dominated problems.

We present our approaches in two directions.  One is a specific type of block
iterative method called the ``cross-wind-block (CWB) iterative method''.  We
divide the computational domain into cross-wind blocks and perform the block
iterative algorithm.  A specific monotone scheme, the EAFE scheme, has been
adopted as the main discretization method in our computation and analysis.
Also, to effectively divide the domain into cross-wind blocks and order them
in the downwind direction, we apply graph theory to our mesh and the stiffness
matrix, and a special depth-first search algorithm, Tarjan's ordering
algorithm, is used.  The convergence of the CWB method is particularly good
for the problems with non-recirculating flows.  The solution for the monotone
schemes, though with only first order accuracy, may be applied to help
construct preconditioners or good initial guesses for higher order schemes,
such as streamline diffusion and SUPG schemes.

The other approach is to discretize the problem with long-thin elements at its
boundary layers.  This method is effective in resolving the spurious
oscillations usually occur near the boundary layers with the standard Galerkin
methods.  One application of this long-thin element method is to be used as a
preconditioner for spectral methods.

    Editor's Note: See or access it at


Date: Thu, 25 Mar 1999 15:56:43 +0000 (GMT)
From: Gavin J Pringle 
Subject: The future of HPC - DIRECT Questionnaire 


Which aspect of the SERVICE do you value most?
How will your NEEDS change over the next three to five years?
What is the ROLE of HPC Centres in Europe?

If questions like this concern you, we would like to invite
you to complete the following questionnaire:

One lucky respondent who will be chosen by lot will receive
their chosen bottle of malt whisky!

EPCC has prepared this questionnaire as a leading partner in
the DIRECT project, which is an EU-funded Concerted Action
in High Performance Computing and Networking, bringing
together fourteen Large-Scale HPC and Data Centres.

Help DIRECT reflect your opinions and future needs to the
decision-makers by taking a few minutes to fill in the form.

The closing date for entry to the prize draw is 30th April 1999.
However responses will also be accepted after this date and we
encourage you to reply  so that your views can be included in this

Also, we would be very grateful if you could encourage your colleagues
to take part in this survey.

We look forward to hearing from you soon.


Date: Sun, 28 Mar 1999 16:44:27 +0200
From: Klaus Stueben 
Subject: Algebraic Multigrid (AMG) Pre-print Available from GMD

Algebraic Multigrid (AMG):
An Introduction with Applications

Klaus Stueben

Schloss Birlinghoven		
D-53757 St. Augustin


For copyright reasons, we cannot put the report on the web for downloading.
However, pre-prints are available.  Just send an e-mail to "".


Since the early nineties, there has been a strongly increasing demand for more
efficient methods to solve large sparse and unstructured linear systems of
equations.  For practically relevant problem sizes, classical one-level
methods had already reached their limits and new hierarchical algorithms had
to be developed in order to allow an efficient solution of even larger
problems.  The purpose of this paper is to give an elementary introduction to
the first hierarchical and purely matrix-based approach, algebraic multigrid
(AMG).  The main idea behind AMG is to extend the classical ideas of geometric
multigrid (smoothing and coarse-grid correction) to certain classes of
algebraic systems of equations.  Besides its robustness and efficiency, the
main practical advantage of AMG is that it can directly be applied, for
instance, to solve various types of elliptic partial differential equations
discretized on unstructured meshes, both in 2D and 3D.  Since AMG does not
make use of any geometric information, it is a "plug-in" solver which can even
be applied to problems without any geometric background, provided that the
underlying matrix satisfies certain properties.  Although the development of
AMG goes back to the early eighties, it still provides one of the most
efficient algebraic methods to solve corresponding problems.  Compared to the
original approach, however, several modifications and extensions have been
introduced.  In addition to the theoretical basics of AMG, we present a
concrete algorithm in some detail and demonstrate its robustness and
efficiency by means of a variety of typical applications.

Klaus Stueben, GMD/SCAI, Schloss Birlinghoven		
D-53757 St. Augustin, Germany,
Tel: (+49) 2241-14-2749, Fax: (+49) 2241-14-2102


Date: Wed, 31 Mar 1999 11:25:22 +0500
From: Craig C. Douglas 
Subject: Copper Mountain Conference on Multigrid Methods Schedule

                      Ninth Copper Mountain Conference
                             Multigrid Methods
                             APRIL 11-16, 1999

                                                             April 1999
                                                       Su  M Tu  W Th  F Sa
                                                                    1  2  3
                                                        4  5  6  7  8  9  8
                                              ----->   11 12 13 14 15 16 17
                                                       18 19 20 21 22 23 24
                                                       25 26 27 28 29 30
          This program is preliminary and subject to modifications.
       Times given here are only estimates, and are likely to change.
     We will try to limit most schedule changes to within the given day.
Participants and speakers should monitor this page and the bulletin board at
                              the conference.
         Speakers should plan their talks to be at most 20 minutes.
           Program Chairs will adhere strictly to this schedule.
We will have a slide projector and an overhead projector for speaker use. We
  will also provide an In Focus Systems Lite Pro 620 multimedia projector,
 which plugs into standard video outlets on a Mac or PC to project its image
   onto the screen. This equipment will also be available for the special
                             evening sessions.
   Wednesday night's cash bar & banquet is in Victoria B in the Pavillon.
     All other activities will be held in the Village Square conference

                              Sunday, April 11

   TUTORIAL     Tutorial series on basic methods, algebraic multigrid, and
                                   parallel techniques.

 9:00am-noon  William L. Briggs A Multigrid Tutorial 

 1:30pm-3:30pmVan Emden Henson An Algebraic Multigrid Tutorial 

 4:00pm-5:00pmJim E. Jones A Parallel Multigrid Tutorial 

   RECEPTION                    Cash bar and light buffet.
 7:00pm-9:00pmVillage Square conference facilities.

                              Monday, April 12

     SESSION 1
 Algebraic Methods                  Chair: Van Emden Henson

 8:00am            Achi Brandt MG Re-Examined 

                   Mark Adams Heuristics for the Automatic Construction of
 8:25am            Coarse Grids in Multigrid Solvers for Finite Element
                   Problems in Solid Mechanics 

                   Michael Bader An Algebraic Multigrid Method with Fixed
 8:50am            Coarse Grid Selection for Convection Diffusion Equations
                   Tim Chartier Algebraic Multigrid Based on Element
 9:15am            Interpolation (AMGe): Coarsening and the AMGe Measure
                   Mike DeLong Recent Results with Algebraic Multilevel
 9:40am            Methods for ASCI Problems 

 10:05am                                 COFFEE BREAK
     SESSION 2
 Algebraic Methods                  Chair: Van Emden Henson
                   Joel Dendy Some Aspects of Multigrid for Mixed
 10:35am           Discretizations 

                   Michael Griebel Algebraic Multi-Grid for Coupled Systems
 11:00am           of PDEs 

                   Jim E. Jones Algebraic Multigrid Methods Based on
 11:25am           Element Agglomeration 

                   Domenico Lahaye On the Use of Algebraic Multigrid in an
 11:50am           Electromagnetic Systems Simulation Package 

                   Jens-Dominik Müller Coarsening 3-D Hybrid Meshes for
 12:15pm           Multigrid Methods 

     SESSION 3
      Fluids                          Chair: Duane Melson
                   Scott R. Fulton Performance of a Self-Adaptive Multigrid
 4:30pm            Algorithm 

                   Klaus Johannsen Robust Multigrid Methods for Convection
 4:55pm            Diffusion Problems with Closed Characteristics 

                   David Kay A Multigrid Preconditioner for the Steady
 5:20pm            State Navier-Stokes Equations 

                   Boris Lastdrager The Sparse-Grid Combination Technique
 5:45pm            for a Time-Dependent Advection Problem 

                   Marius Paraschivoiu Multi-Level Full Potential and Euler
 6:10pm            Formulation for Transonic Flows 

       CIRCUS                        Chair: Craig Douglas
                   Everyone is welcome to talk and/or listen. The program
                   will be determined at the start by polling the
 7:30pm            participants. Contact Craig for suggestions and
                   information, or just show up and be ready to talk,
                   interact, or just listen.

                             Tuesday, April 13

 Algebraic Methods                 Chair: Van Emden Henson
                  Yousef Saad ARMS : an Algebraic Recursive Multilevel
 8:00am           Solver for General Sparse Linear Systems 

                  Serguei Sokol Applying Multi-Coarse Correction with
 8:25am           Suboptimal Operators to CFD Problems 

                  Ray Tuminaro ML: A Multilevel Framework for Parallel
 8:50am           Unstructured Grid Computations 

                  Jun Zhang A Multilevel Block ILU Preconditioning
 9:15am           Technique for Solving General Sparse Linear Systems
 9:40am                                 COFFEE BREAK
 Algebraic Methods                 Chair: Van Emden Henson
                  John W. Ruge Algebraic Multigrid Applied to Multi-Body
 10:10am          Elasticity Problems with Interface Constraints 

                  Petr Vanek Smoothed Aggregation Multilevel Methods and
 10:35am          Their Convergence Analysis 

                  Jan Mandel Fast Computation of Energy Minimal Coarse
 11:00am          Basis Functions by Smoothing and Projection 

                  Marian Brezina Smoothed Aggregation-Based Black-Box
 11:25am          Iterative Solver 

                  Hervé Guillard A Petrov Galerkin Smoothed Aggregation
 11:50am          Method 

 12:15pm          Rudolph Lorentz Total Reduction Revisited 

      Fluids                         Chair: Irad Yavneh
                  Michael Pernice Hybrid Solvers for the Steady State
 4:30pm           Incompressible Navier-Stokes Equations 

                  William J. Rider A Serendipitous Recovery of Linear
 4:55pm           Scaling 

                  Jonathan Rochez A Multigrid Strategy for Accelerating
 5:20pm           Steady-State Computations of Waves Propagating with
                  Curvature Dependent Speeds 

                  Carol S. Woodward Comparison of Parallel
 5:45pm           Newton-Krylov-Multigrid Solvers for Variably Saturated
                  Flow Problems 

                  Shu-hua Chen The Application of the Multigrid Method in a
 6:10pm           Nonhydrostatic Atmospheric Model 

                                 Moderator: Van Emden Henson

 7:30pm           Open discussion on algebraic methods. Everyone is welcome
                  to talk and/or listen.

                            Wednesday, April 14

         SESSION 7
 Parallel Algebraic Methods               Chair: Jim E. Jones
                            Ulrike Meier Yang Coarse-Grid Selection for
 8:00am                     Parallel Algebraic Multigrid 

 8:25am                     Arnold Krechel Parallel Algebraic Multigrid
                            Marc Alexander Schweitzer Three Parallelization
 8:50am                     Strategies for AMG 

                            Dimitri J. Mavriplis A Highly Scalable
 9:15am                     Unstructured Agglomeration Multigrid Algorithm
                            for Viscous Turbulent Flows 

 9:40am                                      COFFEE BREAK
         SESSION 8
    Parallel Techniques                   Chair: Jim E. Jones
                            Colin J. Aro Implicit Mechanical Modeling in
 10:10am                    ALE3D 

                            Craig C. Douglas Accelerating ADI Methods on
 10:35am                    Parallel Processors including Multigrid with
                            ADI as the Smoother 

 11:00am                    Paul O. Frederickson Recent Results on PUMG
                            Ignacio M. Llorente Smoothing Properties for
 11:25am                    Plane Smoothers for Multiblock Grids 

                            William F. Mitchell Approaches to Parallel
 11:50am                    Multigrid with the Full Domain Partition
         SESSION 9
    PDE Reformulations                    Chair: Kirk Jordan
                            Travis Austin Multigrid Ideas in FOSLS Neutron
 4:30pm                     Transport Theory 

                            Markus Berndt Multigrid for FOSLS on the
 4:55pm                     Discontinuous Coefficient Problem with Singular
                            Basis Functions 

                            Zhiqiang Cai Least Squares for the Perturbed
 5:20pm                     Stokes Equations and the Reissner-Mindlin Plate
                            Andrea Codd Steady-State Elasticity-Fluid
 5:45pm                     Coupled Systems 

                            Johannes Korsawe FAS Multilevel Methods for
 6:10pm                     Nonlinear Least-Squares Finite Element

                            Irad Yavneh Towards Automatic Differential
 6:35pm                     Preconditioning and Variable Transformations
                            for PDE Systems 

          BANQUET                        Cash bar and banquet.
 7:30pm                     Victoria B room in the Pavillion.

                             Thursday, April 15

     SESSION 10
 Parallel Techniques                   Chair: Jim E. Jones
                     David Moulton MPI-Based Black Box Multigrid for
 8:00am              Workstation Clusters 

                     I. D. Parsons Scalable Multigrid Methods Using Fortran
 8:25am              90 and MPI 

                     Markus Kowarschik How Much Can Be Gained by Cache
 8:50am              Aware Programming of Multigrid Methods 

                     Jonathan Hu Cache Based Multigrid on Unstructured
 9:15am              Grids 

 9:40am                                   COFFEE BREAK
     SESSION 11
 Parallel Techniques                   Chair: Jim E. Jones
                     Dana Knoll A Parallel, Two-Level, Preconditioner for
 10:10am             Unstructured Grid Elliptic Problems 

                     Stefan Vandewalle A Multigrid Method for Partial
 10:35am             Differential Equations with Time-Delay 

                     Gerhard Zumbusch Parallelization of an Adaptive
 11:00am             Multilevel Code Using Keys and a Space-Filling Curve
                     Guo Qingping Virtual Boundary Condition Forecast
 11:25am             Algorithm in Multigrid Domain Decomposition Parallel

                     Wing Lok Wan Parallel Sparse Approximate Inverse
 11:50am             Smoothers 

     SESSION 12
  Inverse Problems                     Chair: Eldad Haber
                     Eldad Haber Inversion of 3D DC Resistivity Using
 4:30pm              Approximate Krylov-Gauss-Newton Method 

                     Uri M. Ascher A Multilevel Approach to Distributed
 4:55pm              Parameter Estimation 

                     Yaoguo Li Rapid Construction of an Equivalent Source
 5:20pm              Using Wavelet Transforms and Incomplete Conjugate

                     Gregory Newman Computational Techniques for
 5:45pm              Accelerating Solutions of 3D Geophysical Inverse

     SESSION 13
   Special Methods                     Chair: Eldad Haber
                     Zhangxin Chen Convergence Analysis of Multigrid
 6:10pm              Methods for the Wilson Nonconforming Finite Element
                     Linda Stals Performance of Multilevel Methods on
 6:35pm              Equations with Large Jumps in Coefficients 

                              Friday, April 16

       SESSION 14
  Parallel and Algebraic                 Chair: Jim E. Jones
                          Michael Kuhn A Unifying Software Concept for
 8:00am                   Several Parallelization Techniques 

                          Gundolf Haase Parallel AMG for Non-overlapping
 8:25am                   Domain Decomposition 

                          Stefan Reitzinger Algebraic Multigrid Methods
 8:50am                   Based on Element Stiffness Matrices 

                          Ping Wang Parallel Computation of 3D,
 9:15am                   Time-dependent, Thermal Convective Flows 

 9:40am                                     COFFEE BREAK
       SESSION 15
  Decomposition Methods              Chair: Panayot Vassilevski
                          Susanne C. Brenner Lower Bounds for
 10:10am                  Nonoverlapping Domain Decomposition Methods
                          Serge Goossens Two-level Algorithms for
 10:35am                  Overlapping Composite Mesh Difference Methods
                          Thomas Huckle Multigrid Preconditioning and
 11:00am                  Toeplitz Matrices 

                          David E. Keyes Newton-Krylov Methods with
                          Multilevel Preconditioning:
 11:25am                  Algorithm-Architecture Trade-offs in the Number
                          of Levels 

                          Maria Sosonkina Distributed Multilevel Schur
 11:50am                  Complement Preconditioning of General Sparse
                          Linear Systems 

                          Wienands Fourier Analysis of GMRES(m)
 12:15pm                  Preconditioned by Multigrid 

 12:40pm                  End of conference

    Editor's Note: See
    -------------  for last minute changes to the program.


End of MGNet Digest