Send mail to: mgnet@cs.yale.edu for the digests mgnet-requests@cs.yale.edu for comments or help Anonymous ftp repository: casper.cs.yale.edu (128.36.12.1) ftp.cerfacs.fr (138.63.200.33) World Wide Web: http://na.cs.yale.edu/mgnet/www/mgnet.html or http://www.cerfacs.fr/~douglas/mgnet.html Today's editor: Craig Douglas (douglas-craig@cs.yale.edu) Volume 5, Number 10 (approximately October 31, 1995) Today's topics: Are you keeping a shadow or have an active multigrid WWW site? ~/.mailcap file for web access to MGNet Kaskade Code Update 5 Papers (Y. Shapira) Preprint: Multigrid Method for Convection-Diffusion Equation Boundary Value ODE book Algebraic MG Conference (revised paper due date) Workshop in Bulgaria ------------------------------------------------------- Date: Tue, 31 Oct 1995 13:25:18 -0500 (EST) From: Craig Douglas Subject: Are you keeping a shadow or have an active multigrid WWW site? I am aware of at least one partial and one complete shadow of the contents of MGNet. As any of you know who regualrly use the Internet, the world wide web is rapidly reducing the data transfer rate on the Internet to zero. If you are running a shadow and are willing to admit it, please let me know so that I can put a pointer to your machine into the MGNet web pages (for either WWW or anonymous ftp access). While I am updating the web pages for MGNet, if you have a WWW page of interest to the multigrid or domain decomposition communities, please let me know so I can add a hypertext link to it. Thanks, Craig ------------------------------------------------------- Date: Tue, 31 Oct 1995 10:05:28 -0500 (EST) From: Craig Douglas Subject: ~/.mailcap file for web access to MGNet Many people send messages to me asking how to avoid having to store the gzipped files before viewing them when accessing MGNet with a web browser. I have no idea if you are using a Mac, OS/2, or Windows. On the UNIX systems that I use, however, I have a file in my home directory called .mailcap. Mine contains the following lines: audio/*; audiotool %s image/*; xv %s video/mpeg; mpeg_play %s video/gl; xgl %s video/dl; xdl %s application/postscript; ghostview %s application/x-dvi; xdvi %s ------------------------------------------------------- Date: Thu, 12 Oct 1995 09:37:55 +0100 From: erdmann@ZIB-Berlin.DE (Erdmann) To: mgnet Subject: Kaskade Code Update KASKADE ======= Adaptive multilevel-code for linear scalar elliptic and parabolic problems in 1, 2, 3 space dimensions. We included example algorithms for nonlinear methods used in obstacle, porous media or stefan problems. --- a C++ toolbox ---- Authors: Rudolf Beck Rainer Roitzsch Bodo Erdmann Last update: 10th October 1995 References: =========== 1. P. Deuflhard, P. Leinen, H. Yserentant: Concepts of an Adaptive Hierarchical Finite Element Code. IMPACT, 1, 1989. 2. F. Bornemann: An Adaptive Multilevel Approach to Parabolic Equations in Two Space Dimensions. Dissertation, Freie Universitaet Berlin, 1991. 3. F. Bornemann, B. Erdmann, R. Kornhuber: Adaptive Multilevel Methods in Three Space Dimensions. Int. J. Numer. Meths. in Eng., 36, 1993. 4. R. Kornhuber: Monotone Multigrid Methods for Nonlinear Variational Problems. Habilitationsschrift, Freie Universitaet Berlin, 1995. 5. R. Beck, B. Erdmann, R. Roitzsch: Kaskade 3.0, An Object-Oriented Adaptive Finite Element Code Techn. Report TR 95-4, 1995. 5. R. Beck, B. Erdmann, R. Roitzsch: Kaskade 3.0, User's Guide Techn. Report TR 95-11, 1995. Abstract: ========= The KASKADE 3.0 software package solves linear scalar elliptic and parabolic problems in 1, 2, 3 space dimensions with adaptive finite element methods. Furthermore, the toolbox includes extensions for handling systems of equations and example algorithms for nonlinear methods used in obstacle, porous media or Stefan problems. Core of the program is a variety og multilevel/multigrid preconditioners for the arising linear systems. This object-oriented code is written in C++ and can be compiled with Gnu g++, version 2.6.3, and some other compilers. It solves the same mathematical problem classes as its predecessor KASKADE 2.x, which is written in C. The code, a programmer's manual describing the software design, and a user's guide are available by anonymous ftp from the MGNet or from the eLib at the Konrad-Zuse-Zentrum in Berlin. elib: ===== ftp elib.zib-berlin.de in the subdirectories /pub/kaskade/3.0 and /pub/kaskade/Manuals/3.0 MGNet: ====== ftp na.cs.yale.edu in the subdirectories /mgnet/Codes/kaskade/3.0 How to use the code? 1. uncompress 3.0.tar.Z - or - gunzip 3.0.tgz 2. tar -xf 3.0.tar (Creates a directory 3.0 with the sources) 3. cd 3.0 4. make (Uses the make-file 'makefile' to compile and link the executable 'k6' handling 1D-, 2D-, and 3D-problems. In the make-file there are four targets (k1,k2,k3,k6 (default)) to obtain seperate versions for different space dimensions ( - and one that comprises all of them). You find some more information about this in the short documentation in the main file kaskade.cc or in the user's guide. The file kaskade.make is a copy of the makefile. You should use it to define new dependencies of files.) 5. k6 cmd=unit2 (Starts the program, the command file unit2 defines a 2D problem) How to use the Programmer's Manual? 1. uncompress programmer_guide.ps.Z - or - gunzip programmer_guide.ps.gz 2. output on a postscript printer How to use the User's Guide? 1. uncompress user_guide.ps.Z - or - gunzip user_guide.ps.gz 2. output on a postscript printer We are extending the User's Guide (tutorial) continously. Address: Konrad-Zuse-Zentrum Berlin (ZIB) Heilbronnerstrasse 10 10711 Berlin Germany Telefon: 0049+30+89604-215 e-mail: erdmann@zib-berlin.de roitzsch@zib-berlin.de For questions and remarks, please use the e-mail addresses. _______________________________________________________________ COPYRIGHT/Licence ================= You may use or modify this code for your own non-commercial purposes for an unlimited time. In any case you should not deliver this code without a special permission of ZIB. In case you intend to use the code commercially, we oblige you to sign an according licence agreement with ZIB. _______________________________________________________________ Bodo Erdmann Rainer Roitzsch Bodo Erdman | Konrad-Zuse-Zentrum fuer Informationstechnik (ZIB) erdmann@sc.zib-berlin.de | Abt. Numerische Software-Entwicklung Telefon: (030) 89604-215 | Heilbronner Str.10 Fax: -125 | D-10711 Berlin - Wilmersdorf ------------------------------------------------------- Date: Thu, 19 Oct 1995 14:15:31 -0600 From: Yair Shapira Subject: 5 Papers I put 5 papers on mgnet... My new email address is yairs@lanl.gov Best regards Yair Shapira. Towards Automatic Multigrid Algorithms for SPD, Nonsymmetric and Indefinite Problems Yair Shapira, Moshe Israeli and Avram Sidi Computer Science Department, Technion, Haifa $32000$, Israel email: yairs@lanl.gov Abstract A new multigrid algorithm is constructed for the solution of linear systems of equations which arise from the discretization of elliptic PDEs. It is defined in terms of the difference scheme on the fine grid only, and no rediscretization of the PDE is required. Numerical experiments show that this algorithm gives high convergence rates for several classes of problems: symmetric, nonsymmetric and problems with discontinuous coefficients, non-uniform grids and non-rectangular domains. When supplemented with an acceleration method, good convergence is achieved also for pure convection problems and indefinite Helmholtz equations. Editor's Note: in mgnet/papers/Shapira/automg.ps.gz and ------------- mgnet/papers/Shapira/automg.abs * * * * * * * * * * Multigrid Techniques for 3-D Definite and Indefinite Problems with Discontinuous Coefficients Yair Shapira Computer Science Department, Technion -- Israel Institute of Technology, Haifa 32000, Israel email: yairs@lanl.gov Abstract A multigrid method for the solution of finite difference approximations of elliptic PDEs is introduced. A parallelizable version of it, suitable for two and multi level analysis, is also defined, and serves as a theoretical tool for deriving an optimal implementation for the main version. For indefinite Helmholtz equations, this analysis provides a prediction of the optimal mesh size for the coarsest grid used. Numerical experiments show the applicability of the method to 3-d diffusion problems with discontinuous coefficients and highly indefinite Helmholtz equations. Editor's Note: in mgnet/papers/Shapira/automgD3.ps.gz and ------------- mgnet/papers/Shapira/automgD3.abs * * * * * * * * * * Coloring Update Methods Yair Shapira Computer Science Department, Technion -- Israel Institute of Technology, Haifa 32000, Israel email: yairs@lanl.gov Abstract For linear update methods with nonsingular iteration matrices, a coloring method is introduced for which the multicolor iteration matrix is similar to the original one. It is general in the sense that its definition is independent of grids and stencils. A method for transforming eigenvectors of the original iteration matrix to those of the multicolor one is introduced. Applications to unstructured grids and multigrid solution of three-dimensional problems are presented. Editor's Note: in mgnet/papers/Shapira/colors.ps.gz and ------------- mgnet/papers/Shapira/colors.abs * * * * * * * * * * Parallelizable Approximate Solvers for Recursions Arising in Preconditioning Yair Shapira Computer Science Department, Technion -- Israel Institute of Technology, Haifa 32000, Israel email: yairs@lanl.gov Abstract For the recursions used in the Modified Incomplete LU (MILU) preconditioner, namely, the incomplete decomposition, forward elimination and back substitution processes, a parallelizable approximate solver is presented. The present analysis shows that the solutions of the recursions depend only weakly on their initial conditions and may be interpreted to indicate that the inexact solution is close, in some sense, to the exact one. The method is based on a domain decomposition approach, suitable for parallel implementations with message passing architectures. It requires a fixed number of communication steps per preconditioned iteration, independently of the number of subdomains or the size of the problem. The overlapping subdomains are either cubes (suitable for mesh-connected arrays of processors) or constructed by the data-flow rule of the recursions (suitable for line-connected arrays with possibly SIMD or vector processors). Numerical examples show that, in both cases, the overhead in the number of iterations required for convergence of the preconditioned iteration is small relatively to the speed-up gained. Editor's Note: in mgnet/papers/Shapira/iluparallel.ps.gz and ------------- mgnet/papers/Shapira/iluparallel.abs * * * * * * * * * * Two-Level Analysis and Multigrid Methods for SPD, Non-Normal and Indefinite Problems Yair Shapira Computer Science Department, Technion -- Israel Institute of Technology, Haifa 32000, Israel email: yairs@lanl.gov Abstract A convergence theory for Black-Box Multigrid for a class of SPD problems is presented. Improved versions of Black-Box Multigrid for diffusion problems with discontinuous coefficients are defined. A two-level analysis method for several automatic multigrid methods for certain separable problems is introduced. Unlike standard two-level analysis methods, based on Fourier analysis, it is based on spectral analysis, hence applicable to non-normal problems and to certain problems with variable coefficients. For indefinite problems, it provides a way to choose an optimal mesh size for the coarsest grid used and motivates the definition of an improved version of Black-Box Multigrid. Numerical experiments confirming the analysis are reported. Editor's Note: in mgnet/papers/Shapira/tlanalysis.ps.gz and ------------- mgnet/papers/Shapira/tlanalysis.abs ------------------------------------------------------- Date: Tue, 24 Oct 1995 12:26:38 -0400 From: Jun Zhang Subject: Preprint: Multigrid Method for Convection-Diffusion Equation An Accurate and Stable Multigrid Method for Convection-Diffusion Equations Murli M. Gupta, Jules Kouatchou and Jun Zhang Department of Mathematics The George Washington University Washington, DC 20052, USA Abstract: We introduce a high-order compact difference scheme with multigrid algorithm to solve the convection-diffusion equations with constant coefficients. This high-order discretization scheme is shown to be more accurate and stable than the usual five-point discretization scheme. It solves the convection- diffusion equations directly without using any preconditioner or adding any artificial dissipation terms. This method is shown to converge faster than some of the existing methods and to achieve higher accuracy. Numerical experiments are presented to validate the conclusions. Please send your comments to: zhang@math.gwu.edu Editor's Note: in mgnet/papers/Gupta-Kouatchou-Zhang/con_diff.ps.gz and ------------- mgnet/papers/Gupta-Kouatchou-Zhang/con_diff.abs. ------------------------------------------------------- Date: Thu, 12 Oct 95 09:48:54 BST From: "M. Ainsworth" Subject: Article for MGNet VIIth EPSRC NUMERICAL ANALYSIS SUMMER SCHOOL LEICESTER UNIVERSITY, UK 8th-19th July 1996 The Programme ------------- The meeting will comprise two one-week modules, each of which can "stand alone", although it is expected that many participants and speakers will stay for the full two weeks. From Monday to Friday each week there will be three five-lecture courses given by the invited lecturers as follows: Week 1, 8th-12th July 1996 ~~~~~~~~~~~~~~~~~~~~~~~~~~ G. Cybenko (Dartmouth) "Neural Networks" M. Plum (Clausthal) "Eigenvalue Problems for Differential Equations" G. Strang (MIT) "Wavelets" Week 2, 15th-19th July 1996 ~~~~~~~~~~~~~~~~~~~~~~~~~~~ L. Greengard (NYU) "Multipole Methods" C. Schwab (ETH, Zurich) "Hierarchical Modelling" J. Xu (Penn State) "Multilevel and Domain Decomposition Methods" The principal aim of the meeting is to gather together numerical analysts and a team of internationally renowned experts for a period of intensive study and research. It is intended that the lectures should be accessible to people (particularly research students) for whom the material is new, to enable them to acquire reasonable competence in it, thus broadening their research horizons. Those with greater initial knowledge should end up being able to work on significant problems in the area. There will be a substantial amount of time available for research and discussion with the assembled experts, who will make themselves available for consultation in "office hours". Typeset lecture notes will be provided by most of the speakers. In addition, there will be an opportunity for participants to present research seminars on their own work. It is anticipated that there will also be book exhibitions and displays of computer software. Registration forms and further details are available (electronically or by surface mail) from: Dr M. Ainsworth (ain@mcs.le.ac.uk), Mathematics and Computer Science, Leicester University, Leicester LE1 7RH, United Kingdom. ------------------------------------------------------- Date: Wed, 4 Oct 1995 11:49:49 UTC-0700 From: Uri Ascher Subject: Boundary Value ODE book Dear Colleagues, Our book, Numerical Solution of Boundary Value Problems for Ordinary Differential Equations has recently been published with SIAM in the Classics series and is now available. The first edition of this book, published in 1988 by Prentice-Hall, became unavailable in 1993. The current edition contains many small corrections but no major ones. Also, it's in a softcover volume and is significantly cheaper than the original edition. Those of you who are interested in this field may find the book very helpful. Please feel free to contact SIAM for further information: siam@siam.org ISBN 0-89871-354-4 Uri Ascher, Bob Mattheij and Bob Russell ------------------------------------------------------- Date: Fri, 20 Oct 1995 17:37:28 +0100 From: Maya Neytcheva Subject: Algebraic MG Conference (revised paper due date) Announcement and call for papers CONFERENCE ON ALGEBRAIC MULTILEVEL ITERATION METHODS WITH APPLICATIONS June 13-15, 1996, University of Nijmegen, The Netherlands PROGRAM COMMITTEE: Owe Axelsson, Nijmegen, The Netherlands Dietrich Braess, Bochum, Germany Tony F. Chan, Los Angeles, California Richard E. Ewing, College Station, Texas Wolfgang Hackbusch, Kiel, Germany Piet Hemker, Amsterdam, The Netherlands Yuri A. Kuznetsov, Moscow, Russia Ulrich Langer, Linz, Austria Jean-Francois Maitre, Lyon, France Panayot S. Vassilevski, Sofia, Bulgaria David M. Young, Austin, Texas, honorary member Harry Yserentant, Tubingen, Germany ADDRESS FOR CORRESPONDENCE: Prof. Owe Axelsson Faculty of Mathematics and Informatics Toernooiveld 1, NL-6525 ED Nijmegen The Netherlands e-mail: amli96@sci.kun.nl} fax: +31 (0)24 3652140 LOCAL ORGANIZATION COMMITTEE: Owe Axelsson, Ben Polman, Rob Stevenson, Maya Neytcheva, Mariana Nikolova SCOPE: The purpose of the conference is to provide a forum for the presentation and the discussion of recent progress in the analysis, implementation and applications in various fields of algebraic multilevel iteration methods in a broad sense. This includes their implementation on massively parallel computers. TOPICS covered include Algebraic Multilevel Iteration methods for * second and fourth order elliptic scalar equations and systems of equations * mixed variable variational problems * nonselfadjoint problems and indefinite matrix problems * inner-outer iteration methods * parallel implementations, efficiency measures, scalability * robust implementations, i.e. convergence uniform with respect to meshsize parameter and singular perturbation parameters * applications for Navier's equations and Stokes problem * applications outside partial differential equation problems * applications for nonlinear problems, such as electromagnetic field, plastic flow, Navier-Stokes, and Miscible displacement problems. CALL FOR PAPERS: Papers intended for presentation at the conference should be submitted to Owe Axelsson. All papers should be up to 12 pages delivered in a plain LaTeX format preferably as \documentclass[a4paper,12pt]{article} \usepackage{a4wide} and submitted either by electronic mail or on a floppy disk. The submission should be accompanied by a printout sent by ordinary mail. The papers accepted for presentation at the conference are planned to appear in a proceedings volume ready for the conference. Authors who are unable to produce a paper in LaTeX or TeX format are requested to contact the organizers. All papers will be refereed. CALENDAR: Deadline for submission of contribution papers: December 15, 1995. Notification of acceptance: March 15, 1996. GENERAL INFORMATION: The registration fee will be DFL 600,- (currently $350) and includes a copy of the conference proceedings, two lunches and coffee and tea during breaks. The conference language will be English. ------------------------------------------------------- Date: Mon, 4 Sep 95 11:22:22 +0300 From: yalamov@amigo.acad.bg (Plamen Yalamov) Subject: Workshop in Bulgaria FIRST WORKSHOP ON NUMERICAL ANALYSIS AND APPLICATIONS RUSSE, BULGARIA, JUNE 24-27, 1996 Organizers: University of Russe, Association of Bulgarian Mathematicians - Russe Co-organizers: Institute of Mathematics and Center for Informatics and Information Technologies of the Bulgarian Academy of Sciences, Technical University of Gabrovo, Technical University of Sofia Traditionally every 4 years a Conference on Numerical Analysis and Applications is organized in Bulgaria. The present workshop is meant to support this tradition and to serve as an intermediate meeting between these conferences. We would like to give an opportunity for mathematicians and applied scientists to discuss topics of common interest. The workshop will have three tracks: 1. Numerical linear algebra. 2. Numerical methods for differential equations. 3. Numerical modelling. Preliminary list of Invited Speakers: R. Bisseling (Netherlands), L. Brugnano (Italy), S. K. Godunov (Russia), A. Griewank (Germany), A. Hadjidimos (USA), S. Hammarling (UK), W. Hofmann (Germany), A. Karageorghis (Cyprus), Yu. A. Kuznetsov (Russia), R. Maerz (Germany), W. T. Pickering (UK), R. Plemmons (USA), I. V. Puzynin (Russia), G. I. Shishkin (Russia), T. Szulc (Poland), E. E. Tyrtyshnikov (Russia), W. Varnhorn (Germany), V. V. Voevodin (Russia), Z. Zlatev (Denmark). Organizing committee: L. Vulkov (Chair), P. Yalamov (co-Chair), A. Andreev, S. Chernev, P. Ivanova, I. Lirkov, M. Paprzycki, V. Pavlov, S. Romanova, N. Strateva,T. Todorov, Z. Zlatev, K. Zlateva. We would like to invite all interested individuals to ORGANIZE a MINISYMPOSIUM related to one or more of the conference tracks. Please send a minisymposium abstract (approximately one page) and a list of 4-8 speakers to one of the addresses listed below. The deadline for proposals is December 1, 1995. A general call for papers and more details about the meeting will be provided in the future announcements. For more information, please, contact: Plamen Yalamov Marcin Paprzycki Dept. of Mathematics Dept. of Mathematics and CS University of Russe UTPB 7017 Russe Odessa, TX 79762 BULGARIA USA yalamov@iscbg.acad.bg paprzycki_m@gusher.pb.utexas.edu ------------------------------ End of MGNet Digest **************************