Send mail to: mgnet@cs.yale.edu for the digests or bakeoff mgnet-requests@cs.yale.edu for comments or help Current editor: Craig Douglas douglas-craig@cs.yale.edu Anonymous ftp repository: ftp.ccs.uky.edu (128.163.209.106) World Wide Web: http://www.mgnet.org or http://casper.cs.yale.edu/mgnet/www/mgnet.html or http://www.cerfacs.fr/~douglas/mgnet.html or http://phase.etl.go.jp/mgnet or http://www.nchc.gov.tw/RESEARCH/Math/mgnet/www/mgnet.html Today's editor: Craig Douglas (douglas-craig@cs.yale.edu) Volume 9, Number 4 (approximately April 30, 1999) Today's topics: Postcard from Copper Mountain Prometheus-1.0 PLTMG8.1 Copper 99 Contribution (Mavriplis) Copper 99 Contribution (Reitzinger) Copper 99 Contribution (Pernice) Copper 99 Paper (Haase) DD12 Second Announcement International Workshop on Computational Physics Abstracts Due Date MGNet Bibliography (Zumbusch) ------------------------------------------------------- Date: Sat, 17 Apr 1999 09:31:22 -0500 (PDT) From: Craig C. DouglasSubject: Postcard from Copper Mountain Once again many of the multigrid crowd gathered in Copper Mountain for a Colorado conference during April 12-16. Just two of the participants have attended all thirteen. There were many new participants and many from Europe. (Hopefully there will be many Americans at the European Multigrid Conference in Gent at the end of September.) There were about twenty graduate students in attendance of which four took home awards. Number thirteen in the Colorado conference series turned out to be a very important number when Achi Brandt suffered a skiing injury (a very, very superstitious occurance). Achi showed just how tough he is by coming to the conference dinner and talks right afterwards. The conference series is also demonstrating that it is Y2K compliant by announcing already that there will be conferences in the spring of 1900 (iterative methods) and 2001 (multigrid). Of course, the real millenium bug is best seen by pointing your favorite browser at the animation contained in http://www.mgnet.org/resources/mil_bug.gif (my thanks to its creator). The parabolic skis from two years ago were now touted as styled skis (the public seemingly did not appreciate time dependent skiing). The snow at the conference was well conditioned with lots of new snow just before and during the conference, though nobody seemed to have time to notice this fact except for a O(N) sized set who spent O(1) time on the slopes. The 1999 conference had two major themes: algebraic multigrid (organized by Van Henson) and parallel multigrid (organized by Jim Jones). There were 69 talks and 3 tutorials. We had the following number of talks on the special topics: 21 Algebraic Multigrid 8 Parallel Algebraic Multigrid 15 Parallel Multigrid While we did not have any parallel sessions, we certainly had lots of parallel talks. The other sessions were devoted to fluids, pde reformulations, inverse problems, special methods, and decomposition methods. Besides regular sessions, there was the circus and the bulls were raging in the algebraic stock market session, easily surpassing 10,000 on the DOW (Doesn't One Wonder what this means?). The circus had three talks. The Linz group demonstrated their program FEPP, which is a CAD based adpative grid multidimensional finite element multigrid package that can handle industrial quality problems easily. We had a wavelet talk by Isreali. Yavneh discussed a new solution of "determining the height of a blanket that covers piles of junk" (his title, no kidding). During the AMG bull session there were many comments that said that algebraic multigrid works when problems have two features: locality and low dimensional local subspaces. There were also discussions on minor variants. Missing from the session was a discussion of what the various incarnations of algebraic multigrid really are and what they should be trying to accomplish. Continuing a tradition, there was a fossil's, er, FOSLS session. The speakers dug up a number of new topics. My favorite was the talk by Andrea Codd, who described how to simulate a square eye. (What? Your eye is three dimensional? That is simply too complicated just yet to work with.) By converting her problem into a small print, page long first order system and minimizing it, she will be able to get a Ph.D. and go home to Australia. Astonishingly, there is a faculty member at the University of Colorado who is building a square eye in order for her to see how accurate her simulations really are! (I maintained after the talk, and still do, that if she can solve such a hard problem, it would be a real shame for her to go home. Surely there is someone in the United States who can convince her to stay after graduation instead.) There were a number of talks highlighting that we are almost to the point that given part of a multilevel problem, the rest can be generated automatically at what is potentially an acceptable cost. There are also methods that operate even more optimally, but they have an O(n!) step instead of the textbook multigrid efficiency that was once an astonishing O(n)! All encompassing (parallel) multilevel packages that generate problems, visualize the solutions, and can solve industrial strength problems have become a growth industry. While packages have been developed for sale for large amounts of money, there are even some that are free. There were talks describing a number of these packages: Los Alamos, Sandia, Livermore, NIST, Linz ($$$?), and the GMD spinoffs ($$$). This on top of ones from places like Stuttgart/Heidelberg, San Diego, and others who did not give talks. If I missed your package, send me email and I will include it in the web page on codes, http://www.mgnet.org/mgnet-codes.html. Multigrid has always been known as a high performance method. A question has been raised by the two groups caching in on multigrid that maybe it can run 2-5 times faster and get the exact same answer as the traditional implementations. Does this mean that most people have been writing poor codes? No, it means that the hardware architects have been producing faster CPU's with the same old memory speeds. Hence, we need more sophisticated algorithms. Alternately, buying computers with 3-6 times as many processors seems to be an acceptable alternative. Of course, with talks describing how various multilevel algorithms scaled on four to ten thousand actual processors was different from recent conferences. Papers for the special issue are due at ETNA no later than May 17! See you at the 2000 and 2001 conferences. ------------------------------------------------------- Date: Fri, 16 Apr 1999 12:40:35 -0700 From: Mark Adams Subject: Prometheus-1.0 We would like to announce the alpha release of the Prometheus library: a fully parallel multigrid-based linear system solver for unstructured 3D Finite Element problems. Prometheus takes a FE mesh distributed across multiple processors, automatically generates all the coarse representations and operators for standard multigrid algorithms, and solves the linear system. Prometheus is designed for distributed memory machines, and has been used to solve irregular 40 million degree-of-freedom problems, with large (10^4) jumps in material coefficients and incompressible materials (v=0.49), on 960 IBM PowerPC processors with over 50% parallel efficiency. It uses C++ with MPI, is built on PETSc (www.mcs.anl.gov/petsc) and ParMetis (www-users.cs.umn.edu/~karypis/metis), and is correspondingly portable. It has been tested extensively on the Cray T3E and IBM PowerPC cluster; support for more platforms will be added depending on demand. Prometheus has been developed for large scale solid mechanics problems and has only been run on symmetric positive definite systems. However, Prometheus can in principle be used for any problem for which effective Krylov and multigrid algorithms exist (multigrid conference proceedings are one source for such algorithms e.g., www.mgnet.org). Prometheus uses no explicit information about the partial differential equation being solved (it only uses geometric information about the fine grid mesh). The primary purpose of Prometheus is to construct standard multigrid operators for unstructured FE meshes for use in multigrid preconditioners for Krylov solvers. The current version supports only first order hexahedral and tetrahedral elements; the next release will support shell and beam elements as well. Prometheus-1.0 documentation and library can be found at www.cs.berkeley.edu/~madams/Prometheus-1.0. It is freely available. We are interested in interacting with potential users to get feedback on the effectiveness and usability of the algorithms and code, to help in the continued development of Prometheus. Mark Adams (madams@cs.berkeley.edu) Jim Demmel (demmel@cs.berkeley.edu) R.L.Taylor (rlt@ce.berkeley.edu) ------------------------------------------------------- Date: Mon, 26 Apr 1999 10:28:21 -0700 From: "Randolph E. Bank" Subject: PLTMG8.1 I made a few changes to the pltmg8.1 tar file: 1. I put in a separate makefile for linux/g77 2. I changed to GUI logo from pltmg8.0 to pltmg8.1 (oops....) Editor's Note: See http://www.mgnet.org/mgnet-codes-pltmg.html ------------- ------------------------------------------------------- Date: Fri, 09 Apr 1999 16:03:52 -0400 From: Dimitri Mavriplis Subject: Copper 99 Contribution (Mavriplis) copper.99.mavriplis.talk.gz is the name of the file I have put on the MGNet server. This is not a paper, but a copy of my view graphs for the 1999 Copper Mtn MG conference. This is also viewable at http://www.icase.edu/~dimitri/copper.99 A Highly Scalable Unstructured Aggolomeration Multigrid Algorithm for Viscous Turbulent Flows Editor's Note: See http://www.mgnet.org/mgnet-ccmm99.html or access it at ------------- http://www.mgnet.org/Conferences/CopperMtn99/Talks/mavriplis.ps.gz ------------------------------------------------------- Date: Mon, 26 Apr 1999 09:39:59 +0200 From: Stefan Reitzinger Subject: Copper 99 Contribution (Reitzinger) I have put my slides of the 9th Copper Mountain Conference on the MGNet. Unfortunately I have not the time to put a technical report on it, because there is no time to do this in the near future. But I will update it as soon as possible! Hybrid Approaches for Solution of Steady-State Incompressible Navier-Stokes Eqations Editor's Note: See http://www.mgnet.org/mgnet-ccmm99.html or access it at ------------- http://www.mgnet.org/Conferences/CopperMtn99/Talks/reitzinger.ps.gz ------------------------------------------------------- Date: Mon, 19 Apr 1999 14:09:19 -0600 From: Michael Pernice Subject: Copper 99 Contribution (Pernice) Following the suggetion you made at Wednesday night's dinner, I've downloaded a postscript copy of my CuMtn99 talk. Algebraic Multigrid Methods Based on Element Stiffness Matrices Editor's Note: See http://www.mgnet.org/mgnet-ccmm99.html or access it at ------------- http://www.mgnet.org/Conferences/CopperMtn99/Talks/pernice.ps.gz ------------------------------------------------------- Date: Mon, 26 Apr 1999 14:20:39 +0200 From: Gundolf Haase Subject: Copper 99 Paper (Haase) A Parallel AMG for Overlapping and Non-overlapping Domain Decomposition Gundolf Haase Johannes Kepler University Linz Abstract There exist several approaches for the parallel solving of huge systems of linear equations resulting from the discretization of 2nd order elliptic pdes. We distinguish between overlapping and non-overlapping decompositions based on the distribution of finite elements. On the other hand, there exists a great demand on Algebraic Multigrid solvers (AMG) which have as input only matrix and right hand side or, as a substitute, the appropriate information per element. In this paper we propose a parallel AMG algorithm using overlapping or non- overlapping data decompositions. Editor's Note: See http://www.mgnet.org/mgnet-ccmm99.html or access it at ------------- http://www.mgnet.org/Conferences/CopperMtn99/Papers/haase.ps.gz ------------------------------------------------------- Date: Sun, 11 Apr 1999 08:17:15 +0900 From: "dd12" Subject: DD12 Second Announcement Second Announcement and Call for Papers 12th International Conference on Domain Decomposition Methods October 25-29, 1999 Chiba University Chiba, Japan Domain Decomposition (DD) has served as an organizing principle for many concepts and methodologies in mathematics, computer science, and computational science and engineering. And also DD will contribute to originate new concepts and methodologies in related fields mentioned above, which will give a clue to understand and solve complex problems existing in our real world at present. The objective of this conference is to promote understanding and use of DD for the solution of problems arising in various fields of science and engineering and to promote interaction between researchers throughout the above-mentioned disciplines. The conference will include invited plenary talks by leading experts in the field from academia, research institutions, and industry, as well as mini-symposia, contributed papers and graduate student paper sessions. In particular, it will focus on: Conference Themes Domain decomposition in science and engineering Theoretical developments Multilevel methods Parallel machine architecture Parallel algorithms and their implementation Demonstrations and evaluations of large-scale codes Application Topics Acoustics and electromagnetics Aerospace applications Automotive applications Biological applications Control applications Coupled phenomena Dimension reduction Grid generation Mathematical surface problems Micro-electronics applications Optimization Parallel processing Particulate flows Ship applications Conference Deadlines Proposals for mini-symposia April 30 Proposals for contributed papers April 30 Proposals for exhibition May 31 Early registration August 31 Late Registration September 30 PLEASE NOTICE THE CHANGE OF DEADLINES for mini-symposia and contributed papers to April 30. Proposals for Mini-symposia Mini-symposia will be 2 hour sessions of 3-5 speakers focussing on a single topic. The organizer of each mini-symposium will invite the speakers and decide on the topics to be addressed. Proposals for mini-symposia can be submitted via email to dd12@applmath.tg.chiba-u.ac.jp. Session title, organizer, names of speakers and a summary of the topic should be included in the email. Proposals for Contributed Papers If you wish to present a paper (including graduate student paper), please submit an abstract (one page up to 400 words in plain text or Postscript file) via email to dd12@applmath.tg.chiba-u.ac.jp. Invited Speakers Christine Bernardi (University Pierre et Marie Curie, France) Dietrich Braess(Ruhr-University, Germany) Annalisa Buffa (University of Pavia, Italy) Martin Gander (Ecole Polytechnique, France) Marc Garbey(University Lyon 1, France) Roland Glowinski(University of Houston, U.S.A.) Takashi Kako(University of Electro-Communications, Japan) Tshuguo Kondoh(Toyota Central Research and Development Labs., Japan) Ulrich Langer (Johannes Kepler University Linz, Austria) Jacques Louis Lions (French Academy of Science, France) Yvon Maday (University Pierre et Marie Curie, France) Kenichi Miura(Fujitsu Limited, Japan) Masahisa Tabata(Kyushu University, Japan) Olof Widlund(New York University, U.S.A.) Conference Web Page The web site of the conference contains up-to-date information: http://applmath.tg.chiba-u.ac.jp/dd12/ (See also http://www.ddm.org for the history and archives of the conference series.) Scientific Committee P. E. Bjorstad (Chairperson, Bergen, Norway) T. F. Chan (UCLA, USA) P. Deuflhard (ZIB, Germany) R. Glowinski (Houston, USA) R. Hoppe (Augsburg, Germany) H. Kawarada (Chiba, Japan) D. E. Keyes (ODU, USA) Y. Kuznetsov (Houston, USA) J. Periaux(St Cloud, France) O. Pironneau (Paris VI, France) A. Quarteroni (Lausanne, Switzerland) Z. C. Shi (Academia Sinica, China) O. Wildlund (Courant Inst., USA) J. Xu (Penn State, USA) Local Organizing Committee I. Hagiwara (TIT) T. Ikeda (Ryukoku Univ.) H. Imai (Tokushima Univ.) T. Kako (UEC) H. Kawarada (Chairperson, Chiba Univ.) H. Koshigoe (Chiba Univ.) M. Mori (Kyoto Univ.) M. Nakamura (Nihon Univ.) H. Okamoto (Kyoto Univ.) M. Tabata (Kyushu Univ.) T. Takeda (UEC) G. Yagawa (Univ. of Tokyo) Registration Fee Academics and Industrialists Early 26,000 Late 30,000 Full time students Early 13,000 Late 15,000 Accompanying person Early 10,000 Late 12,000 Registration fee includes proceedings, book of abstracts, lunches, refreshments, welcome party, and banquet. Contact Information Mrs. Akemi Tonomura, Institute of Applied Mathematics, Chiba University, 1-33 Yayoicho, Inage-ku, Chiba, 263-8522, JAPAN Email : dd12@applmath.tg.chiba-u.ac.jp TEL&FAX: +81-43-290-3505 ------------------------------------------------------- Date: Mon, 26 Apr 1999 09:47:20 -0500 From: Zhangxin CHEN Subject: International Workshop on Computational Physics Abstracts Due Date The deadline for submitting an abstract has been extended from April 30 to May 30, 1999. See http://www.mgnet.org/conferences/iwcp-fftpm.0899.html for more information. ------------------------------------------------------- Date: Mon, 19 Apr 1999 20:02:57 +0200 (MSZ) From: Gerhard Zumbusch Subject: MGNet Bibliography (Zumbusch) @TechReport{GWZumbusch_1991a, author = {G. W. Zumbusch}, title = {Adaptive parallele {M}ultilevel-{M}ethoden zur {L}\"{o}sung elliptischer {R}andwertprobleme}, institution = {SFB 342, TU M\"{u}nchen}, OPTaddress = {Germany}, year = {1991}, number = {342/19/91 A}, } @MastersThesis{GWZumbusch_1992a, author = {G. W. Zumbusch}, title = {Adaptive parallele {M}ultilevel-{M}ethoden zur {L}\"{o}sung elliptischer {R}andwertprobleme}, school = {Mathematisches Institut, TU M\"{u}nchen}, address = {Germany}, year = {1992}, type = {Diplomarbeit}, } @TechReport{GWZumbusch_1993a, author = {G. W. Zumbusch}, title = {Symmetric Hierarchical Polynomials for the h-p-Version of Finite Elements}, institution = {Konrad-Zuse-Zentrum}, address = {Berlin, Germany}, year = {1993}, number = {SC-93-32}, } @TechReport{GWZumbusch_1994a, author = {G. W. Zumbusch}, title = {Visualizing Functions of the h-p-version of finite elements}, institution = {Konrad-Zuse-Zentrum}, address = {Berlin, Germany}, year = {1994}, number = {TR-94-05}, } @InProceedings{GWZumbusch_1995a, author = {G. W. Zumbusch}, title = {Adaptive h-p approximation procedures, graded meshes and anisotropic refinement for Numerical Quadrature}, booktitle = {Proceedings of The First European Conference on Numerical Mathematics and Advanced Applications, ENUMATH 95}, editor = {F. Brezzi and J. Periaux and R. Glowinski and R. Rannacher and Yu. Kuznetsov}, year = {1995}, } @PhdThesis{GWZumbusch_1995b, author = {G. W. Zumbusch}, title = {Simultanous h-p Adaptation in Multilevel Finite Elements}, school = {Fachbereich Mathematik und Informatik, FU Berlin}, address = {Germany}, year = {1995}, note = {published by Shaker, Aachen, 1996}, } @Article{GWZumbusch_1996a, author = {G. W. Zumbusch}, title = {Symmetric Hierarchical Polynomials and the Adaptive h-p-Version}, journal = {Houston Journal of Mathematics}, year = {1996}, pages = {529--540}, note = {Proceedings of the Third International Conference on Spectral and High Order Methods, ICOSAHOM'95}, } @TechReport{CSchutte_MDinand_GWZumbusch_RBrinkmann_1995a, author = {Ch. Sch\"{u}tte and M. Dinand and G. W. Zumbusch and R. Brinkmann}, title = {Dynamics of {E}rbium-doped Waveguide Lasers: Modelling, Reliable Simulation, and Comparison with Experiments}, institution = {Konrad-Zuse-Zentrum}, address = {Berlin, Germany}, year = {1995 }, number = {SC-95-24}, } @InProceedings{MKorzen_RSchriever_KUZiener_OPaetsch_GWZumbusch_1996a, author = {M. Korzen and R. Schriever and K.-U. Ziener and O. Paetsch and G. W. Zumbusch}, title = {Real-Time 3-D Visualization of Surface Temperature Fields Measured by Thermocouples on Steel Structures in Fire Engineering}, booktitle = {Oroceedings of International Symposium Local Strain and Temperature Measurements in Non-Uniform Fields at Elevated Temperatures}, pages = {253--262}, year = {1996}, editor = {J. Ziebs and J. Bressers and H. Frenz and D. R. Hayhurst and H. Klingelh\"{o}ffer and S. Forest}, publisher = {Woodhead Pub}, } @TechReport{GWZumbusch_1996b, author = {G. W. Zumbusch}, title = {Multigrid methods in {D}iffpack}, institution = {Sintef Applied Mathematics}, year = {1996}, number = {STF42 F96016}, address = {Oslo, Norway}, } @TechReport{GWZumbusch_1996c, author = {G. W. Zumbusch}, title = {Schur Complement Domain Decomposition Methods in {D}iffpack}, institution = {Sintef Applied Mathematics}, year = {1996}, address = {Oslo, Norway}, } @TechReport{GWZumbusch_1996d, author = {G. W. Zumbusch}, title = {Overlapping Domain Decomposition Methods in {D}iffpack}, institution = {Sintef Applied Mathematics}, year = {1996}, address = {Oslo, Norway}, } @InProceedings{AMBruaset_HPLangtangen_GWZumbusch_1998a, author = {A. M. Bruaset and H. P. Langtangen and G. W. Zumbusch}, title = {Domain Decomposition and Multilevel Methods in {D}iffpack}, booktitle = {Proceedings of Domain Decomposition Methods 9, DD9}, pages = {655--662}, year = {1998}, editor = {P. E. Bj{\o}rstad and M. S. Espedal and D. E. Keyes}, publisher = {Domain Decomposition Press}, address = {Bergen, Norway}, note = {also as report STF42 F96017, Sintef Applied Mathematics, Oslo, 1996 }, } @InProceedings{MGriebel_GWZumbusch_1997a, author = {M. Griebel and G. W. Zumbusch}, title = {Parnass: Porting gigabit-{LAN} components to a workstation cluster}, booktitle = {Proceedings of the 1st Workshop Cluster-Computing}, pages = {101--124}, year = {1997}, editor = {W. Rehm}, number = {CSR-97-05}, series = {Chemnitzer Informatik Berichte}, organization = {TU Chemnitzer}, } @InProceedings{MGriebel_GWZumbusch_1997b, author = {M. Griebel and G. W. Zumbusch}, title = {Parallel multigrid in an adaptive {PDE} solver based on hashing}, booktitle = {Proceedings of ParCo '97}, pages = {589--599}, editor = {E. D'Hollander and G.R. Joubert and F.J. Peters and U. Trottenberg}, publisher = {Elsevier}, year = {1997}, } @InProceedings{MGriebel_GWZumbusch_1998a, author = {M. Griebel and G. W. Zumbusch}, title = {Hash-Storage Techniques for Adaptive Multilevel Solvers and their Domain Decomposition Parallelization}, booktitle = {Proceedings of Domain Decomposition Methods 10, DD10}, pages = {279--286}, year = {1998}, editor = {J. Mandel and C. Farhat and X.-C. Cai}, number = {218}, series = {Contemporary Mathematics}, publisher = {AMS}, } @InProceedings{MGriebel_GWZumbusch_1998b, author = {M. Griebel and G. W. Zumbusch}, title = {Adaptive Sparse Grids for Hyperbolic Conservation Laws}, booktitle = {Proceedings of Seventh International Conference on Hyperbolic Problems, Theory, Numerics, Applications}, year = {1998}, publisher = {Birkh\"{a}user}, } @InProceedings{TSchiekofer_GWZumbusch_1999a, author = {T. Schiekofer and G. W. Zumbusch}, title = {Software Concepts of a Sparse Grid Finite Difference Code}, booktitle = {Proceedings of the 14th GAMM-Seminar Kiel on Concepts of Numerical Software}, year = {1999}, editor = {W. Hackbusch and G. Wittum}, series = {Notes on Numerical Fluid Mechanics}, publisher = {Vieweg}, } @InProceedings{GWZumbusch_1999a, author = {G. W. Zumbusch}, title = {Dynamic loadbalancing in a lightweight adaptive parallel multigrid PDE solver}, booktitle = {Proceedings of the 9th SIAM Conference on Parallel Processing for Scientific Computing (PP99)}, year = {1999}, publisher = {SIAM}, } @InProceedings{MASchweitzer_GWZumbusch_MGriebel_1999a, author = {M. A. Schweitzer and G. W. Zumbusch and M. Griebel}, title = {Parnass2: A Cluster of Dual-Processor {PC}s}, booktitle = {Proceedings of the 2nd Workshop Cluster-Computing}, year = {1999}, editor = {W. Rehm and T. Ungerer}, number = {CSR-99-02}, series = {Chemnitzer Informatik Berichte}, organization = {TU Chemnitzer}, } Editor's Note: These will be reformatted and entered into the bibliography; ------------- see http://www.mgnet.org/mgnet-bib.html soon. ------------------------------ End of MGNet Digest **************************