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:    casper.cs.yale.edu (128.36.12.1)

World Wide Web:  http://na.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.ccs.uky.edu/mgnet

Today's editor:  Craig Douglas (douglas-craig@cs.yale.edu)

Volume 7, Number 7 (approximately July 31, 1997)

Today's topics:

     Changes at Yale / Missing E-mail to MGNet
     Mirroring of MGNet (Experimental Mirror in Japan)
     CV Listing Available
     Paper on a Multi-Level Method for Sparse Linear Systems
     Kaskade 3.1 code
     Publist Oosterlee
     Benchmarking Update
     Conference in Xi'an

*****************************************************************************
***** August is traditionally a slow month.  Please send contributions. *****
*****************************************************************************

-------------------------------------------------------

Date: Thu, 31 Jul 97 10:12:45 -0400
From: Craig Douglas 
Subject: Changes at Yale / Missing E-mail to MGNet

Yale's computer science department is in the process of retiring a number of
its file servers.  A complete reorganization and house cleaning is in
progress.  This is the first really significant change in the Yale CS machine
structure in many, many years.  Things are not going as smoothly as had been
originally hoped for.

In particular, casper.cs.yale.edu (or na.cs.yale.edu) is one of the servers
that is being retired.  It will be turned off sometime between the end of this
week and the end of the month.  I expect that the day that casper is turned
off, that MGNet will be missing in the new environment until certain new links
are set up.  Hopefully, casper will not be turned off while I am in China,
where I will not be able to check on the progress.

When everything is done, there should be aliases for the old machine names.
In the meantime, you might want to use one of the mirror sites.  If you want
to upload a file, please use ultra.cs.yale.edu instead of casper.cs.yale.edu.

I am currently making all changes to MGNet on the http://ccs.uky.edu/mgnet
site first since it will not be affected by the upheaval at Yale.

There is one message in this digest that was supposed to be in the May issue.
I have had one too many e-mailboxes since January.  If you submit something
for the digest and do not get a confirmation, please try again.  If you think
I missed a message, please let me know.  This is particularly true for this
month's issue.  Somewhere in the machine reorganization at Yale, files with
all of my Yale e-mail from June 7 - July 16 disappeared.  Hence, there might
be at least one message missing from this digest.

-------------------------------------------------------

Date: Tue, 29 Jul 1997 10:00:16 +0900
From: Akira Nishida 
Subject: Mirroring of MGNet (Experimental Mirror in Japan) 

Dear Prof. Douglas,

I am not sure if it arrives at the correct address.  If you have already read,
please disregard this message.  We are working on the Phase (Parallel and
High-performance computing Research Activities in the Electrotechnical
Laboratory, Japan) project.  The understanding of multigrid method is being
deepened also in Japan recently.  With your approval, we should like to mirror

    ftp://casper.cs.yale.edu/mgnet
    and
    http://casper.cs.yale.edu/mgnet/www

on our archive at

    ftp/http://phase.etl.go.jp

officially, and provide this repository for use in studies of high performance
computing in Japan and Asia- Pacific region.  We will quote the source of the
repository and give proper credit.

We look forward to receiving your permission and thank you in advance.

Yours faithfully,
Akira Nishida

NISHIDA Akira                       
Department of Information Science, Faculty of Science, 
the University of Tokyo
E-mail: nishida@is.s.u-tokyo.ac.jp

    Editor's Note: This is working.  It is mirroring casper until the new
    -------------  machine takes over.  Note that this is experimental and
                   could disappear.  I checked the site and found the transfer
                   rate impressive.  There are other neat sites mirrored there.
                   If you are in the Pacific rim, I highly recommend this
                   site for MGNet access.

-------------------------------------------------------

Date: Thu, 31 Jul 97 10:17:23 -0400
From: Craig Douglas 
Subject: CV Listing Available

I have received a number of curriculeum vitaes in the last few months.  In the
last month several of these people's advisors have contacted me to see if I
can find jobs for these people.  If you want to list your CV on MGNet, I will
be happy to list it as long as you are doing multigrid or domain decomposition
related work.  I take plain text, LaTeX, html, PostScript, and PDF formats.

-------------------------------------------------------

Date: Fri, 15 Jul 97 09:05:15 -0600
From: Yair Shapira 
Subject: Paper on a Multi-Level Method for Sparse Linear Systems

                A Multi-Level Method for Sparse Linear Systems

                                 Yair Shapira
                        Los Alamos National Laboratory
                               Mail Stop B-256
                             Los Alamos, NM 87545

                                   Abstract

A multi-level method for the solution of sparse linear systems is introduced.
The definition of the method is based on data from the coefficient matrix
alone.  An upper bound for the condition number is available for certain SPD
problems.  Numerical experiments confirm the analysis and illustrate the
efficiency of the method for diffusion problems with discontinuous
coefficients for which the discontinuities are not aligned with the coarse
meshes.

    Editor's Note: in mgnet/papers/Shapira/amg.ps.gz
    -------------

-------------------------------------------------------

Date: Thu, 31 Jul 1997 11:15:37 +0200
From: Erdmann 
Subject: Kaskade 3.1 code

Please, be so kind to rename the actual directory mgnet/Codes/kaskade/3.0 to
mgnet/Codes/kaskade/3.x and add the files

     856496 Jul 31 09:36 3.1.1.tgz
       5746 Jul 31 09:39 README-3.x
      88324 Jul 31 09:39 programmer_guide.ps.gz
    1271964 Jul 31 09:44 user_guide.ps.gz

Bodo Erdmann
Konrad-Zuse-Zentrum fuer Informationstechnik (ZIB)
e-mail: erdmann@zib.de        | Abt. Scientific Software
Telefon: (030) 84185 - 315    | Takustr. 7
Fax:     (030) 84185 - 125    | D-14195 Berlin - Dahlem

    Editor's Note: This replaces version 3.0.
    -------------

-------------------------------------------------------

Date: Thu, 15 May 97 11:18:16 +0200
Subject: Publist Oosterlee
From: Kees Oosterlee 

So here is an overview of some of my papers over the last 3 years.

    Editor's Note: I have too many e-mailboxes (about to be cured).  This
    -------------  got lost in them.  The entries will be reformatted and
                   entered in the database.  Sorry about the delay...

@ARTICLE{Oo-We94-1,
         AUTHOR="{C.W. Oosterlee and P. Wesseling}"
        ,TITLE="{Steady incompressible flow around objects in general
                 coordinates with a multigrid solution method}"
        ,JOURNAL="Num. Methods for Part. Diff. Equations"
        ,volume=10
        ,pages={295-308}
        ,YEAR=1994
}

@ARTICLE{Oo-We95-1,
         AUTHOR="{C.W. Oosterlee and P. Wesseling}"
        ,TITLE="{On the robustness of a Multiple Semi-coarsened Grid method}"
        ,JOURNAL="Zeitschrift Angew. Math. und Mech."
        ,volume=75
        ,pages={251-257}
        ,YEAR=1995
}

@ARTICLE{,
         AUTHOR="C.W. Oosterlee and H. Ritzdorf"
        ,TITLE="{Flux difference splitting for three-dimensional steady
                 incompressible Navier-Stokes equations in
                 curvilinear coordinates}"
        ,JOURNAL="International Journal for Numerical Methods in Fluids"
        ,VOLUME= "23"
        ,YEAR="1996"
        ,PAGES="347-366"
}

@ARTICLE{,
         AUTHOR="R. Hempel and R. Calkin and R. Hess and W. Joppich and
                 C.W. Oosterlee and H. Ritzdorf and P. Wypior and W. Ziegler
                 and N. Koike and T. Washio and U. Keller"
        ,TITLE="{Real applications on the new parallel system NEC Cenju-3}"
        ,JOURNAL="Parallel Computing"
        ,VOLUME= "22"
        ,YEAR="1996"
        ,PAGES="131--148"
}

@ARTICLE{Oos95-1,
         AUTHOR="C. W. Oosterlee"
        ,TITLE="{The convergence of parallel multiblock multigrid methods}"
        ,JOURNAL="Appl. Numerical Mathematics"
        ,volume=19
        ,pages={115-128}
        ,YEAR=1995
}

@ARTICLE{Oos97-1,
         AUTHOR="C. W. Oosterlee"
        ,TITLE="{A GMRES-based plane smoother in multigrid to solve 3D anisotropic 
                 fluid flow problems}"
        ,JOURNAL="Journal of Comp. Physics"
        ,volume=130
        ,pages={41-53}
        ,YEAR=1997
}

@inproceedings{Oo_Ri_Sc_St_94_3,
  author = "C. W. Oosterlee and  H. Ritzdorf and  A. Sch{\"u}ller
             and B. Steckel",
  title = "Experience with a parallel multiblock multigrid
           solution technique for the {Euler} equations" ,
  booktitle = "{Proceedings of the 10th GAMM Seminar Kiel
               `Fast solvers for flow problems'}" ,
  editor = "W. Hackbusch and G. Wittum",
  address = "Kiel" ,
  year = "1995",
  publisher = "Vieweg, Braunschweig",
  pages = "192--203",
  note = "Ser. Notes on Num. Fluid Mech. {\bf 49}"
}

@inproceedings{Oos96R,
  author = "C.W. Oosterlee and H. Ritzdorf",
  title  = "{A robust parallel solver for 3D fluid flow problems using
             a high-level communications library}",
  BOOKTITLE="{Parallel Computing:
                     State-of-the Art Perspective (PARCO95)}"
  ,YEAR="1996"
  ,EDITOR="{D'Hollander, Joubert, Peters, Tystram}"
  ,PAGES={77--84}
  ,ORGANIZATION={PARCO95}
}

@inproceedings{Oos96RBE,
  author = "C.W. Oosterlee and H. Ritzdorf and H.M. Bleecke and B. Eisfeld",
  title  = "{Benchmarking the FLOWer code on different parallel and
             vector machines}",
  BOOKTITLE="{Parallel Computational Fluid Dynamics:
              Implementations and Results Using Parallel Computers}"
  ,YEAR="1996"
  ,EDITOR="{A. Ecer, J. P{\'e}riaux, N. Satofuka and S. Taylor}"
  ,PAGES={281--288}
  ,ORGANIZATION={Elsevier, Amsterdam}
}

-------------------------------------------------------

Date: Thu, 31 Jul 97 10:17:23 -0400
From: Craig Douglas 
Subject: Benchmarking Update

Many of yo do not realize this, but the MGNet mailing list was started
originally to discuss benchmarking efforts for multigrid.  I am not sure
that any real messages have ever been put through MGNet on this topic.
The benchmarking mailing lists are up.  So far, all of the messages have
been of the form, "I would like to be on the mailing list for..."

One effort is to provide SPEC with some large applications for
benchmarking.  If you have any possibilities, please send e-mail to
mgnetbc@ccs.uky.edu if you have an ideas that you are willing to share.

Another of the efforts requires defining a problem, its solution, and a list of
what multigrid solvers can solve this problem.  In order to do this, a problem
description needs to be developed.  There are several existing formats that
can be studied (e.g., the sparse matrix field has a number of repositories of
problems).  Please send e-mail to mgnetb@ccs.uky.edu if you have an ideas that
you are willing to share.

To start this off, consider the following inappropriate boundary value
problem:

    - ( a(x) u  )  - b(x) u  + c(x) u = f(x)
              x  x         x

        alpha(0) u (0) + beta(0) u(0) = g(0)
                  x

        alpha(1) u (1) + beta(1) u(1) = g(0)
                  x

Information that would be needed to store a problem in a database would
include the following information:

    1.  Discretization method
    4.  Number of levels
    2.  Information about grids
    3.  Suggestions for interpolation and multilevel algorithms
    4.  Coefficient data or formulas
    5.  Right hand side (data or formulas)
    6.  Solution (data or formulas for the finest or all levels)
    7.  How the solution was derived

Item 3 could be considered optional since the point is to come up with
algorithm independent problems.  The problem definition needs to work on 1 or
more processors, too.

While the toy problem is usually solved on uniform grids, this should not be
assumed to be the case when defining a problem definition.  Plenty of problems
have to use nonuniform, overlapping, and/or nonnested grids.

Here are some questions that I would like to see discussed on the
mgnetb@ccs.uky.edu mailing list.

    What else is needed for this toy problem?

    Should something be dropped from the description?

    Should we assume that all computers are IEEE compliant so that binary data
    can be transferred at a much lower expense than ASCII?  (Should both be
    available?)

    How should the description be stored so that it is easily downloaded via
    the web?

-------------------------------------------------------

Date: Tue, 22 Jul 1997 10:44:07 -0500
From: Zhang-xin CHEN 
Subject: Conference on Bifurcation Theory and its Numerical Analysis

             First Announcement and Call for Papers

               Second International Conference on
           Bifurcation Theory and its Numerical Analysis

                June 29--July 3, 1998
                Xi'an Jiaotong University
                Xi'an, P.R.CHINA

                Hosted by
                Research Center for Applied Math.
                Xi'an Jiaotong University
                P.R.CHINA
Conference Chairs
        Honorary Chairman
        K. C. Chang, Honorary Chairman (Peking University, CHINA)
        S. N. Chow, Chairman (Georgia Institute of Technology, USA)
        Kaitai Li, Co-Chairman (Xi'an Jiaotong University, CHINA)
Scientific Committee
        P. Chossat      CNRS-Universite de Nice, FRANCE
        S.N. Chow       Georgia Institute of Technology, USA
        M. Golubitsky   University of Houston, USA
        K. Kirchgassner Universitaet Stuttgart, GERMANY
        W.E. Lanford    University of Guelph, CANADA
        K.T. Li         Xi'an Jiaotong University, CHINA
        Y. Li           University of Iowa, USA
        Z. Mei          Universitaet Marburg, GERMANY
        W.M. Ni         University of Minnesota, USA
        P.H. Rabinowitz University of Wisconsin, USA
        George Reddien  Southern Methodist University, USA
        George Sell     University of Minnesota, USA
        A. Spence       University of Bath, U. K.
        Yingfei Yi      Geogia Institute of Technology, USA
Organization Committee
        A.X. Huang, Chair, Xi'an Jiaotong University, CHINA
        Julian Wu, Vice Chair, U.S. Army Research Office (ARO), USA
        Z.Q. Xu, Vice Chair, NSF, CHINA
        Zhangxin Chen, Southern Methodist University, USA
        S.C. Hu, Southwest Missouri State University, USA
        Y.C. Ma, Xi'an Jiaotong University, CHINA
        Jun Zou, Chinese University of Hong Kong, CHINA
Objectives
    The aim of this conference is to bring together active researchers
with different backgrounds to discuss recent and prospective advances
in the bifurcation theory and its numerical analysis.

History: The First International Conference on Bifurcation and Its Numerical
Analysis was held at Xi'an Jiaotong University in July, 1988.  The
international scientific committee consisted of M. Golubitsky, G. Iooss,
D.D. Joseph, K. Kirchgassner, K.T. Li, J. Marsden, J. Rappaz and Z.C. Shi.

Call for Papers
    Prospective authors are invited to submit 500 word abstracts of their
papers by Feb. 1, 1998 to:
                Dr. Hou Yanren
                Research Center for Applied Mathematics
                Xi'an Jiaotong University
                Xi'an, 710049, P.R. CHINA
                Email: ktli@xjtu.edu.cn
Location
    The conference will be hosted by Research Center of Applied Mathematics
at Xi'an Jiaotong University, Xi'an, China. The conference will take place
in the International Conference Hall at Xi'an Jiaotong University.

Accommodation and Tours
    It is planned to accommodate conference participants in the Jianguo Hotel
(US$ 50/night), the Empire Hotel (US$ 40/night) or the Foreign Guest House of
Xi'an Jiaotong University (US$ 25/night).

    Tours around Xi'an city, including the Qin Terra-cotta soldiers, the Big
Pagoda, the city wall of Xi'an, the Bell Tower, Huaqing hot spring and the
Shaanxi History Museum, will be arranged during the meeting. All above tours,
including the traffic fees and tickets, may cost US$ 50 or so.

Conference Registration
    The registration fees are US$ 250 for participants and US$ 150 for
students. The checks made payable and registration form written to:
                Professor Huang Aixiang
                Research Center for Applied Mathematics
                Xi'an Jiaotong University
                Xi'an, 710049, P.R. CHINA
Registration Form for BTNA'98
        June 29, 1998
        International Conference Hall
        Xi'an Jiaotong University, Xi'an, 710049, CHINA

        First Name      ___________________________________
        Last Name       ___________________________________
        Affiliation     ___________________________________
        Street Address  ___________________________________
        City/State/Zip  ___________________________________
        Country         ___________________________________
        E-mail Address  ___________________________________
        Fax             ___________________________________

        [] I'll be attending the conference, please find enclosed a check
           of US$ 250.
        [] I am interesting in attending the conference, please keep me
           informed.
        [] I will submit. Title of the paper is
           _______________________________________________________________
           _______________________________________________________________

------------------------------

End of MGNet Digest
**************************