Conference nicctr::dxml

> What math_lib are available on Alpha/UNIX?

  DXML supports LAPACK.

> How can I get them and detail information?

  Typing:

  man lapack

  at the prompt, on a system where DXML is installed, is good way to
  get started. :-)

    Thank you Dwight-san,
    
    I found  some eigen value solver for dense and triangle matrix.
    Is there math_lib which support eigen value solver of sparce matrix?
    
    Regards
    mamoru
    

Hi Nakano-san,

>    Is there math_lib which support eigen value solver of sparce matrix?

Yes, but not in DXML. I suggest searching netlib or, better yet, contacting
Susanne Balle. I don't know her e-mail address, but sending mail to her at
[email protected] should work. Susanne works in John Kreatsoulas's
expertise center and is an externally recognized expert on this very topic.

Best Regards,

	- Dwight -

    Customer,EPSON R&D found out NCSLIB which can handle very high density
    matrix from spare matrix.
    (their simulation model is very large. If they take skyline method,the
    size of their simulation model require over 10GB memory.)
    BCSLIB is supporting SGI,CRAY and sun.
    But EPSON would like to use Alpha.
    
    The conatct person is  
    
  ��Mr.��Roger��Grimes,��BCSLIB��developer���at boeing o

>>During��the one hour meeting, he told us that they are working on
>>a parallel��version of the program, which will support hundreds
>> of processors by the end of the year.
>>
>> On the computer platform issue, he said that currently they are
>>supporting Cray, SGI, and Sun.  Digital and Fujitsu are not supported
>>at this moment.
>
  �������ô��Ԥϡ�Ji��Wang��EPSON��Palo��Alto��Lab. �Ǥ��������ߺ���
�Ƥ���ץ���
���BCSLIB����Ȥ��Ƥ��ꡢ���Ū��SGI��SUN���Ȥ����äˤʤäƤ��ޤ���

����������ꤤ���ޤ���
						���桡��
						�����������ץ���ʳ��
						��ư�ҳ�ȯ��





% =================== Internet headers and postmarks ===================
% Received: by decjpn.tko.dec.com id AA19807; Thu, 8 May 97 22:53:22 +0900
% Received: from epsongw2.epson.co.jp (epsongw2.epson.co.jp [202.248.17.20]) by jnet-gw-2.dec-j.co.jp (8.8.3+DEC-J/JNET-GW-970311.1) with SMTP id WAA19033; Thu, 8 May 1997 22:59:37 +0900 (JST)
% Received: from epsongw.epson.co.jp by epsongw2.epson.co.jp (8.6.12+2.4W/3.4W-MX950731-Epson Mail Gateway) id WAA08546; Thu, 8 May 1997 22:54:01 +0900
% Received: from swapex1.exc.epson.co.jp by epsongw.epson.co.jp (8.6.12+2.4W/3.4W-97040115) id WAA05222; Thu, 8 May 1997 22:48:50 +0900
% Received: by swapex1.exc.epson.co.jp with SMTP (Microsoft Exchange Server Internet Mail Connector Version 4.0.995.52) id <[email protected]>; Thu, 8 May 1997 22:47:12 +090
% Message-Id: <c=US%a=_%p=EPSON%[email protected]>
% From: Imai Tsutomu <[email protected]>
% To: "'[email protected]'" <tkov50::nakano>
% Subject: Boeing's Library                             
% Date: Thu, 8 May 1997 22:47:11 +0900
% X-Mailer:  Microsoft Exchange Server Internet Mail Connector Version 4.0.995.52
% Mime-Version: 1.0
% Content-Type: text/plain; charset="iso-2022-jp"
% Content-Transfer-Encoding: 7bit

    Nakano-san,

>   Customer,EPSON R&D found out NCSLIB which can handle very high density
>   matrix from spare matrix.
>   (their simulation model is very large. If they take skyline method,the
>   size of their simulation model require over 10GB memory.)
>   BCSLIB is supporting SGI,CRAY and sun.
>   But EPSON would like to use Alpha.


    For processing skyline matrices of this size, you should consider the
    Fast Matrix Solver (FMSlib) package distributed by Dr. Ronald Young of
    Multipath Corporation. Dr. Young has been a Digital ISV for many years.
    His product is of very high quality and is well optimized for Alpha
    platforms. FMSlib has some eigenvalue capability, probably enough to
    meet your EPSON's needs. Please find information about fmslib on the
    web at:

	http://www.fmslib.com
   
    or contact Dr. Young at

	Dr. Ronald Young
	Multipath Corporation
	P.O. Box 8210
	Incline Village, NV. U.S.A.
	89452-8210

	Tel. 702-831-4400
	Fax  702-831-4401

	E-Mail	[email protected]

    As far as I know, we have no business relationship with BCSLIB. Sorry.
    However, because FMSlib specializes in very large out of core linear
    systems, I think Boeing may even be an FMSlib customer.

    BTW - Dr. Young and Randy Doering of the Washington D.C. recently solved
    a 100,000 X 100,1000 dense complex linear system on a very large 8400
    system at the Washington, D.C. benchmark center. Either he or Randy will
    be very happy to talk about it.

    Good Luck.

    - Dwight -