krj@na.toronto.edu (Ken Jackson) (03/11/89)
From: nagmag%ukc.ac.uk@NSS.Cs.Ucl.AC.UK
Date: Fri, 10 Mar 89 10:09:10 EST
Subject: NAgMAg V89 #4
Source-Info: From (or Sender) name not authenticated.
Sender: trh%ukc.ac.uk@NSS.Cs.Ucl.AC.UK
NAgMAg Friday, March 10 1989 Volume 89 Issue 4
Today's Topics
Linear Programming - a new software announcement
Re: GAMS Classification Scheme
Nested calls to D01AJF
Reply: NAG Graphical Supplement for NOS/VE
Convective Boundary problems
Toolpack Mailing list
Problem with D02EBF
%% Just a couple of things before you get to the interesting
%% stuff. Thanks to NAG staff for answering queries but please
%% would you send a copy of the reply to nagmag as well
%% as the original poster of the query just
%% in case other readers are interested in the reply.
%%
%% I'll also keep a list of unanswered (via nagmag) queries
%% here -- just as a gentle reminder:-)
%% If individuals have had a reply from NAG please forward
%% them to me for inclusion in the digest.
%%
%% UNANSWERED QUERIES
%% ==================
%%
%% V89 #3 Online Workstation Help System
%% V89 #3 GAMS Scheme
---------------------------
Date: 2-MAR-1989 14:57:53 GMT
From: DAVID@vax.nag.co.uk
Subject: Linear Programming - a new software announcement
NAGUA members may be interested to hear of the latest software
announcement from NAG. Below please find an extract from the first
draft of an information note on FORTLP. Within the next few weeks an
information pack will be posted to the 50 or so sites who have
already expressed a keen interest in the software. If you would like
to be included in this mailshot please contact David Sayers at NAG.
FORTLP Linear Programming Package - Information Note
Introduction.
The NAG Library contains excellent routines for the solution of
dense linear programming problems. It also has an instructive
routine to solve small integer linear programming problems. For many
applications however the problem size becomes very large and the NAG
routines are no longer the most appropriate. Firstly because sparse
techniques are not employed and secondly because the user-interface
demands industry-standard data formats.
This gap in the NAG problem solving environment has now been
admirably filled by FORTLP, a flexible package, originating from
Brunel University and which is now available from NAG.
The FORTLP Product
Nag is offering the FORTLP software in its Fortran source
form on a variety of media. It has been successfully tested and
installed on a number of machine ranges including IBM PC, DEC Vax
and Cray systems. The system is therefore suitable for all classes
of machines ranging from PC to supercomputer.
The only additional requirement of user site is the availability of
a Fortran compiler capable of accepting INTEGER*2 constructs. (This
is outside the Fortran standard but is commonly available with most
compilers and is exploited by the software to enable it to handle
very large problems with maximum efficiency.)
Uniquely amongst the NAG offerings FORTLP rests between a package,
which solves a users problem given the data, and a collection of
subroutines to enable the user to tailor the solution to his own
particular requirements. The installation guide explains both uses
and the accompanying manual gives considerable internal detail on
the working of the system.
The Support Service
An optional support service is offered and this will entitle the
user to periodic updates as the system continues to expand and
develop. For example the next release, planned for summer 1989, will
include the Forrest and Tomlin LU update for even better
preservation of sparsity.
We believe that the high quality research into algorithms and
software and prospects for further technological breakthroughs, such
as the full exploitation of the sparse BLAS concept, make the
Support Service an attractive proposition for most sites.
March 1989
---------------------------
Date: Fri, 3 Mar 89 14:26 GMT
From: FRIEDLI@czheth5a.earn
Subject: Re: GAMS Classification Scheme
In response to Tim's inquiry (V89 #2): I was in contact with David
Kahaner recently and he informed me that they are in the
process of editing a new version of GAMS which will
include the ACM algorithms and other new things as well.
The new volume will appear either this summer or towards
the end of the year.
Armin Friedli, Friedli@czheth5a.BITNET, ETH Zurich.
---------------------------
Date: 6-MAR-1989 14:06:36 GMT
From: PHYHJR@vaxb.heriot-watt.ac.uk
Subject: Nested calls to D01AJF
We had a rather nasty experience using D01AJF recently. The problem
took a long time to track down and seems to be due to the use of
the common block AD01AJ. It appears that calling the routine from
within its integrand function does not work. (the integrand contains
another integral).
In the documentation under Further Comments there is a note
`Labelled COMMON block AD01AJ is used by this routine and must therefor
e
be avoided by the user',
perhaps this should be amended to indicate that recursive calls to
the routine will not work.
Of course such calls may not be standard f77 but many compilers
support this.
Harvey Richardson
e-mail : JANET - phyhjr@uk.ac.hw.clust
INTERNET - phyhjr%clust.hw.ac.uk@cunyvm.cuny.edu
BITNET - phyhjr%clust.hw.ac.uk@ukacrl.bitnet
mail : Physics Department
Heriot-Watt University
Riccarton, Edinburgh, EH14 4AS, UK.
phone : (+44) 031-451-3047
---------------------------
Date: 08 MAR 89 12:18:15
From: Z3000PA@awituw01.earn
Subject: Reply: NAG Graphical Supplement for NOS/VE
Dear Mr Partl
I understand from Guenter Baszenski that the above will be available in
about 1 month with the GKS interface as standard.
Yours sincerely
Tim Nash
---------------------------
Date: Thu, 09 Mar 89 10:25:41 GMT
From: "Chan Basaruddin <061 720 7263>" <MBBGPTB@cms.umrcc.ac.uk>
Subject: Convective Boundary problems
I'm solving a parabolic problem in 2-D (which is linear), but
it has got a special boundary conditions i.e of hyperbolic type:
a (du/dt) + b (du/dx) + (du/dy) = 0.
Which nag routine should I employ ?
Chan
Manchester
---------------------------
Date: Fri, 10 Mar 89 13:35:55 GMT
From: Tim Hopkins <trh@uk.ac.ukc>
Subject: Toolpack Mailing list
In case any of you missed to announcement on the news recently,
there is an electronic mailing list been set up for discussing
Toolpack. It looks as though there are people out there using
it at last and thinking about how to improve it.
Anyone interested e-mail
dph@lanl.gov
If you have problems getting through -- mail me and I'll post
them on.
Tim.
-----------------------------
From: "Alphonse P. Magnus" <ANNU76@buclln11.earn>
Subject: Problem with D02EBF
To: T Hopkins <nagmag-request@ukc.ac.uk>
The following ODE solver went into a loop on NAG Mark12 on the
IBM 4381 (VM) of Louvain-la-Neuve (we just received Mark13,
not yet installed, perhaps the problem has been settled)
PROGRAM NGTD02
C test D02EBF
REAL*8 T,T0,TEND,Y(2),TOL,W(222)
COMMON T0,TEND,N
EXTERNAL FC,CJAC,OUT1,OUT2
N=1
C
TOL=1.0D-8
T0=0
TEND=1
Y(1)=1
PRINT '('' t0,tend : '',2F15.10)' , T0,TEND
IR=0
NJAC=1
T=T0
IFAIL=111
CALL D02EBF(T,TEND,N,Y,TOL,IR,FC,NJAC,CJAC,OUT1,W,222,IFAIL)
PRINT *,' ifail=',IFAIL
C
TOL=1.0D-8
T0=0
TEND=1
Y(1)=1
PRINT '('' t0,tend : '',2F15.10)' , T0,TEND
IR=0
NJAC=1
T=T0
IFAIL=111
CALL D02EBF(T,TEND,N,Y,TOL,IR,FC,NJAC,CJAC,OUT2,W,222,IFAIL)
PRINT *,' ifail=',IFAIL
STOP
END
SUBROUTINE FC(T,Y,F)
REAL*8 T,T0,TEND,Y(N),F(N)
COMMON T0,TEND,N
C ridiculously simple exemple y'=y
F(1)=Y(1)
PRINT '('' t : '',F20.15,'' y:'',1PE14.7)',T,Y
RETURN
END
SUBROUTINE CJAC(T,Y,JAC)
REAL*8 T,T0,TEND,Y(N),JAC(N,N)
COMMON T0,TEND,N
JAC(1,1)=1
RETURN
END
SUBROUTINE OUT1(TSOL,Y)
REAL*8 TSOL,T0,TEND,Y(N)
COMMON T0,TEND,N
PRINT '('' tsol: '',F20.15,'' y:'',1PE14.7)',TSOL,Y
TSOL=TSOL+0.1*(TEND-T0)
IF( ABS( (TSOL-TEND)/(T0-TEND) ) .LT.1.0D-5) TSOL=TEND
C This one works fine
RETURN
END
SUBROUTINE OUT2(TSOL,Y)
REAL*8 TSOL,T0,TEND,Y(N)
COMMON T0,TEND,N
PRINT '('' tsol: '',F20.15,'' y:'',1PE14.7)',TSOL,Y
TSOL=TSOL+0.1*(TEND-T0)
IF( TSOL.GT.TEND ) TSOL=TEND
C should stop here, but runs into a loop
RETURN
END
%% A quick run on the unix version gives the same results.
%% Tim
---------------------------------------
%%
%% Replies or submissions to nagmag@uk.ac.ukc
%% Distribution changes to nagmag-request@uk.ac.ukc
%%
%% END OF ISSUE
--------------------------------------
Reposted by
--
Kenneth R. Jackson, krj@na.toronto.edu (on Internet, CSNet,
Computer Science Dept., ARPAnet, BITNET)
University of Toronto, krj@na.utoronto.ca (CDNnet and other
Toronto, Canada M5S 1A4 X.400 nets (Europe))
(Phone: 416-978-7075) ...!{uunet,pyramid,watmath,ubc-cs}!utai!krj