krj@na.toronto.edu (Ken Jackson) (04/15/89)
From: nagmag%ukc.ac.uk@NSS.Cs.Ucl.AC.UK
Date: Fri, 14 Apr 89 10:43:20 EDT
Subject: NAgMAg V89 #7
Sender: trh%ukc.ac.uk@NSS.Cs.Ucl.AC.UK
NAgMAg Friday, April 14 1989 Volume 89 Issue 8
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%% The official electronic digest of the NAG Users Association %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Today's Topics
C Version of Nag Library / Nag help system
naghelp
New Naghelp
nag on-line
IBM CMS NAG library
Help Systems
Nag Help
Languages
NAG Fortran Library for parallel execution on Cray X-MPs
HELP systems
2-D interpolation
%%
%% It must be all the chocky eggs -- look articles galore. Many
%% thanks to all who contributed. The article from Mike on the planned
%% replacement help system provides ample excuse to come along to
%% York in September plus the opportunity to put forward your
%% ideas as to what's really wanted -- so fire your wish lists
%% to nagmag immediately!!
%%
%% The mailing list now contains 150 addresses -- I hope they are
%% all getting through. There are a couple of Bitnet sites whose
%% original requests were forwarded to me by our postmaster -- so
%% if your posted article isn't in this issue it may be taking a
%% little longer than usual to get to me!!
%%
%% Happy number crunching,
%% Tim
%%
%%
---------------------------
Date: Mon, 10 Apr 89 16:45:55 BST
From: Harvey Richardson <harvey@sun.physics.heriot-watt.ac.uk>
Subject: C Version of Nag Library / Nag help system
Well here is a contribution if you really need them that badly.
I suppose the majority of scientific programs are written in Fortran.
This may be due to an unwillingness to learn a new language but there
may be good reasons. Personally I stick with Fortran for the
following reasons:
1) There are major subroutine libraries available in Fortran. Nag
is one example. Also the many graphics packages.
2) Fortran (77) is standardised to a point that standard conforming
programs will run on any decent system. I am wary of using a
language (C) which appears to be loosely defined (K&R).
3) There is the promise of Fortran8X on the horizon if the wrangling
stops and a standard is ratified.
Point (1) will improve as other language libraries become available.
No doubt Numerical Recipes in C would not have been published if
there was insufficient demand.
Perhaps ANSI C will make point (2) irrelevant.
Anyway I cannot see the many `traditional' Fortran programmers
changing to C. Maybe there is demand from programmers on UNIX
systems who have more incentive to use C. Certainly the possibility
of argument checking is good news. (Of course Fortran8X will also
provide this facility).
I am not a C expert so the points above are only my impression and
should not be taken too seriously.
I only use the Nag Help system for checking IFAIL messages or looking
for simple functions like for example Bessel functions. To choose a
major routine from the help system alone is not feasible. A UNIX man
version would be nice. As would a hierarchical help system.
Another point on Nag Documentation. I would suggest an additional
section to the introduction. This could be a simple example
problem showing how the choice of routine was made, how it was
coded and tested. I think many new users find the Library a
daunting prospect and this would be useful.
One final point.
Should any Nag routine fail with an arithmetic overflow
error internally (not in a user supplied function). I have met this
a couple of times. I realise that to check all possible arithmetic
statements for such a possibility may be too much but wonder if there
is a policy on such matters.
Harvey Richardson
e-mail : JANET - harvey@uk.ac.hw.phy.s
- phyhjr@uk.ac.hw.clust
INTERNET - harvey%s.phy.hw.ac.uk@cunyvm.cuny.edu
BITNET - harvey%s.phy.hw.ac.uk@ukacrl.bitnet
mail : Physics Department
Heriot-Watt University
Riccarton, Edinburgh, EH14 4AS, UK.
phone : (+44) 031-451-3047
---------------------------
Date: Mon, 10 Apr 89 15:56:06 PDT
From: Jerry Berkman <jerry@violet.berkeley.edu>
Subject: naghelp
> 2) What sort of facilities would users like to see added to the NAG
> help system?
I want the NAG "help" facility to be so complete that users do not need
to ever look at a manual. Here at U.C. Berkeley, most users do not
have manuals. We put a copy in the library, but this is inconvenient.
Also, the pages for commonly used routines sometimes get ripped off.
- Jerry Berkman, U.C.Berkeley, jerry@violet.berkeley.edu
---------------------------
Date: Mon, 10 Apr 89 16:53:43 BST
From: MRO2@phoenix.cambridge.ac.uk
Subject: New Naghelp
Ok, Tim, I'll bite.
As you know, I am working on a replacement for the NAG
HELP system. This is very near to completion and some pilot
implementations should be out this summer.
Perhaps you would like to beta test it?
The new system will be loosely based on the keyword searching algorithms
of current NAG HELP but with rather more facilities, including
backtracking and context-sensitive break-handling.
Rather less obviously it will overcome some of the built-in restrictions
of the existing system and will be capable of expansion, both in terms
of the size and complexity of the database, and the user interface.
It will also be more portable than the existing system, being
more easily adaptable to different filing systems.
I will be talking about and demonstrating the new system at NAGUA
this year. Rather than go into any more detail, it might be interesting
to hear what facilities potential users would require. If these are not
provided or planned then they can still be considered for later versions.
---------------------------
Date: Mon, 10 Apr 1989 12:05:06 EDT
From: "Leigh, Anne" <ANNE@ukcc.earn>
Subject: nag on-line
Tim
I find the NAG-HELP facility very useful, and I recommend it to the
students. They can print off their own copy of the documentation
this way. I would like to see the examples added to the documentation,
so they could print those off also.
Anne
---------------------------
Date: Mon, 10 Apr 89 12:28:23 EDT
From: "Leigh, Anne" <ANNE@ukcc.earn>
Subject: IBM CMS NAG library
Tim,
I hate to bother you with this, but is the vector version of Mark13
available yet for the IBM CMS environment?
The reason I ask is we got our serial versions recently, but no
vector version.
A packing list was not included, so I don't know if we were supposed
to get the vector version and it was omitted inadvertently.
(We currently have a Mark12 vector version on our system).
Anne
---------------------------
Date: Tue, 11 Apr 89 08:22:08 bst
From: Joe Morice <MORICE@prl.philips.co.uk>
Subject: Help Systems
I have a few comments about on-line help systems:
We tinkered with the supplied Nag help files to incorporate them into
IBM's way of doing it under VM/CMS. It was not a success, nobody
(as far as I know) used it. My suspicion is that our users find
it easier to read things in a book or to come and ask.
On our Apollo systems, we have offered the SLATEC library (we did this
because there are no limitations on its distribution and use).
A help system, somewhat similar in style to what Tim described as
being supplied by IMSL, comes free with SLATEC so we installed it.
A few people have used it, we know this because they have come and
asked for help with it! But it is reasonably effective because the
Apollo's windowing system allows you to have the help,
your program and the results all on the screen at the same time.
My conclusions to all of this are as follows:
(1) On-line help is a very poor substitute for well-informed
personal help.
(2) Clearly written manuals are very useful.
(3) If effort is available to make libraries easier to use a more
effective way is via program generators eg PDE/Protran,
ELLPACK and Computer Associates Codebook (a graphics package).
Joe Morice
---------------------------
Date: 11-APR-1989 10:45:53 GMT
From: AEIC0456@v1.qub.ac.uk
Subject: Nag Help
Tim,
Don't get disheartened that not too many are firing lots of
messages over to you - not everything can be applicable to everybody!
You asked about help facilities in the latest bulletin: I haven't
tried the IMSL package but I have used the NAG Help. I have found
it very useful and have employed it for serious routine selection. On
many occasions a user seeking advice on how to approach a particular
problem has been pleasantly surprised by not being pointed at a manual,
but rather looking on as an Advisor peruses the NAG Help, stopping every
now and then to toggle a printer on and off so that the user can take
away a short listing of specific information to help him in his plans.
George Munroe, User Support, Computer Centre
The Queen's University of Belfast
Belfast BT7 1NN, N Ireland
---------------------------
Date: Tue, 11 Apr 89 16:22 BST
From: Bruce Stephens <STEPHENS@rmcs.cranfield.ac.uk>
Subject: Languages
I was interested to see that NAG are pushing some C products. This
suggests a few questions:
Fortran is not a remarkable language for writing quality, reusable
software, so should NAG be exploring other languages?
Of course they are; there is already an Ada language library, and C
seems to be supported to some extent. But are Ada and C
representative of the most appropriate languages for reliable
numerical software. I'm willing to believe that they are the ones
most likely to be wanted by people, but (and this is my real question)
should NAG not try to lead the market by choosing
a language that they know is a good one for the production of software?
Probably they'll argue that they can manage quite well with Fortran, and
they're probably right, but it seems to me that a better designed
language, such as C++, would make the writing of the library easier,
make the result more powerful, and make the library easier to use.
But should NAG be in the business of persuading its customers to
choose a particular language?
Bruce Stephens.
---------------------------
Date: Thu, 13 Apr 89 14:35:49 ITA
From: Valeria Frontini <AFT0@icineca.earn>
Subject: NAG Fortran Library for parallel execution on Cray X-MPs
AT CINECA, where a Cray X-MP/48 (four CPUs) is installed, we
compiled the NAG Fortran Library source code with the option
ALLOC=STACK (CFT compiler) that allows jobs executing in
parallel to access the NAG routines.
Till now we had about 55 accesses to the Library (unfortunately
without knowing which routines were called), and it seems the users
have no problems.
When will NAG provide an official (Cray) version?
Can they tell us their plans?
Are there any other users that had experiences about it?
We are now looking for which routines were used by our users,
and may be soon we will have more informations on this.
Valeria Frontini
---------------------------
Date: Thu, 13 Apr 89 12:29:08 bst
From: no-name <HELLON@prl.philips.co.uk>
Subject: HELP systems
I attended the recent IMSL user group Europe meeting (Amsterdam),
at which they demonstrated their Online Documentation Facility.
The package is very powerful and I was particularly impressed with
the power of the keyword search.
Boolean operations can be applied to lists of routine names
generated from keyword searches to help obtain specific routine
names. This would be an invaluable feature if manuals were not
readily available.
Craig Hellon
---------------------------
Date: Thu, 13 Apr 89 16:14:45 bst
From: Joe Morice <MORICE@prl.philips.co.uk>
Subject: 2-D interpolation
I'm interested in software for 2-D interpolation. At present we use
Robert Renka's algorithm, publised as ACM 624 but this seems to have
one drawback - it can introduce spurious local extrema.
If one has gridded data then F.N. Fritch's BIMOND does a good job,
the problem is that I have scattered data.
Any suggestions?
---------------------------
%% For further information about the NAG Users Association
%% please contact:
%%
%% Janet Bentley, Administrator NAGUA,
%% Shore Lane Farm, Blackstone Edge Old Road,
%% LITTLEBOROUGH, Lancashire, OL15 0LQ, UK.
%%
%% 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