[ut.na] NAgMAg v89 #21

krj@na.toronto.edu (Ken Jackson) (09/15/89)

From:	nagmag%ukc.ac.uk@NSFNET-RELAY.AC.UK
Date:	Thu, 14 Sep 89 13:26:10 EDT
Subject: NAgMAg v89 #21


NAgMAg		Wednesday,  September 13 1989	Volume 89   Issue 21

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%   The official electronic digest of the NAG Users Association   %%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

Today's Topics
                           Re: NAgMAg v89 #20
                       Transputer based NAG Library
                  E04NCF (problem type LS1, case: M < N)
                      Apollo Version of NAG Library

%%
%% Here we go again -- holidays and conferences over -- those of
%% you who didn't make York really missed out. The Medieval Banquet
%% was very good and there was a unique opportunity to see members
%% of both NAG and NAGUA making fools of themselves:-) Hopefully
%% somebody will post a "thoughts on NAGUA at York" article in
%% the near future.
%%
%% In the past I have gone to some lengths to ensure that bounced
%% digests do finally get through to their intended recipients. As
%% the mailing list has grown, and with it the number of trashed 
%% digests, I find I really haven't enough time to unravel the 
%% cryptic information from intermediate mailers. If you think
%% you have missed digests therefore send me a message and I'll
%% try and get back copies through to you.
%%
%% Tim
%%
---------------------------

Date: 22 Aug 89  22:21:40 bst
From: J.Blair-Fish@ed.ac.uk
Subject: Re: NAgMAg v89 #20

The message I wrote about NAG libraries on the transputer to the
transputer bulletin board may be misleading in wording. For a single
processor the only available NAG library uses INMOS FORTRAN. For other
compilers some routines have been ported by individuals with or without
the knowledge of NAG or NAG site representatives. No one to my
knowledge is distributing unofficial NAG libraries.

It should be noted that Mike Delves' company, NAS, which distributes an
OCCAM Numerical Library, is not related to NAG but has a very similar
name.

John Blair-Fish

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

Date: Tue, 29 AUG 89 17:56:03 GMT
From: CAROLINE@vax.nag.co.uk
Subject: Transputer based NAG Library

re: NAG on the Transputer    from Jeremy Du Croz, NAG Ltd
    ---------------------

We wish to correct and expand on the message which was quoted by David
Morse in NAgMAg #20.

Mark 12 of the NAG Fortran Library has been implemented to run on a
single transputer, with an IBM PC front-end, using the INMOS 3L
compiler. This software is available from NAG Ltd. The implementation
was done by Liverpool University.

Mark 13 of the NAG Fortran Library is currently being implemented,
again to run on a single transputer, using the Meiko Fortran compiler:
   a) on a Meiko Computing Surface
   b) on a Meiko PC card 
This implementation is being done by the Cambourne School of Mines. As
soon as this work has been completed and certified, it will be
available

There are no other implementations of the NAG Fortran Library on
transputers which are authorised by NAG.

We are aware that some individual users have transferred selected NAG
Library routines to various transputer-based systems, having obtained a
copy of the source-text from their local NAG Site Representative. In
such cases they should always sign a NAG source-text release form, and
it should be understood that the routines are only to be used for bona
fide research purposes.

For the future, NAG are currently working with Liverpool University on
the development of a parallel numerical library for the
transputer-based Supernode machine. The work is being undertaken as
part of the EC-funded Esprit 2 project, Supernode 2. This parallel
library will cover the same principal areas of numerical computation as
the NAG Fortran Library, but, at least initially, it is likely to be
much less extensive in its coverage and not directly compatible with
the NAG Fortran Library. The library will be callable from Fortran
programs, and is intended to be transportable to other transputer-based
systems. When completed, the library will be marketed by NAG Ltd.
Liverpool University developed a small pilot library of parallel
numerical routines under the Supernode 1 project, and this is available
from NA Software Ltd.

NAG is also planning other collaborative research projects to develop
parallel numerical software for various types of parallel architecture,
including transputer-based systems.

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

Date:     Wed, 30 Aug 89 15:20:23 EDT
From:     Colin Campbell (Computing Sers) <campbell@watdcs.uwaterloo.ca>
Subject:  E04NCF (problem type LS1, case: M < N)

We've run into a wall using NAG 13 routine E04NCF on IBM CMS.  As a
test problem consider the online example program "EX1".  We made the
following changes to make it the case "M < N":

 
        ---- change 42          PARAMETER       (M=10,N=9,NCLIN=3)
        ==== to 42              PARAMETER       (M= 8,N=9,NCLIN=3)
 
        ---- change 44          PARAMETER       (NROWC=NCLIN,NROWA=M)
        ==== to 44              PARAMETER       (NROWC=NCLIN,NROWA=10)
 
When we run it, the output is:
 
         Parameters
         ----------
         Problem type...........       LS1
         ...
         Workspace provided is     IW(     9),  W(   261).
         To solve problem we need  IW(     9),  W(   261).
 
         *****  Parameter  M  is invalid in routine  F01YDY  *****
                Value supplied is      8
            The input parameters contained  1 error(s)
         ** ABNORMAL EXIT from NAG Library routine F01YDY: IFAIL =    -1
         ** NAG hard failure - execution terminated
 
What gives?  The documentation says we can have M < N, but the routine
won't let us:

	 M - INTEGER
	      ... If the problem is specified as type LS1, ....
	      Note that all possibilities (M < N, M = N, and M > N) are
	      allowed in this case.

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

%%
%% I got this out of a recent copy of the na digest 
%% run by Cleve Moler. Waterloo are certainly on our
%% distribution list -- I will however post a personal
%% reply when nag reply. Tim
%%

From: Paul Calamai <phcalamai@watfun.waterloo.edu>
Date: Sun, 3 Sep 89 16:24:44 EDT
Subject: Apollo Version of NAG Library

Perhaps someone out there in na land can help me with a
problem I'm having.

The University of Waterloo has obtained a site licence for
the NAG library.  My research group has an Apollo DN3500
running OS SR10.1 with all three environments installed
(ie. bsd4.3, sys5, aegis) but with bsd4.3 as the default
SYSTYPE.  The implementation of NAG we received contained a
compiled, tested, ready-to-use version of NAG Mark 13 Fortran
double-precision library for APOLLO DOMAIN hardware running
v9.7 AEGIS with fortran compiler v9.66 with the LBR utility.
We have v10.1 fortran.  Because the v9.7 AEGIS implementation
is worthless under SR10.1 we have recompiled the sources twice
(once using options -compress -cpu 3000 and once using
options -dba -cpu 3000) and recreated the library
using ar.  The problem is that revalidating the library
by running  all the example problems identifies a number of
problems that fail and many (~20) that produce no output
beyond a header (identifying the routine) and the column
headings for a table that never gets printed.

I'm wondering if anyone else has gone through this exercise
with more success then we've had.  Any recommendations
would be appreciated.  If you need more details please
email me and I'll do my best to provide them.

Thanks

Paul H. Calamai
Dept. of Systems Design Engineering
U of Waterloo
(519)885-1211 x3182
phcalamai@water.uwaterloo.ca
phcalamai@water.waterloo.edu
na.calamai@patience.stanford.edu

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

%%   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.
%%
%% or
%% Caroline J Foers
%% NAGUA and Service Coordinator, NAG Ltd,
%% Wilkinson House, Jordan Hill Road,
%% Oxford, OX2 8DR, U.K.
%%
%% e-mail   caroline@uk.co.nag.vax

%%
%%   Replies or submissions to          nagmag@uk.ac.ukc
%%   Distribution changes to    nagmag-request@uk.ac.ukc
%%
%%   END OF ISSUE 


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

Reposted by

Prof. Kenneth R. Jackson,      krj@na.toronto.edu   (on Internet, CSNet, 
Computer Science Dept.,                              ARPAnet, BITNET)
University of Toronto,         krj@na.utoronto.ca   (on CDNnet and other 
Toronto, Ontario,                                    X.400 nets (Europe))
Canada   M5S 1A4               ...!{uunet,pyramid,watmath,ubc-cs}!utai!krj
(Phone: 416-978-7075)                               (on UUCP)
(FAX: 416-978-4765)