Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!neat.ai.toronto.edu!krj From: krj@na.toronto.edu (Ken Jackson) Newsgroups: ut.na Subject: NAgMAg V89 #14 Message-ID: <89Jun21.144236edt.11720@neat.ai.toronto.edu> Date: 21 Jun 89 18:42:24 GMT Distribution: ut Organization: Department of Computer Science, University of Toronto Lines: 188 NAgMAg Wednesday, June 21 1989 Volume 89 Issue 14 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% The official electronic digest of the NAG Users Association %% %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% Today's Topics Accuracy of the F02BJF Sparse Matrix Package Response to Volterra Integral Equation Request Usage Statistics NAGUA meeting update %% Still hot in Kent I've replaced the rice on the allotment %% (sown after the monsoon earlier this month) with cacti. %% %% A few more articles this week -- please keep them rolling in. %% I believe it really helps to put a digest out approximately %% weekly as contributers then get a reasonably fast turn round %% on queries. %% %% Tim --------------------------- Date: Wed, 14 Jun 89 13:21:49 BST From: MBBGPTB@cms.manchester-computing-centre.ac.uk Subject: Accuracy of the F02BJF I recently used the generalized eigenvalue routine F02BJF. For the accuracy control I called the function X02AAF(0.0e0) and I expect using this I will get a reasonably accurate result. What I found was a very inacurate result. My matrix is quite small (19 x 19) and has a pair of complex e'values; the rest of the eigenvalues are real. Using F02BJF it turned out that all the eigenvalues are complex with quite significant imaginary part. (I ran my program on IBM-CMS with double length). Have I done something wrong ? --------------------------- Date: Thu, 15 Jun 89 12:06:38 EET From: J}rgen Bruchhaus Subject: Sparse Matrix Package I need FORTRAN-subroutines for the usual operations on sparse matrices ( including factorization ) . Does anyone know about a freely available package ? Thanks in advance , Juergen Bruchhaus Institute of Mathematics University of Cologne Weyertal 86-90 D-5000 Koeln 41 ( BITNET : MI004@DK0UMI1 ) Acknowledge-To: --------------------------- Date: Thu, 15 JUN 89 16:19:53 GMT From: CAROLINE@vax.nag.co.uk Subject: Response to Volterra Integral Equation Request The following item is in response to the request from Harvey Richardson, Heriot-Watt University in issue 12. Routines for Volterra Integral Equations ---------------------------------------- Algorithms for the numerical solution of Volterra integral equations of the first and second kind are under consideration in the D05 chapter of the Fortran library. In fact, a routine for the solution of nonlinear convolution integral equation of the second kind, which occurs in real life problems, t (1) y(t) = f(t) + I k(t-s)g(s,y(s)) ds a has been developed and will appear in Mark 14 of the Fortran library. This routine (D05BAF) uses a variety of reducible linear multistep methods of Adams and backward differentiation formulae type; the user should supply a tolerance for the desired accuracy. Equation (1) is a subclass of the general second kind Volterra equation t (2) y(t) = f(t) + I K(t,s,y(s)) ds a with K(t,s,y(s)) = k(t-s)g(s,y(s)). To develop efficient and reliable codes for Volterra equations, the structure of the kernel K(t,s,y(s)) should be considered and the choice of the method is very important. For example, in solving (1) the number of function evaluations are reduced significantly due to the convolution nature of the kernel. For kernels with weak singularities of the form -1/2 k(t,s,y) = (t-s) H(t,s,y) special methods exist and a separate routine will be developed; again the structure H(t,s,y) is important. Routines for the solution of the general equation (2) are being considered. In the meantime we are preparing codes to exploit commonly occuring problems with special kernel structures. Mishi Derakhshan 15 June 1989 --------------------------- Date: Fri, 16 Jun 89 16:33:40 EDT From: Steve Hotovy Subject: Usage Statistics I read about tracking subroutine calls in Issue 12 of NagMag. We are very interested in doing something like this here at Cornell. Could you point me in the right direction for more information? Steve Hotovy CNSF %% Correct me David if I'm wrong, but I thought all implementations %% of routine call accounting had been site dependent `hacks'. These %% are therefore very machine/system dependent. I'd like to see NAG %% put the hooks in the sources to allow such accounting to be %% implemented. Would others be interested in this? Tim --------------------------- Date: Tue, 20 JUN 89 15:30:24 GMT From: CAROLINE@vax.nag.co.uk Subject: NAGUA meeting update NAGUA'89 -------- Conference latest - Julian Gallop from Rutherford Appleton Laboratory will be giving the presentation on 'Library Management in a Network' - Ron Furzeland from Koninklijkeshell-Laboratorium, Amsterdam, one of the speakers on Industrial Applications of NAG Software, will be speaking on 'Differential Equation Software in the Petrochemical Industry: Requirements and Needs'. Just in case any of you are not yet NAGUA members (although I'm sure you all are, as NAGMAG is a NAGUA service), membership of just #30, or $60 in North America, entitles you to a discount of #30 at the one day tutorial AND a discount of #30 at the main meeting. So if you will be coming to the whole conference or even just the main meeting, then it certainly makes sense to join. There, that's my sales pitch for now. Caroline Foers --------------------------- %% 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