From research!csnet!csnet-relay.arpa!krj%toronto.csnet 11 Nov 85 11:15:17 est Original-From: Ken Jackson Subject: ODE solver for NetLib To: golub@na Cc: igladwel@ucl-cs.arpa, na.dongarra@su-score.arpa, na.grosse@su-score.arpa, tehull%toronto.csnet@CSNET-RELAY.ARPA Via: Csnet-Relay; 14 Nov 85 3:44-EDT Re: Your request of October 9 for codes (ODE solvers in particular) for NetLib. As you may know, in the mid 1970's, Tom Hull, Wayne Enright, and I wrote a nonstiff IVP solver called DVERK based upon explicit Runge-Kutta formulas; this Fortran 66 code has been in the IMSL library since 1976. We are willing to submit it to NetLib if you are interested in having it. If so, please let me know the procedure for submitting codes to NetLib and any "standards" to which they should conform. We are currently in the process of revising and extending DVERK, but this probably won't be finished for a year or two (depending in part upon whether or not we get a M.Sc. student to help with this project). We are willing to submit that program to NetLib, when it is completed, as a replacement for DVERK. From research.att.com!pandora.llnl.gov!petzold Tue Jun 25 13:41:45 PDT 1991 Received: by pyxis; Tue Jun 25 16:41 EDT 1991 Received: by inet.att.com; Tue Jun 25 16:41 EDT 1991 Received: by pandora.llnl.gov (4.0/LLNL-1.17) id AA09056; Tue, 25 Jun 91 13:41:45 PDT Date: Tue, 25 Jun 91 13:41:45 PDT From: petzold@pandora.llnl.gov (Linda R. Petzold) Message-Id: <9106252041.AA09056@pandora.llnl.gov> To: ehg@research.att.com Subject: Re: DASSL in NETLIB Yes, it is the move that I am worried about - whether I will be organized enough to accomplish this for some time afterwards. Anyway, the following eight messages will be DDASSL and SDASSL (double and single precision). The files are: DDASSL - main body of DASSL source DLINAL - Linear algebra routines needed (these are from LINPACK) DAUX - Machine constant routine and error handler (machine constants from Bell, Error handler from SLATEC - this is the new error handler from SLATEC - XERRMSG) DDEMO - demonstration program -- and the same for single precision version. I am leaving them in separate files because I don't know how you want to handle those routines that are in common with LINPACK and SLATEC. -- thanks! Linda From research.att.com!pucc.PRINCETON.EDU!ERENJ.BITNET!GDBYRNE Thu Jun 27 10:06:56 EDT 1991 From: "George D. Byrne" Subject: Submitting code to netlib To: Eric Grosse Eric, I am interested in submitting a single and a double presion version of the ODE solvers VODE and VODPK to netlib. These might be called svode, dvode, svodpk, and dvodpk . I have read over the instructions for submitting codes and have a few questions. Would you please telephone me to guide me a bit? I would like to have the codes submitted BEFORE ICIAM 91. My telephone number is 908-730-3115. Thank you. George Byrne From janus.cs.umn.edu!petzold Sat Aug 10 15:09:36 CDT 1991 Received: by pyxis; Sat Aug 10 16:09 EDT 1991 Received: by inet.att.com; Sat Aug 10 16:09 EDT 1991 Received: from janus.cs.umn.edu by cs.umn.edu (5.61/1.14) id AA24912; Sat, 10 Aug 91 15:09:40 -0500 Received: from constantine.cs.umn.edu by janus.cs.umn.edu; Sat, 10 Aug 91 15:08:59 CDT Date: Sat, 10 Aug 91 15:09:36 CDT From: "Linda Petzold" Message-Id: <9108102009.AA01366@constantine.cs.umn.edu> Received: by constantine.cs.umn.edu; Sat, 10 Aug 91 15:09:36 CDT To: ehg@research.att.com Subject: missing subroutine in netlib DASSL (ugh!) Eric, I recently discovered that the following small subroutine was missing from the version of SDASSL and DDASSL that I sent you for NETLIB. Sorry. -- Linda INTEGER FUNCTION I1MACH (IDUM) INTEGER IDUM C----------------------------------------------------------------------- C THIS ROUTINE SETS THE STANDARD ERROR MESSAGE UNIT NUMBER. C THIS IS A SIMPLIFIED VERSION OF THE SLATEC LIBRARY MACHINE CONSTANT C ROUTINE I1MACH. C----------------------------------------------------------------------- I1MACH = 6 RETURN END ------m janus.cs.umn.edu!petzold Sat Aug 10 16:14:30 EDT 1991 From-: Eric Grosse 908-582-5828 To-: janus.cs.umn.edu!petzold Date-: Sat Aug 10 16:11:10 EDT 1991 Subject: missing subroutine in netlib DASSL (ugh!) If it's ok with you, I'd just as soon leave that out. The trouble is that someone loading dassl will get a version of i1mach that quietly returns junk if they're trying to find out the number of characters per word, etc. We've had a tradition in netlib of removing *1mach, blas, and linpack from files before depositing them in the collection to avoid problems of this sort. But I'll put it in if you really want, though I think it would be best to at least print a warning message if IDUM isn't the expected value. Best wishes, Eric From janus.cs.umn.edu!petzold Sat Aug 10 15:18:27 CDT 1991 Received: by pyxis; Sat Aug 10 16:18 EDT 1991 Received: by inet.att.com; Sat Aug 10 16:18 EDT 1991 Received: from janus.cs.umn.edu by cs.umn.edu (5.61/1.14) id AA25977; Sat, 10 Aug 91 15:18:32 -0500 Received: from constantine.cs.umn.edu by janus.cs.umn.edu; Sat, 10 Aug 91 15:17:50 CDT Date: Sat, 10 Aug 91 15:18:27 CDT From: "Linda Petzold" Message-Id: <9108102018.AA01370@constantine.cs.umn.edu> Received: by constantine.cs.umn.edu; Sat, 10 Aug 91 15:18:27 CDT To: ehg@research.att.com, petzold@janus.cs.umn.edu Subject: Re: missing subroutine in netlib DASSL (ugh!) No problem. Leave it out, if this is your policy. -- Linda From cwi.nl!bsom Fri Aug 30 11:15:39 +0200 1991 Received: by pyxis; Fri Aug 30 05:15 EDT 1991 Received: by inet.att.com; Fri Aug 30 05:15 EDT 1991 Received: by charon.cwi.nl with SMTP; Fri, 30 Aug 1991 11:15:41 +0200 Received: by forel.cwi.nl via EUnet; Fri, 30 Aug 91 11:15:39 +0200 Date: Fri, 30 Aug 91 11:15:39 +0200 From: bsom@cwi.nl Message-Id: <9108300915.AA06700@forel.cwi.nl> To: ehg@research.att.com Subject: contribution Dear Dr. Grosse, Herewith, I submit a contribution for the netlib public library. It concerns an ODE solver, which is, of course, best fitted into netlib/ode. Although the code has not yet reached the sophisticated level of a production-quality code, I think it is of interest to include it into netlib/ode since the present contents of this library aims at problems that are either stiff or nonstiff. There are not many codes for the problems 'in between'. This is exactly the class of problems RKC is designed for. If you are going to install the code, I would like to suggest the name: RKC. For the index and keywords file, I suggest the following: *** for the keyword list (ODE/RKC) Parabolic PDEs, Reaction-diffusion equations, method of lines, Explicit time stepping, Runge-Kutta-Chebyshev method. *** for the index in netlib/ode RKC - Explicit time integration (double precision version). RKC is designed to solve the initial-value problem for a system of first-order differential equations. It is based on (a family of) second-order explicit Runge-Kutta methods with increased real stability boundary. Consequently, for a successful application, the eigenvalues of the Jacobian matrix should be in the vicinity of the negative real axis. For instance, ODEs originating from semi-discretization of (diffusion-dominated) parabolic partial differential equations or from semi-discretized reaction-diffusion equations are examples RKC is designed for (especially if such problems have more than one spatial dimension). In conclusion, RKC aims at problems (with a negative spectrum of the Jacobian) which are not 'stiff enough' to justify the overhead involved when an implicit method is selected but which are 'sufficiently stiff' to exclude the use of an explicit method with a (fixed) small stability region. (Sommeijer, CWI Amsterdam, August 1991). The actual code of RKC will be sent by separate email (it is of moderate size). I hope that these emails contain all the information you need. If not, please let me know. With kind regards, Ben Sommeijer Centre for Mathematics and Computer Science (CWI) Amsterdam, The Netherlands email: bsom@cwi.nl From cs.umn.edu!petzold Thu Jun 25 15:14:19 0500 1992 Received: by pyxis; Thu Jun 25 16:13 EDT 1992 Received: by inet.att.com; Thu Jun 25 16:13 EDT 1992 Received: by cs.umn.edu (5.61/1.14) id AA05486; Thu, 25 Jun 92 15:14:19 -0500 Date: Thu, 25 Jun 92 15:14:19 -0500 From: "Linda Petzold" Message-Id: <9206252014.AA05486@cs.umn.edu> To: petzold@cs.umn.edu, wmc@research.att.com Subject: Re: DASRT solver Info on DASRT: DASRT is, like DASSL, a public-domain code. It has been extensively tested although it has not had nearly as many users as DASSL. It is in good shape and could go in NETLIB if you so desire (in fact, this would be convenient for me, too). -- Linda .