Summary of MOPAC7 on a PC



 =======================================================================
  I sent the following message to CCL on Nov. 29, 1994, seeking help
  in using MOPAC on a Pentium.  I have received a number of useful
  suggestions which are summarized below.  Thanks to all.
 =======================================================================
 =======================================================================
  Original Message:
 =====================================================================
 Date: Tue, 29 Nov 1994 14:38:11 -0500 (EST)
 From: David Close <R29CLOSE (- at -) ETSU.EAST-TENN-ST.EDU>
 Subject: CCL:MOPAC on PC's
 To: chemistry (- at -) ccl.net
 Organization: East Tennessee State University
   Dear Netters:
   I would like to know if others have had success with MOPAC or AMPAC
 on a PC.  I am trying to use Ivar Koppel's MOPAC 7.1 on my Gateway
 Pentium (P5-66).  Ivar sent me the FORTRAN code which I compiled using
 Microsoft's Powerstation 32-bit compiler.  The program executes, but
 gives results that differ from some of the popular tests examples.
   I need to know if others have had similar problems.  Some users have
 mentioned the use of compilers from Microway or Lahey.  I am wondering
 if this is important.  Also it would help to know if Ivar's program has
 given reliable results in other situations.  Then there is this Pentium
 division error that everyone is discussing.  Can this influence MO
 calculations?
   Regards,  Dave Close.
 ========================================================================
 Reply from Ken Fountain about using the Lahey compiler.
 ========================================================================
 Date:        Tue, 29 Nov 94 15:01:33 CST
 From:        "FOUNTAIN, KEN" <SC18%NEMOMUS.bitnet (- at -)
 ACADEMIC.NEMOSTATE.EDU>
 To:          <R29CLOSE (- at -) ETSU.EAST-TENN-ST.EDU>
 Subject: Re: CCL:MOPAC on PC's
 In-Reply-To: In reply to your message of TUE 29 NOV 1994 03:38:11 CST
 I think it can.  If you send me the code for MOPAC7.1 I can compile it
 with a Lahey compiler with a DX (- at -) 66mHz chip and we can compare results.
 Ken Fountain
 Northeast Missouri State University
 Science Hall
 Kirksville, MO 63501
 =======================================================================
 Reply from Janusz Mrozek warning about MICROSOFTS POWERSTATION FTN
 compiler.
 =======================================================================
 Return-Path: <MROZEK (- at -) TRURL.CH.UJ.EDU.PL>
 From: <MROZEK (- at -) Trurl.ch.uj.edu.pl>
 Organization:  Faculty of Chemistry, UJ
 To: David Close <R29CLOSE (- at -) ETSU.EAST-TENN-ST.EDU>
 Date:          Wed, 30 Nov 1994 08:22:23 MET+2
 Subject:       Re: CCL:MOPAC on PC's
 Dear Dr. Close,
     Compiler is the thing that makes a real difference when moving
 from mainframe codes to the PC. In my opinion Microsoft should be
 legally forbidden to touch Fortran compilers. We are running for more
 than two years an OS/2 version of MOPAC 6.0 on our PC machines, and
 it reproduces the results from other machines to the very last digit,
 provided such a thing as reproducible AMPAC/MOPAC results do exist.
 We are using MICROWAY FTN compiler (an older version) and I think
 this is one of the best compilers I ever worked with. If you are
 interested in a running MOPAC 6.0 code you can ftp it as an anonymous
 user from:  Trurl.ch.uj.edu.pl, directory is mopac6/os2. The MOPAC 7
 code should be there in about the beginning of the year.
 Best regards,
 Janusz Mrozek
 ====================================================
 Janusz Mrozek,
 Department of Theoretical Chemistry
 & Department of Computational Methods in Chemistry
 Faculty of Chemistry,
 Jagiellonian University
 3 R. Ingardena St.
 30-060 Cracow, Poland
 ========================================================================
 Reply from Ivar Koppel about various compliers used on MOPAC7
 ========================================================================
  30 Nov 1994 11:20:43 UTC+0100
 Date: Wed, 30 Nov 1994 11:20:43 UTC+0100
 From: Ivar Koppel <rockoppel (- at -) roca.csic.es>
 Subject: mopac
 That should be a lot of time ago when I posted you that Mopac..
 Can you remember me some details? I think you got the sources from
 tera.chem.ut.ee. Some months ago somebody posted note on CCL that
 mopac 7 contains some bugs, and I fixed those, but at the very
 beginning there were unmodified sources at tera. What compiler
 are you using? I have compiled the code on both using MS Powerstation
 Fortran 1.0 and f2c+gcc combination, the results differ a bit. (sometimes
 Powerstation version did 1 more iteration cycle (on large molecule) than
 the gcc compiled version, but differencis on final heat of formation
 were less than 0.000xx Kcal/mol. Right now I m using the same sources
 on Linux, gcc+f2c , and results are fine. The only problem detected
 so far that in eigenvector following calculation the calculation aborts
 sometimes, giving message "CONSULT PROGRAMMER, etc". But I haven't
 noticed incorrect results. At present time I'm working in Madrid,
 so perhaps you can use my E-mail address here to send some more
 details. For verification you can use mopac6 from www.ccl.net,
 directory should be /pub/chemistry/software/dos/mopac_for_dos or something
 I haven't got any complaints for that program. The sources for mopac7 are
 also available at solaris.ims.ac.jp , but I think they are unmodified.
 To compile them under dos you must comment the DATE and TIME calls, I
 think in SECOND.F
  Best regards,
  Ivar
 =======================================================================
  Reply from Klaus-Peter Gulden.  Note:  I have had problems contacting
  him.  All my messages at the address given have come bouncing back.
 =======================================================================
 Return-Path: <gulden (- at -) CHEMIE.UNI-WUERZBURG.DE>
 Date: Wed, 30 Nov 1994 16:19:27 +0100 (MET)
 From: Klaus-Peter Gulden <gulden (- at -) wocx03>
 Subject: Re: CCL:MOPAC on PC's
 I've ported MOPAC6.0 on Linux! I don't know whether this seems to be an
 appropriate solution for you. All test files have been successfully
 calulated. It's compiled with the f2c converter and subsequent gcc
 compilation. Only minor changes were neccessary to make the program
 run on a LinuX-platform.
 You don't mention your operating system, but if you are interested I would
 like to give either the sources or the executables.
 > mentioned the use of compilers from Microway or Leahey.  I am wondering
 > if this is important.  Also it would help to know if Ivar's program has
 > given reliable results in other situations.  Then there is this Pentium
 > division error that everyone is discussing.  Can this influence MO
 > calculations?
 Are the sources to MOPAC7.1 free to the public? In this case you can give
 the FORTRAN-files and I try to port it on LinuX. So after that we can see
 whether the numerical problems are due to the PENTIUM-division bug or
 due to the Microsoft SUPER POWER HYPER ..... (and so on) Compiler (or
 what do the comercials say?).
 ok, see you and bye
 Klaus-Peter
  ------------------------------------------------------------------------
 | Klaus-Peter Gulden | Institut fuer Org. Chemie, Universitaet Wuerzburg |
 | Am Hubland         | mail  :  gulden (- at -) chemie.uni-stuttgart.de
 |
 | D-97074 Wuerzburg  | voice :  (0049)-(0)-931-888-4750                  |
 | Germany            | fax   :  (0049)-(0)-931-888-4750                  |
  ------------------------------------------------------------------------
 =======================================================================
 Reply from S. Shapiro reporting similar problems with MOPAC 6.
 =======================================================================
 Return-Path: <toukie (- at -) ZUI.UNIZH.CH>
 Date: Wed, 30 Nov 1994 16:32:25 +0100 (MET)
 From: "Hr Dr. S. Shapiro" <toukie (- at -) zui.unizh.ch>
 Sender: toukie (- at -) zui.unizh.ch
 Subject: MOPAC 7 for PC
 Dear Prof. Dr. Close;
      Thank you for replying so promptly to my request for contact.
      I have been using a PC version of MOPAC 6.0, and wished to
 "upgrade" to
 MOPAC 7.  I obtained the FORTRAN files from Prof. Dr. Koppel, but was unable
 to compile it into a PC-executable form.  I then sent it to colleagues in
 Belgium and Israel, and apparently they haven't been able to compile it,
 either.  The issue rested at that stage until I saw your posting in CCL.
 Under these circumstances, I should be very grateful if you would send me
 a unencoded zipped version of your working version of MOPAC 7 compiled for
 execution on a PC fitted with MS-DOS 6.22.  If you wish, I can send you a
 copy of my version of MOPAC 6.0, which I _know_ works properly, and then
 we can compare notes  Please note that my server will store very large
 files  for me until I elect to download them.
    By the eay, have you ever made a stab at trying to compile the latest
 version of MOPAC, viz. MOPAC93, on a DOS machine?  Do you know just what
  are the differences between the capabilities of MOPAC 6.0, MOPAC 7, and
  MOPAC93? I would also be very interested in learning of other responses
 that you may receive to your CCL posting.
  Thank you for your time.  I look forward to hearing from you in the near
 future.
 Your sincerely,
 (Dr.) S. Shapiro
 Institut fuer orale Mikrobiologie und allgemeine Immunologie
 Zentrum fuer Zahn-, Mund-, und Kieferheilkunde der Universitaet Zuerich
 Plattenstrasse 11
 Postfach
 CH-8028 Zuerich, Switzerland
 =======================================================================
 Most helpful reply from Rainer Koch offering to help check some of
 my poor results.
 =======================================================================
 From: Rainer Koch <cri (- at -) rz.uni-jena.de>
 Message-Id: <199412010732.IAA12207 (- at -) cnve.rz.uni-jena.de>
 Subject: Re: CCL:MOPAC on PC's
 To: R29CLOSE (- at -) ETSU.EAST-TENN-ST.EDU (David Close)
 Date: Thu, 1 Dec 94 8:32:48 MEZ
 Would you be so kind to send me some of your faulty results? I have done a
 port of MOPAC 93 to PC`s some time ago using Powerstation. It would be
 therefore interesting to see whether it is a problem of a) MOPAC 7 (Ivar`s
 or  the original code) b) Pentium FDIV c) FORTRAN Powerstation.
         Rainer Koch
 ******************************************************************************
 *       Rainer Koch                                                          *
 *       Institut fuer Organische und Makromolekulare Chemie                  *
 *       Universitaet Jena                                                    *
 *       Humboldtstr. 10                                                      *
 *       07743 Jena                                                           *
 *       Germany                                                              *
 *                                                                            *
 *       E-mail: cri (- at -) rz.uni-jena.de
 *
 *       E-mail: koch (- at -) organik.uni-erlangen.de
 *
 *       Fax:    0049/0-3641-6-35600                                          *
 ******************************************************************************
 From: Rainer Koch <cri (- at -) rz.uni-jena.de>
 Message-Id: <199412020855.JAA05080 (- at -) cnve.rz.uni-jena.de>
 Subject: MOPAC x/PC
 To: R29CLOSE (- at -) ETSU.EAST-TENN-ST.EDU (David Close)
 Date: Fri, 2 Dec 94 9:55:08 MEZ
 Dear Dave,
 Anyway, if you are interested in source code and/or executables for a variety
 of compilers, just ask us :-) we (my collegue Bernd Wiedel and myself)
 have ported MOPAC 6 to PC`s back in 1991 using Mircoway Fortran and a Phar Lap
 DOS-Extender (published at QCPE as QCMP 113). Other versions for Microsoft
 Powerstation (Fortran) and the gcc compiler exist as well as versions of
 MOPAC 7 and MOPAC 93. With these programs you can run MOPAC in a DOS box
 under Windows even in the background. There is no need to change to Linux
 unless you want to.
 Hope this will be of some help.
                 Best regards, Rainer.
 **************************************************************************
 =========================================================================
 NOTE:  I have had several more exchanges with Rainer (exchanging test
 outputs.  Rainer is convinced that the problems I am having are mot
 very significant, and that indeed my version of MOPAC 7 is ok.  He
 has suggested a number of things to try (which I will do as soon as
 possible.
 ========================================================================
 More from Ivar Koppel after I wrote him.
 ========================================================================
 Return-Path: <rockoppel (- at -) ROCA.CSIC.ES>
 I used mopac 7.1 with powerstation for calculating heats of formation,
 and the results are the same that of mopac 6 on convex or VAX. Maybe
 on case when molecule converges badly the mopac 7 was better.
 f2c+gcc means using fortran to C translator and GNU C compiler from
 Open Software Foundation. Linux is a free Unix clone, it has now
 Watcom fortran but I have used the free fortran to C and gnu C
 combination, the penalty is told to be 10% perfomance increase.
 BTW, I have never carried out reaction path calculations using mopac 7 so
 I dont know how it works. The "porting" of mopac 7 to DOS means that
 I commented some lines in SECOND.F and somewhere where the DATE call is made,
 otherwise its the standard code for sun. I got it from solaris.ims.ak.jp,
 and as I found out that only a few people know that mopac 7 is freely
 available I compiled it and said that such thing is available. By the
 way, J.Stewart is always offically saing that mopac 93 is the right thing
 for production work :-). Ok, enough of jokes, I m myself very much interested
 in finding out what is the problem with mopac 7.
 Best regards,
 ========================================================================
 Reply from Wally Reiher expressing similar problems getting MOPAC to
 work on a PC.
 ========================================================================
 From: wallyr (- at -) netcom.com (Walter E. Reiher III)
 Message-Id: <199412022021.MAA18168 (- at -) netcom11.netcom.com>
 Subject: Re: CCL:MOPAC on PC's
 To: R29CLOSE (- at -) ETSU.EAST-TENN-ST.EDU (David Close)
 Date: Fri, 2 Dec 1994 12:21:02 -0800 (PST)
 David,
 I have been struggling to port MOPAC93 to Windows NT and DOS (with a DOS
 extender, of course) using the Watcom compiler and having absolutely no
 luck.  I have gone through all kinds of contortions trying to figure out
 why the program returns results which differ substantially from the
 results I obtained with the same code on an SGI.  In some cases,
 minimizations which converge on the SGI don't on the Pentium.
      I have also been struggling with Watcom techincal support on this
 one and am not pleased with their response so far, but I haven't yet
 given up on them.
      After this experience, I was considering purchasing MS FORTRAN
 PowerStation, but after reading your post, I'm not so sure.  At least
 the PowerStation demo compiler I got compiled the program on the first
 try, which is more than I could do with the Watcom compiler--I just
 can't figure out how to link it with the Microsoft linker.  Do you have
 a linker script you can share with me?
 I am very interested in hearing your experiences with this--I'd
 appreciate seeing your summary of responses.  I'd be happy to share more
 details with you if I can be of any help.
 Wally
 ========================================================================
 Walter E. Reiher III, Ph.D.                            WallyR (- at -)
 netcom.com
 Consultant in Computational Chemistry
 P.O. Box 61056                                        voice 408-720-0240
 Sunnyvale, CA 94088                                     fax 408-720-0378
 ========================================================================
 SUMMARY:
 ========================================================================
   That's the summary so far.  I obviously have a lot of follow up work
 to do.  Some have suggested switching to OS/2.  OS/2 WARP is available
 for ca. $75. but must be used with Windows NT.  After New Year there
 is supposed to be a complete operating system from IBM that doesn't
 depend on Windows.
   Others have suggested MICROWAY's FTN compiler.  They cost $695. for
 the 486 version and $995. for the Pentium version.  That's a lot of
 money to spend for something the may or may represent an improvement.
   Several people suggested LINUX, but Rainer Koch says this is not
 necessary.
   In reading the above messages it appears that other are having pro-
 blems with getting MOPAC to work on a PC.  I am going to continue
 testing the version I have, and they I will summarize the results.
 Thanks again.
 David Close
 ========================================================================