Summary of data flow responses



 >>>>>> Summary of Responses to inquiry about data flow
 programs <<<<<<<<
 Thanks to all for your interest. There seem to be relatively few users
 of data-flow programs for molecular modeling at present, but I did get
 a few interesting responses.
 First a few comments of my own.
 Here at the CTC we use IBM Data Explorer (DX) heavily for general visualization
 purposes. We also have AVS and MSI chemistry view but I'm less experienced
 with those products. DX falls somewhere between a low-level programming
 language for graphics and a high-level application with pull-down
 menus and a command line. This type of product is best suited for
 research groups who do a lot of graphics, especially animation, and need
 real flexibility, control and variety. Data flow programs are a real
 time-saver over lower-level programming in C! I would not recommend them
 to casual graphics users who would be better served with a specific
 program like Insight or Quanta. High-level applications can be written
 with a data-flow program (like MSI Chemistry viewer), but MSI is the only
 one I know about.  These kinds of programs have a bright future, but
 they're still too new.
 What is data-flow programming?
      Instead of writing a program as a text file which you then compile
      and run, data flow programming uses a Graphical User Interface (GUI)
      to visually construct a network of boxes (modules) connected by pipes.
      When the visual program is run, data starts at the top of the network
      (usually from a data-reading module of some kind) and flows downward
      through the pipes. Whenever a module is encountered, the data is
      modified. A standard set of modules for performing such tasks as
      isosurface construction, interpolation, mapping, coloring, annotation
      and other data and data structure manipulations is usually included
      with the GUI. At the bottom of a network, one usually combines all the
      data pipes together and passes them to an image display module.
 Advantages
    o  Easier than low-level graphics programming in C
    o  Trivial to modify and play what-if games
       (you just point, click on and move boxes)
    o  Can link in your own programs as modules (with some restrictions)
    o  Natural coarse-grain parallelism (if you have multiple processors)
    o  Can combine and compare many different data sets in a greater
       variety of ways than application-specific programs
    o Can add control panels customized to your application
    o Works well for animation (In my experience with DX)
    o Anonymous ftp sites for free modules exist for both AVS and
      IBM Data explorer.
 Disadvantages
    o Gives new meaning to the expression "plumber's nightmare"
      As in any programming language, it's easy to make an uncommented,
      disorganized mess. Personally, however, I find a well-organized
      network easier to understand than a directory fill of C files.
    o Lower-level than application-specific programs
      You won't see the word chemistry unless you write it yourself.
      The MSI chem viewer, however, is a product specifically for chemistry
      that runs under AVS. As far a I know this is the only commercial
      data-flow application for chemistry. Others may follow soon.
    o At least in IBM Data Explorer, the strict adherence to
      "pure-function semantics" discourages iterative or recursive
      calculations. Modules must not have internal states that
      change from one call to the next. The output results must be
      purely a function of the input data. Potential function evaluation
      is ok but trajectory calculation is tricky.
    o I find these programs to be not quite as interactively responsive
      as a custom-written C application using hardware polygon rendering.
 ********  List of replies  *********
 >Is there anyone else out there using data-flow type programs
 >for molecular modeling?
 >
 >        IBM Visualization Data Explorer
 >        SGI explorer
 >        AVS
 >
 >How do you like them? What are their strengths and weaknesses?
 ----------------
 From: jas' at \`medinah.atc.ucarb.com (Jack Smith)
 Subject: Re: Molecular modeling with data flow programs
 Status: RO
 Richard:
    We have just started looking at them.  We currently have SGI's Explorer
 and downloaded the MOPAC Viewer.  We also just got IBM's Data Explorer in
 for evaluation and plan to look at AVS ChemistryViewer (from MSI)
 relatively soon.
    I started using SGI Explorer to analyze MOPAC and DMol results for metal
 surface reactions.  I am particularly intrigued with opaque renderings of
 electron density.  I plan on making a QuickTime movie of the complete
 reaction.  (I make QuickTime movies of reactions pretty regularly on a
 CAChe system.)
    The visual programming concept is quite appealing and has a bright
 future, but I don't think it's quite ready for the casual user yet,
 although close.  The "scripting" language in SGI Explorer is
 definitely
 geared toward C programmers -- which pleases me, but not my
 strictly-FORTRAN or non-programmer friends.
    I had quite a bit of experience with LabVIEW on the Mac in years past
 (for doing data acquisition and control) and I found that data flow
 programming is quite powerful, but there are some subtleties that one has
 to become aware of.  Things are done asynchronously and with more degree of
 autonomy than in conventional sequential programming -- it's not just
 visual programming!  Things like global variables, for example, become more
 ambiguous (or ill-defined).  On the other hand, I think it's ideally suited
 for exploiting parallelism, and would highly encourage experimenting with
 data-flow architectures (visual or not) as a way to recognize opportunities
 for course grain parallelism.
   BTW, it was at the Cornell Theory Center that I first got to experiment
 with IBM's Data Explorer.  If you have never had a chance to experiment
 with LabVIEW on a Mac (also now available on the PC, and soon on SGIs), I
 encourage you to do so -- National Instruments has quite a headstart in
 visual data-flow programming and has quickly outgrown it's original domain
 of digital signal processing (DSP).
 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
   JACK A. SMITH
   ......................................................................
   Union Carbide Corp.           ||  Phone:    (304) 747-5797
   Advanced Technical Computing  ||  FAX:      (304) 747-5448
   P.O. Box 8361                 ||
   S. Charleston, WV  25303      ||  Internet: jas' at \`medinah.atc.ucarb.com
 -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
 From: jle' at \`world.std.com (Joe M Leonard)
 Message-Id: <199212072106.AA01159' at \`world.std.com>
 To: richard' at \`TC.Cornell.EDU
 Subject: Re:  Molecular modeling with data flow programs
 Status: RO
 I'd love to learn whether you hear of folks using these codes.  I was a
 Stardent when we did the Chemistry support for AVS, although I
 am only aware of 1 company using it commercially (MSI).  Henry Rzepa's
 group in the UK seems to be using SGI's explorer, and Doug SMith's
 group at U Toledo's using AVS.  I think only AVS has explicit Chemistry
 datatype support to date...
 Joe Leonard
 714-955-2120
 jle' at \`world.std.com
 From: "Bill,Molecular & Math Sciences, 508-467-6538  08-Dec-1992
 1248" <desimone' at \`mr4dec.enet.dec.com>
 To: richard' at \`TC.Cornell.EDU, desimone' at \`mr4dec.enet.dec.com,
         loh' at \`mr4dec.enet.dec.com
 Apparently-To: richard' at \`TC.Cornell.EDU
 Subject: AVS uses in modeling
 Status: R
 Richard,
 There are two places that I know of that are using AVS in conjunction with
 their modeling efforts:
 	UCSF (Jim Caldwell) - AMBER related biomolecular work
 	nm%decwrl::"caldwell' at \`cgl.ucsf.edu"
 	Scripps
 Angela, can you send Richard contacts for Scripps.
 Richard, you might also talk to Joe Leonard of Wavefunction, Inc.
 nm%decwrl::"jle' at \`world.std.com"
 	Bill DeSimone
 	Molecular Sciences Applications Marketing
 	DEC
 From: davem' at \`hydra.convex.com (Dave Mullally)
 Message-Id: <9212141739.AA29220' at \`hydra.convex.com>
 To: richard' at \`TC.Cornell.EDU
 Subject: Re:  Molecular modeling with data flow programs
 Status: R
 Hi Richard,
 I have written a number of AVS modules which are used by customers
 who have AVS on a Convex.  The 'data flow' part of AVS is actually
 a problem in general.
 Unless one has to, one does not want to know the details of how data
 flows in the model.  All that people really want to do is to see
 their own molecule and some limited number of properties.  To get
 people to use this type of program, you have to get make the
 data flow invisible in the final interface.
 I hope that this helps.
 Dave Mullally
 Convex Computer Corp.
 Date: Tue, 8 Dec 92 14:35:22 EST
 To: richard' at \`TC.Cornell.EDU
 Cc: desimone' at \`mr4dec.enet.dec.com
 Apparently-To: richard' at \`TC.Cornell.EDU
 Subject: AVS users in modeling
 Status: RO
 Hi Richard,
 Art Olson of SCripps is a heavy user of AVS for protein modeling. Art may be
 reached at "olson' at \`scripps.edu".
 Also, Ethan Merritt of U Washington is a user of AVS - he has converted RASTER3D
 to run with it - "MERRITT' at \`GOUDA.BCHEM.WASHINGTON.EDU" is his
 address.
 Regards,
 Angela
 From: cushing' at \`smoked.cse.ogi.edu (Judy Bayard Cushing)
 To: richard' at \`TC.Cornell.EDU
 Subject: Re:  Molecular modeling with data flow programs
 Status: RO
 i would appreciate being sent a summary or compendium of responses to
 your request.  also, can you tell me about ibm visualization data explorer?
 thanks!
 judy
 From: msdrl!andose' at \`uunet.UU.NET (Joe Andose)
 Subject: Molecular modeling with data flow programs
 To: richard' at \`TC.Cornell.EDU
 Date: Tue, 8 Dec 92 11:38:40 EDT
 X-Mailer: ELM [version 2.2 PL0]
 Status: RO
 Richard,
 I saw your query on the CompuChem Mail Exploder. We briefly looked at SGI
 Explorer for ad hoc data visualization in chemistry. However, Explorer has no
 pre-constructed data types useful for chemistry at this time. AVS on the other
 hand has an extensive data structure defined for chemistry, and one company
 (Molecular Simulations Inc.) even has a commercial molecular modeling product
 based upon AVS. At the moment, we are not using either program. However, I am
 personally interested in the use of SGI Explorer for this purpose, and so would
 appreciate your forwarding any information you may receive related to this as
 a result of your query.
 Thanks,
 Joseph
 --
 Dr. Joseph D. Andose
 Merck Research Laboratories
 Bldg 50SW-100
 126 E. Lincoln Avenue, P.O. Box 2000
 Rahway, NJ 07065
 (908) 594-4391
 (908) 594-4224 (FAX)