From chemistry-request@www.ccl.net Mon Sep 28 04:48:06 1998 Received: from bioreason8 (dialup9.trail.com [209.194.109.34]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with SMTP id EAA10667 Mon, 28 Sep 1998 04:48:03 -0400 (EDT) Received: from bioreason.com (localhost [127.0.0.1]) by bioreason8 (950413.SGI.8.6.12/950213.SGI.AUTOCF) via ESMTP id CAA08309; Mon, 28 Sep 1998 02:44:11 -0700 Sender: dalke@bioreason.com Message-ID: <360F5A6A.8273635@bioreason.com> Date: Mon, 28 Sep 1998 02:44:10 -0700 From: Andrew Dalke Organization: Bioreason, Inc. X-Mailer: Mozilla 4.05 [en] (X11; U; IRIX 6.2 IP22) MIME-Version: 1.0 To: "LOEFFLER,DR,GERALD FEM BENAT" CC: chemistry@www.ccl.net Subject: Re: CASP vs. The Tower of Babel References: <4195A3C6D39BD11189C30020AFFC17588F3BD9@vieexch1.vie.at.bic> Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Gerald Loeffler stated: > As a concrete example: It would be almost trivial to design > an OO framework for Energy Minimization [...] > > But is the resulting code efficient? It's certainly a lot > slower than FORTRAN! (For language reasons and because you > loose opportunities for source-level optimizations if components > have to be truly independent of each other.) There's some recent (within the last 4 years or so) work done with C++ templates to push the performance of that langauge to near-FORTRAN speeds using a technique called "template expressions". Interestingly, templates are pretty orthogonal to objects, and I think that's a pretty important distinction. Most of the things you're talking about, > minimization strategies, potentials of mean force, integration > algorithms, Monte Carlo moves, Genetic Algorithms are algorithms which are "genericized" using templates, not objects. The biggest project I know of using templates for computational work is POOMA from Los Alamos. I also recall someone from UCSD(?) working on a framework for biomolecular modeling, but I've forgotten both the name of the project and the people working on it. > So why is there no collaboration on the code-level between > computational scientists? I know several reasons for why people might not distribute their sofware: 1) not worth it to them (eg, documentation, proper packaging, support for multiple platforms, setting up web pages and ftp site, etc can be a lot of work, even if you know what you're doing) 2) worried their contributions won't be properly recognized, 3) want to commercialize their work, especially if there is a patent associated with it 4) the software was written and maintained by grad students who want to graduate -- it often is (just) enough to get the work done or there's no single owner/developer that "owns" the software through the transitions. Of course, this is a minature version of the same problem. Then, when someone does make the source available, most of the time it is under what I consider to be a too restrictive license. For example: "cannot be redistributed" means I cannot make changes or improvements to the software and give the changes to others. Sometimes I can send the changes back to the author and have them incorporated into the main development. But what happens if no more development is done on the project? Or if the development path is different than that one I'm taking? "educational use only" or "not for commercial use" There's a couple problems here. I don't really care who uses my software but if I base some of my work on a library that is not for commercial use, I have to restrict with whom I can collaborate. (There's also sticky points like, is a research group in a for-profit hospital non-commercial?) The other big problem is I'm in a company now, so I cannot use that class of software. Yet I develop and release software that others can freely use. This most recently bit me when I decided I wanted to experiment with mmCIF for use in Python (perhaps as part of Konrad Hinsen's MMTK). I didn't want to implement a parser myself so looked at the mmCIF code from Rutgers. It was only for educational use, which means I couldn't even look at the source. Finally, even if the license isn't onerous, I agree that there's still the problem you pointed out: > traditionally computational scientists have been _lousy_ > programmers. This also falls into several categories. The biggest complication is that there are a lot of self-taught programmers in this field while the field of software development/engineering has advanced beyond what most of those people know. The example I gave earlier, template expressions in C++, is one very powerful feature that few computational scientists know about, much less use. Suppose you're a computational chemist (this is CCL, after all :) but learned programming mostly on your own. Here's some problems I can almost guarantee will be in your code, and which make software hard for others to use: 1) non-thread-safe functions 2) no or poor exception detection and handling 3) containers with only one iterator 4) no concern for namespace collision 5) reimplementation of standard library functions 6) writing your own command/control/macro language 7) poor definition of who "owns" certain data elements, especially strings (these are not the standard beginning programming problems of hard- coded array sizes, excessive use of global variables, fragile input parsers, etc.) BTW, my own code contained all of the above problems until I learned why they should be avoided :) Even then, there is rarely good documentation, especially of the internals and reasons for making certain design decisions Another difficulty is realizing that there are common approaches to certain tasks. If you want others to use your software you have to use those approaches. For a basic example, if you write something that implements a binary tree, you should probably have items like "parent", "node", "left_child" and "right_child". Real situations are more complicated than that. People are starting to recognize and classify larger organizational patterns in software (see the "Design Patterns" book). By being aware that you are using those patterns, and explicitly designing your software to support that description, you help others who need to work with your code. I'm willing to bet that few in this forum have read that book, but it is an important one in the software engineering field. Finally, it takes a long time to learn the sense of what makes a good interface and how to tell if what you're developing is specific to you or is useable with somebody else's view of the world. (My common example here is a visualization program good for biomolcules but has no concept of bond order or ring detection so poor for chemists.) While I'm here, let me point out a problem I've had. I like working on software "infrastructure"; how to tie different pieces of software together into a larger whole. I think the computational chemistry needs some work along these lines, and the lack of a set of commonly used libraries for this field is helping restrict new development. Who will pay for this work? Or, in a university environment, there's not enough physical science to make it worth a chemistry or physics degree, nor enough computer science to make it worth a computer science degree. > (Regarding unwillingness to give away source code at all: With Java > you can even give away the "object code" (byte code) (maybe after > running it through an obfuscator) and it will work on any Java > platform...) Please no. I ask that this not be common. Let me quote from the QCPE distribution here: > Commercial software firms will usually > only make executable code available. As QCPE > is attempting to promote the development of the > field of computational chemistry it has always > been our philosophy to provide source code. > Our reasoning for this is simple. Unless the > user has source code one has no idea about the > science or should one say "alleged" science > which is contained in the system. > > Richard W. Counts/QCPE Andrew Dalke dalke@bioreason.com --- Disclaimer: Opinion only, and not necessarily the viewpoint of my employers nor even myself all the time. From chemistry-request@www.ccl.net Mon Sep 28 10:59:13 1998 Received: from earth.ox.ac.uk (darwin.earth.ox.ac.uk [163.1.22.6]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id KAA12393 Mon, 28 Sep 1998 10:59:12 -0400 (EDT) Received: from metropolis.earth.ox.ac.uk (metropolis [163.1.22.59]) by earth.ox.ac.uk (8.9.1a/8.9.1) with SMTP id PAA28864 for ; Mon, 28 Sep 1998 15:59:08 +0100 (BST) From: Keith Refson MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Message-ID: <13839.42044.95834.577681@metropolis> Date: Mon, 28 Sep 1998 15:59:08 +0100 (BST) To: "Comp. Chem. List" Subject: CCL:Fitting a potential by a family of functions... In-Reply-To: <360B95AC.167E@cennas.nhmfl.gov> References: <360B95AC.167E@cennas.nhmfl.gov> X-Mailer: VM 6.62 under Emacs 20.3.1 Content-Transfer-Encoding: 7bit boufer writes: > I want to fit a potential~(of the Coulomb type) by some family of > functions. My problem is, I cannot use the classical least square method > in which we start by writing : > > D = Int (True Potential - approximation)^2 dv > > since in this case the true potential is not square integrable and hence > yield to a divergence. Just remove the divergence using a weighting function, eg D = Int ((True Potential - approximation)/r)^2 dv will remove the r -> 0 divergence. You will need to devise something more complicated to include this as well as the r -> oo divergence Keith Refson -- ------------------------------------------------------------------------------ Email: Keith.Refson@ | Tel: +44 1865 272026 | Dr Keith Refson, | earth.ox.ac.uk | Fax: +44 1865 272072 | Dept of Earth Sciences | Spam: root@cyberpromo.com | Parks Road, Oxford OX1 3PR, UK| ------------------------------------------------------------------------------ From chemistry-request@www.ccl.net Mon Sep 28 11:25:11 1998 Received: from bioc1.msi.com (bioc1.msi.com [146.202.0.2]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with SMTP id LAA12579 Mon, 28 Sep 1998 11:25:10 -0400 (EDT) Received: by bioc1.msi.com (931110.SGI/930416.SGI) for chemistry@www.ccl.net id AA28418; Mon, 28 Sep 98 08:19:35 -0700 Received: from mica.msi.com(146.202.6.217) by bioc1.msi.com via smap (V2.0) id xma028407; Mon, 28 Sep 98 08:19:32 -0700 Received: from jnauss-pc.msi.com by mica.biosym.com.msi.com (4.1/SMI-4.1) id AA13270; Mon, 28 Sep 98 08:32:02 PDT Message-Id: <3.0.3.32.19980928081810.006b0e4c@146.202.6.217> X-Sender: jnauss@146.202.6.217 X-Mailer: QUALCOMM Windows Eudora Pro Version 3.0.3 (32) Date: Mon, 28 Sep 1998 08:18:10 -0700 To: Bernd Engels , chemistry@www.ccl.net (Computational Chemistry List) From: "Jeffrey L. Nauss" Subject: Re: CCL:books for theor.chem. In-Reply-To: <199809231236.OAA15662@rs5.thch.uni-bonn.de> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" At 02:36 PM 9/23/98 +0200, Bernd Engels wrote: >I would be interested in an overview about >good books about Theoretical Chemistry. I am >mainly interested in books for undergraduate >students but other comments are also >highly appreciated. I have found the following to be most useful: 1. "Computational Chemistry", 1995, GH Grant and WG Richards, Oxford, Oxford University Press A thin book (92 pages) but provides a nice overview of the field. Probably most suited for undergraduates. 2. "Molecular Modeling: Principles and Applications", AR Leach, 1996, Essex, Addison Wesley Longman Limited A much more robust book (595 pp) that goes into more detail than Grant and Richards. Gives more examples and a more solid theoretical foundation for the techniques. My choice for a graduate level textbook. 3. "Dynamics of Proteins and Nucleic Acids," JA McCammon and SC Harvey, 1987, Cambridge, Cambridge University Press. Always a classic - an excellent (though a little dated) description of simulations of proteins and nucleic acids. Good low level introductions to theory with some nice examples and applications. 4. "Computer Simulation of Liquids", MP Allen and DJ Tildesley, 1987, Oxford, Clarendon Press. Only for the hardcore simulation junkies. An excellent description of the nuts and bolts of simulation code. Would be excellent for a course where the students will write programs for simulations. 5. "Forcefield-Based Simulations," April 1997, San Diego, Molecular Simulations Inc. 6. "Discover 2.9.8/96.0/4.0.0. Forcefield Simulations: User Guide, Part 1", September 1996, San Diego, Molecular Simulations Inc. In my experience with teaching molecular modeling at the University of Cincinnati, I found these user guides from MSI to be very helpful. They give a clear explanation of forcefield based simulations. -- Jeffrey L. Nauss, PhD Phone: (619) 799-5555 Product Specialist, Simulations Fax: (619) 458-0136 Molecular Simulations Inc. E-mail: jnauss@msi.com 9685 Scranton Road San Diego, CA 92121 From chemistry-request@www.ccl.net Mon Sep 28 12:45:59 1998 Received: from gatekeeper.es.dupont.com (gatekeeper.es.dupont.com [192.26.233.2]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id MAA12985 Mon, 28 Sep 1998 12:45:59 -0400 (EDT) From: fredvc@esalp2.dnet.dupont.com Received: from ns1.es.dupont.com (ns1.es.dupont.com [138.196.90.13]) by gatekeeper.es.dupont.com (8.9.0.Beta3/8.9.0.Beta3) with SMTP id MAA04967 for ; Mon, 28 Sep 1998 12:45:58 -0400 Received: by ns1.es.dupont.com; id AA12093; Mon, 28 Sep 98 12:45:56 -0400 Message-Id: <9809281645.AA12093@ns1.es.dupont.com> Received: from esalp2.dnet; by esds01.dnet; Mon, 28 Sep 98 12:45:57 EDT Date: Mon, 28 Sep 98 12:45:57 EDT To: smtp@"boufer@cennas.nhmfl.gov" Cc: fredvc@esalp2.dnet.dupont.com, chemistry@www.ccl.net Apparently-To: chemistry@www.ccl.net Subject: RE: CCL:Fitting a potential by a family of functions... Fitting a potential by a family of functions........ If there is a problem wiht the function not be square-integrqable, you will have to discretize the true potential and perform your least squares fit using summatiions rather than integrations. ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ FREDERIC A. VAN-CATLEDGE Scientific Computing Division || Office: (302) 695-1187 or 529-2076 Central Research & Development Dept. || The DuPont Company || FAX: (302) 695-9658 P. O. Box 80320 || Wilmington DE 19880-0320 || Internet: fredvc@esvax.dnet.dupont.com -------------------------------------------------------------------------------- Opinions expressed in this electronic message should ***> NOT <*** be taken to represent the official position(s) of the DuPont Company. *****> ANY OPINIONS EXPRESSED ARE THE PERSONAL VIEWS OF THE AUTHOR ONLY. <***** ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ From chemistry-request@www.ccl.net Mon Sep 28 15:34:30 1998 Received: from mail3.cityweb.de (mail3.cityweb.de [193.189.224.211]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id PAA19066 Mon, 28 Sep 1998 15:34:28 -0400 (EDT) Received: from cityweb.cityweb.de (ess8-m103-96.pool.cww.de [195.71.103.96]) by mail3.cityweb.de (8.9.1/8.9.1/powered by Telemedia) with SMTP id VAA01676 for ; Mon, 28 Sep 1998 21:34:27 +0200 (MET DST) Message-Id: <199809281934.VAA01676@mail3.cityweb.de> X-Sender: warzecha@mail.essen.cityweb.de X-Mailer: QUALCOMM Windows Eudora Pro Version 4.0 Date: Mon, 28 Sep 1998 21:33:46 +0200 To: chemistry@www.ccl.net From: "Klaus-D. Warzecha" Subject: CCL:RE:help! (was: plotting of surfaces with VB) In-Reply-To: <199809251524.MAA00341@huckel.ifi.unicamp.br> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by www.ccl.net id PAA19068 At 12:24 25.09.98 -0300, Marcio Cyrillo wrote: >[...] 2 - Having a collection of data of the following type: variable1, variable2, energy(variable1, variable2), is there a free (public) routine to plot the corresponding surface? [...] Marcio, among the controls shipped with VB, graph.ocx + graph server probably won't work for potential surfaces, whereas MSChart.ocx might. It's however not the simplest they ever made and the documentation could indeed be a bit better. A commercial alternative might be a package named "First Impression" by Visual Components. Again, never used it or saw it in action myself, but you might check their web site at http://www.visualcomp.com for further information. Best wishes, Klaus -- Klaus-D. Warzecha Max-Planck-Institut für Strahlenchemie P. O. Box 101365 D-45413 Mülheim an der Ruhr Germany fax: x49/x208-306-3951 email: warzecha@cityweb.de From chemistry-request@www.ccl.net Mon Sep 28 16:09:36 1998 Received: from rzusuntk.unizh.ch (rzumail2.unizh.ch [130.60.128.10]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id QAA22807 Mon, 28 Sep 1998 16:09:35 -0400 (EDT) Received: [from rzusgi.unizh.ch (chburger@rzusgif.unizh.ch [130.60.68.19]) by rzusuntk.unizh.ch (8.8.5/SMI-5.25) with SMTP id WAA10814 for ; Mon, 28 Sep 1998 22:07:32 +0200 (MET DST)] Date: Mon, 28 Sep 1998 22:06:28 +0200 (MDT) From: "Dr. Peter Burger" To: CHEMISTRY@www.ccl.net Subject: parallel ADF-PVM on Linux/Intel? Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CCLers, has anyone of you successfully installed the parallel version of the DFT program ADF running on Linux/Intel with either PVM or MPI and is willing to share his/her experience/provide hints? Many thanks in advance Peter -------------------------------------------------------- Peter Burger University of Zuerich chburger@aci.unizh.ch From chemistry-request@www.ccl.net Mon Sep 28 18:26:46 1998 Received: from cs.CS.NMSU.Edu (cs.CS.NMSU.Edu [128.123.64.2]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id SAA05137 Mon, 28 Sep 1998 18:26:44 -0400 (EDT) Received: from oso.cs.nmsu.edu (ajones@oso [128.123.64.79]) by cs.CS.NMSU.Edu (8.8.6/8.8.6) with SMTP id QAA02620 for ; Mon, 28 Sep 1998 16:26:39 -0600 (MDT) Date: Mon, 28 Sep 1998 16:26:39 -0600 (MDT) From: Abraham Jones To: chemistry@www.ccl.net Subject: Plotting Contours! Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Greetings, Hello to all fellow users of gnuplot. I want to say a big *thank you* very much to all the people who answered my post about contouring data with gnuplot. :-) Just about all of my questions where answered and I'm now generating contour plots with gnuplot. I must say that Sigma Plot generates much nicer contour plots and it's much easier to use. All of my f77 number crunchers run on a unix machine. Now I can use gnuplot to fine tune the contours and Sigma Plot generates the final pretty print for class. This semester I'm taking a class in low speed aero and we're doing a lot of potential flow; ergo lots of contour plots. I wish the school would break down and get Sigma Plot for Unix machines. Is it that expensive? Anyway, again many thanks to all the people who sent me tips on how to use gnuplot. Abraham Jones