From chemistry-request@ccl.net Mon Aug 26 10:02:14 1991 From: rs0thp@rohmhaas.com (Dr. Tom Pierce) Subject: Re: IUPAC & theoretical chemistry To: Chemistry@ccl.net Date: Mon, 26 Aug 91 9:36:22 EDT Status: R I wish to state that I am not categorically against the resolving of standards for describing Molecular structure. At the moment the ASTM subcommittee E49.51 is attempting to set (describe)a standard for chemical structure information exchange. What I am against is the rush to 'set' standards. I have heard that there are questions about the Crystallographic Information File format, specifically about it missing data that is not used in crystallography but is used in substructure searching. The American Mathematics Society and much of the Physics community use as a 'standard' TEX for journal submission doesn't prove that the chemistry journals think TEX is sufficient or should be a standard. The problems that chemical journals have with solely using TEX shows that setting standards should be done carefully and with much testing. I suggest that we (computational chemists) try-out the different representations and experimentally discover what works and what doesn't work. Let's not try to set Internatinal Standards before we have found workable national standards. -- -------------------------------------------------------------------- Sincerely, Thomas Pierce Computer Applications Research | rs0thp@rohmhaas.com Internet Bldg 64C | rs0thp@rohvm1 Bitnet Rohm and Haas Co. | (215)-785-8989 Voice Bristol, PA 19007 | (215)-781-4204 Fax < Artificial Intelligence is Better than None! - AAAI Slogan > From chemistry-request@ccl.net Mon Aug 26 10:12:03 1991 Date: Mon, 26 Aug 91 09:51:04 EDT From: enrico@watson.ibm.com To: CHEMISTRY@ccl.net Subject: Program for building coiled coils. Status: R Does anybody has a program that can build a coiled coil structure and give me the cartesian coordinates of the atoms starting from the sequence of aminoacids of the two chains ? Any help will be appreciated. Thanks in advance. Dr. Omar G. Stradella Center for Scientific Engineering Computations IBM Corp. Kingston, NY Internet: enrico@watson.ibm.com BITNET: ENRICO@YKTVMZ From jkl@ccl.net Mon Aug 26 10:47:30 1991 Date: Mon, 26 Aug 91 10:16:56 -0400 From: jkl@ccl.net Subject: Re: IUPAC and Theoretical/Computational Chemistry To: chemistry@ccl.net Status: R I am forwarding the message to the list: ---------- Begin Forwarded Message ---------- >From EDELSON@EVAX3.ENG.FSU.EDU Mon Aug 26 09:56:02 1991 Date: Mon, 26 Aug 1991 9:58:38 EDT From: EDELSON@EVAX3.ENG.FSU.EDU Subject: Re: IUPAC and Theoretical/Computational Chemistry To: chemistry-request@ccl.net With regard to the comments on Journals accepting papers on floppies: ACS had a symposium on this problem several years ago, and the conclusions reached at that time are still valid today, viz: 1. It is much easier for the Journal to enforce its editorial standards by marking up hard copy than by editing floppies which may have been prepared by a wide variety of incompatible word or text processors... 2. The cost of keyboarding in from hard copy represents only a small fraction of the production cost of a Journal.\ 3. There is only one Journal which is produced by conventional printing processes (i.e. not photocopied from manuscript) which accepts computer- compatible manuscripts, and that is Computer and Mathematics, with Applications.This Journal accepts manuscripts in TEX, and I suspect that is only because the Editor, Ervin Rodin, runs them off himself for photocopying by the publisher. Other journals in the Computers & series do not accept such input, because the publisher, Pergamon, does not have facilities for handling TEX. The more pertinent problem is distribution of software described in published articles. Computers and Chemistry has discontinued publishing complete codes, and leaves this distribution to individual arrangements between author and reader. A condition of publication is that the author agree to honor all such requests. David Edelson, editor, Computers and Chemistry ----------- End Forwarded Message ----------- From chemistry-request@ccl.net Mon Aug 26 11:48:59 1991 From: Elizabeth Chelkowska Subject: Eigenvectors To: CHEMISTRY@ccl.net Date: Tue, 27 Aug 91 1:37:30 EST Status: R Dear fellow netters, I would like to use Gaussian90 with STO-3G basis to obtain eigenvectors in orthogonal basis as well as Coulomb,exchange and some special integrals for valence orbitals. My molecules (with experimental geometries) are very small, therefore I can risk storage of all integrals. I've spent few days trying to get the required output without any luck. I would appreciate if someone could give me hints how to prepare input for Gaussian90 to get printout of: 1) molecular orbitals in orthogonal basis 2) charge distribution in terms of molecular orbitals 3) molecular J(PP,QQ), K(PQ,PQ) and J(PQ,RS), K(PS,RQ) for valence orbitals. Thank you in advance Elizabeth =============================== ryba@tasman.cc.utas.edu.au || =============================== From chemistry-request@ccl.net Mon Aug 26 16:26:37 1991 Date: Mon, 26 Aug 91 15:09:24 -0500 From: milfeld@hermes.chpc.utexas.edu (Kent Milfeld) To: chemistry@ccl.net Subject: ibm-6000 general info Status: R Response to Venkat about IBM-6000: OUR SITE BOUGHT 2 IBM RISC machines: Two new RISC-architecture machines have been installed at CHPC (The Center for High Performance Computing, The University of Texas System). Both are IBM RISC System/6000 POWERservers, model 550S. The purpose of these machines at at our site is to provide a platform for users to develop and execute scalar-bound programs. Because the RISC System/6000 computer (RS/6000) employs several advanced architectural and "superscalar" features, it can execute up to four instructions per cycle. For scalar-bound code this means that an RS/6000 processor does a good job when competing against a single Y-MP processor; several of our benchmarks confirm this. For those problems and codes which do not map well into vector or course-grain parallel methods some users may want to explore the POWER (Performance Optimized With Enhanced RISC) architecture as an alternate processing approach. Through instruction scheduling of the compilers, fine grain parallelisms within scalar-bound programs are transformed into "superscaler" code which takes advantage of the instruction-level parallelism within the hardware. ARCHITECTURE: The architecture includes some of the concepts found in CISC machines, and uses 184 instructions. To a purist this may not be a RISC machine, but to others it is a second generation RISC design. The increased number of intructions is due to a higher degree of pipelining and parallelism found in the POWER architecture. However, the number of instructions represents an optimal size for the product (cycles per instruction) * (path length), a design parameter which IBM believes is a measure of a machine's true performance. CONFIGURATION at our site: Our machines have the following configuration and important features: Memory: 128 Mbyte (Physical) Disk: 860 Mbyte Internal Drive 2 x 1.3 Mbyte External Drives (NPI, National Peripherals, Inc.) Clock Cycle: 41 MHz Operating System: AIX version 3.1 Floating Point Unit: IEEE Format Double Precision Floating-Point Engine Thirty-two 64-bit Floating-Point Registers Memory caching: Separate Data and Instruction Caches (Base price for 550 with 2x32Mb RAM, 800MB disk is ~$37,000) (NPI disks, we bought ~4 Mbyte for ~$15,000) Manuals: A complete set of documentation for the RISC System/6000 POWERserver is supplied on a CD-ROM, and is available either through the "man" page command or the InfoExplorer system. The InfoExplorer system (Info) provides documentation through an information hierarchy of hypertext which uses the X Window System as a display platform. A tutorial and guide window have been set (at our site) to automatically appear when the InfoExplorer program is initiated. SOFTWARE: Our site has the following software: IBM AIX XL FOTRAN: xlf IBM AIX C cc, xlc, c89 & (bsdcc, added by CHPC) Graphics X Window System Version X11R4 (AIX X-Windows) graPHIGS Batch Queue NQS Science Libary ESSL (I suggest you purchase this library) In general, quantum chemistry codes have been shown to execute well on the RS/6000 machines. We have installed three quantum chemistry codes on these two machines, CADPAC, Gaussian-90 and GAMESS. Later, I'll have some benchmarks for these programs on our new machines. GRAPHICS: The IBM use the Silicon Grapics, Inc. (SGI) gl graphics language, and provides PHIGS on top of it. Also, IBM provides the Motif and Xt toolkits with the X Windows System. Polygen has a beta release on the IBM machines for Quanta/Charmm. I believe it is being released now, or very soon. BENCHMARKS: By scaling some results from our old X-MP/14se machine and using some results from an IBM 540 (Feb. 1991) we found Gaussian-88 to run about 2.5X slower than a single processor Y-MP (for SCF and SCF optimizations). For 7 different problems the ratios were between 1.6 and 4.9. Two scalar-bound codes ran only 2.7 and 1.4X slower. However, vector code certainly does not enjoy such a significant comparison. Margaret L. Simmons and Harvey J. Wasserman wrote a paper entitled: "Los Alamos Experiences with the IBM RISC SYSTEM/6000 Workstations" (LA-11831-MS)(mls@lanl.gov). In it, various vector operations (v=v+s, v=v+v, v=v*v+v, v=v*v+v*v, v=s*v+s*v, etc.) achieved MFLOPS ratings in the range of 4 to 22, depending upon the vector length and other factors. Just to give you an idea of what a Cray Y-MP can do, in the CHPC Newsletter, Vol. 6, no. 6 the results of matrix multiplies (using 3 DO loops) were shown for matrices of order 767 to 801. There are 6 different ways to index (vectorize) the problem. When bank and section conflicts were not a problem, the six different methods ranged >from about 150 to 200 MFLOPS. With various "canned" routines you can achieve close to 300 MFLOPS on a single processor (That is why they call the Y-MP a Super Computers!). Some references: "RS/6000 Storage System to Debut" UNIX Today! June 24, 1991 (If you need lots of disk space) "IBM's RISC LINEUP: PACKED WITH POWER" Digital Review, October 1, 1990. (Benchmarks -- on "old" machines) "A look at IBM's New Filesystem", UNIX Today!, October 1, 1990. (Has some good info about caching.) If you need to have lots of disk space for a single file, say 1Gbyte for quantum chemistry codes, make sure the that your workstation and file system can support this. On some disk systems you cannot write across devices--if you have a string of disks, each of size X, you may not be able to write a file which is larger than X. On the IBM you logically "connect" disks an overcome this limitation. I hope that this information will help. Kent Milfeld Computation Chemist Center for High Performance Computing mileld@chpc.utexas.edu PS Don't rule out HP's new machine(s). From chemistry-request@ccl.net Mon Aug 26 16:48:23 1991 Date: Mon, 26 Aug 91 15:31:53 CDT From: shepard@dirac.tcg.anl.gov (Ron Shepard) To: chemistry@ccl.net Subject: Machine-Readable Manuscript Submissions to Journals Status: R I think that the time has come for support of machine-readable submission of manuscripts. I have been submitting Mac floppies along with my manuscripts for 6 years now, and judging by the amount of proofreading that is still required at the galley stage, they have never been used by a publisher. However, I'm not sure of the role of IUPAC, or any other standards organization, in this effort. Basically, I think that as soon as word processors are sufficiently easy to use (spell checking, WYSIWYG, etc.), sufficiently powerful to do what is required (text formatting, graphics, equations, etc.), and sufficiently popular among chemists, then journals should accept documents in that format. This level of capability was reached about 6 years ago in the Mac world, and I expect it is by now not too far away in the PC/Windows or workstation/Xwindows worlds. I'd bet that with sufficient pressure from the user community, the capability to read various formats would be provided to the journal publishers by the various word processing software vendors, so there would be essentially no external costs associated with the software. This might be facilitated with the (implicit) threat that their software would not be used unless they adopt this policy. Also, if the publisher were to provide, say, a 1-line statement at the end of each manuscript (e.g. "This manuscript was prepared and formatted using QWERTY version 12.34."), this would keep everyone happy. At any rate, we're talking $50-$100 for software here for each program (e.g. typical Mac street prices for Word, MacWrite, FullWrite, WriteNow, etc.), not the $1000-$50000 for high-end publishing systems that are being used now (e.g. Publisher, Framemaker, etc.). The hardware question might be handled in the same manner (e.g. "This manuscript... on an Apple Macintosh."). Any other suggestions? One thing that should be avoided is the specification of a SINGLE standard. For example, although there are many TEX devotees in the unix world, others consider it a step backward into the 70's. Individuals like to stick with whatever they're using for a few years, or longer depending on how much effort they have invested in learning it, before switching to a newer technology. Cabilities of individual WP programs also change; MacWrite circa 1984 is not the same as MacWrite circa 1991, and this tends to keep users loyal. Another aspect of this is that OSI, and in particular GOSIP, the government-mandated version, also contain some document-preparation specifications that are due to kick into effect over the next couple of years. This could either be the answer to the problems, by defining the low level file formats or at least standardizing an interchange format, or it could be a situation of trying to freeze the technology at too early of a level (such as ADA or Xwindows). In the case of ADA, which was mandated by DOD in the late 70's, it developed the support of neither programmers nor vendors, and has not flourished in the past 15 years. With Xwindows, which was based on bitmap-level technology instead of higher-level primitive descriptions, it will be this fall at the earliest before outline fonts and object-oriented drawing, using PEX, will be included, and even these are low-level tools from a scientist's point of view; consequently it will probably be another few years before these features are mature and fully supported -- about a decade to undo the original mistake. It might not have been possible to predict these trends when the standards were defined originally (well, maybe they were, but remember, things are always much clearer in hindsight). Anyway, it calls into question the idea of inviting a standards organization to define how you should do your work. Finally, one other thing should be mentioned in this context. With machine-readable documents, it will also be possible to "publish" journals completely in electronic form, with nary a piece of paper between author and reader. This is beginning to occur even now as authors place their manuscripts on anonymous FTP servers. If this becomes commonplace, and if journals continue to refuse to accept machine-readable documents, maybe authors will just bypass the journal publication step entirely. That is, if it's too much of a hassel to proofread the galleys and pay for page charges and reprints, then why bother? This could have interesting consequences on Science (cap "S"), perhaps returning us to a time when the primary information exchange involved personal correspondence (which could be good), but reducing or eliminating the peer-review step (which could be bad). Also, of course, exactly how would you include these in your resume; and if you can't include them in your resume, then how do you get a job, or submit a grant...? So to the journals: Please start accepting machine-readable WP manuscripts in order to save science as we know it. -ron shepard shepard@tcg.anl.gov [Obviously, these views are personal, and I beg the indulgence of vendors whose products I've mentioned without (T) or (R) or (C) designations.] From chemistry-request@ccl.net Mon Aug 26 21:54:50 1991 From: bernhold@orange.qtp.ufl.edu Date: Mon, 26 Aug 1991 21:43:00 -0400 To: chemistry@ccl.net Subject: electronic publishing Status: R An idea prompted by the discussion of electroinc submission of manuscripts... (Wouldn't it be nice if...) The editors or publisher of a journal would make available (electronically) a bibliography of articles they publish in a "standard" form such as unix refer or BibTeX. Ideally it would be updated with each issue of the journal and available by anonymous FTP or simiilar method. Yes, this is like what Current Contents and the disk version they produce, but the purpose is different: To avoid the tedious and error-prone task of hand entering references into a paper. Current Contents on Disk probably has most of the information, but I doubt if the format is publically available (though I have _not_ checked). Also, they "sanitize" the text so that ther are no diacriticals, superscripts, subscripts, etc. which makes it less than ideal. I suggest refer or BibTeX because they are fairly common in some circles and I am not familiar with the bibliographic capabilities of an PC-based wordprocessors. They have mechanisms for representing special characters and methematical symbols. They both function by maintaining one or more text files comprising the bibliography and paper itself is kept separately, refering to the citations using a short key. I'm sure they're not the only ones with these capabilities, so remember, its just an _idea_. I wonder how the Current Contents folks would take to this? Is it possible they have some kind of arrangement which gives them exclusive to this kind of data from publishers? I have no idea how an idea like this would effect Current Contents sales, since you'd have to spend a chunk of time collecting all of the bibliographies in order to equal Current Contents each week. -- David Bernholdt bernhold@qtp.ufl.edu Quantum Theory Project bernhold@ufpine.bitnet University of Florida Gainesville, FL 32611 904/392 6365