From milan@helix.nih.gov Fri Feb 19 18:25:15 1993 Date: Fri, 19 Feb 93 23:25:15 -0500 From: milan@helix.nih.gov (Milan Hodoscek) Message-Id: <9302200425.AA14694@helix.nih.gov> To: G. Ravishanker Subject: Re: Parallel AMBER G. Ravishanker writes: > > Jim > > We have just finished porting Wesdyn (an MD program derived from GROMOS86 > - heavily modified and vectorized for Cray) to run in parallel under > Network Linda. It is premature for me to put this out, but I could not > resist. Our timings show reasonable numbers for first attempt. This > resulted from the work by myself and Tim Mattson (Scientific Computing > Associates who market Linda). We see speedups of 30% for two nodes and 50% > for 3 nodes. Our initial tests were on EM (10-50 steps) on a dodecamer > sequence of DNA with ~2400 water molecules in a hexagonal prizm box with > periodic boundary condition with a cutoff of 11.5 A (7.5 A to 11.5 A > switching function). We are going to run more benchmarks on 5ps of MD, and > on Ethernet vs FDDI. > > We have several things to sort out including the distribution and other > things, so treat this as very preliminary. Things look pretty good at this > point and we are looking at possibilities of further speedups. > We have parallelized CHARMM23 (standard version) program with general approach so the same program runs on variety of platforms, like Intel Hypercube(gamma) or Delta machines and clusters of workstations. We observe the following performance on our standard benchmark (carboxy myoglobin with 3830 water molecules or 14000 atoms, 1 ps or 1000 steps): machine number of processors speedup(*) intel gamma 32 28 intel gamma 128 80 or 60% of theoretical 128 HP/730 2 1.9 HP/730 4 3.7(FDDI) IBM RS/6000-320 4 3.1(ethernet) Some of the details are reported in Chemical Design Automation News, Volume 7, Number 12, December 1992 more will follow elsewhere. Sincerely yours -- Milan Hodoscek From arne@mango.mef.ki.se Sat Feb 20 15:47:00 1993 Date: Sat, 20 Feb 93 15:47:00 GMT From: arne@mango.mef.ki.se (Arne Elofsson) Message-Id: <9302201547.AA11335@mango.mef.ki.se> To: CHEMISTRY@ccl.net Subject: e>mail cc_post1 Subject: Reference I am looking for a reference where Kollman and co-workers (I think) compared the polar/apolar ratio of a protein surfaceas a test of the acuracy for MD-simulations Greatfull for any help arne@mango.mef.ki.se From JWANG@ac.dal.ca Sat Feb 20 13:57:21 1993 Date: 20 Feb 1993 17:57:21 -0400 (AST) From: JWANG@ac.dal.ca Subject: Looking for curve fit program To: chemistry@ccl.net Message-Id: <01GUYC7RT4AQ00H5ZI@AC.DAL.CA> Dear Netters, I need a program to do an arbitrary curve fit for over twenty parameters. It would be best if it can ran on the UNIX system and show a 3D picture in X window. Do you have any idea to obtain this kind of program? Thanks in advance. Jian Wang jwang@ac.dal.ca From rupley!local@cs.arizona.edu Sun Feb 21 02:24:17 1993 Date: Sun, 21 Feb 93 02:24:17 GMT From: rupley!local@cs.arizona.edu Message-Id: <9302210224.AA02075@rupley> To: chemistry@ccl.net Subject: Re: Looking for curve fit program : I need a program to do an arbitrary curve fit for over twenty : parameters. It would be best if it can ran on the UNIX system and : show a 3D picture in X window. Do you have any idea to obtain this kind : of program? Try S, with export of the results to xgobi. Both are AT&T. Xgobi is free, from statlib. S costs you something, for sources from AT&T or a binary from StatSci. If your arbitrary curve is indeed a curve, i.e., a nonlinear function, would you believe a 20 parameter fit, even with >>10*nparm data (:-)? John Rupley Biochemistry rupley@cs.arizona.edu From mail Sat Feb 20 16:29:48 1993 Date: 20 Feb 1993 16:20:32 -0500 (EST) From: spollack@ester.mse.uc.EDU (S.K. Pollack) Subject: CI calculations To: chemistry@ccl.net (OSC Chemistry List Exploder) Message-Id: <9302202120.AA10042@ester> Netters: We are trying to understand the polymerization chemistry of 1,2,3 -butratriene. Part of out study requires that we evaluate the singlet-triplet energy difference for the twisted and planar forms. I am a closed shell type (cations and neutrals) and would like to find some good tutorial level references on CI calculations. We have GAMES, ZINDO and MOPAC-6.0 running here, all capable of various kinds of CI. From conversations with my collegues in Chemistry here at Cincinnati, there are a number of such CI virgins out there who would appreciate such info. There are a number of references in the GAMMESS docs, but the introductory ones are more theoretical than applied. Also the Pople school (of which I am a grandchild) are more oriented to MP and for these molecules I am not sure if the values we are getting are reasonable. Any and all help is appreciated. Bye the way, I will post some c-codes for converting SPARTAN archives and input files to coordinate input file for all the above packages. SPARTAN is a good fron end for these as well. Steve -- _______________________________________________________________________ Steven K. Pollack | University of Cincinnati | Department of Materials Science | I notice that you are still & Engineering | using polymers! 498 Rhodes Hall ML#12 | - Lt. Commander Montgomery Scott Cincinnati, OH 45221-0012 | U.S.S. Enterprise spollack@ester.mse.uc.edu | ________________________________________________________________________