From chemistry-request@server.ccl.net Fri Dec 14 23:26:59 2001 Received: from rwcrmhc52.attbi.com ([216.148.227.88]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBF4Qxi20870 for ; Fri, 14 Dec 2001 23:26:59 -0500 Received: from C1353359A ([12.228.114.158]) by rwcrmhc52.attbi.com (InterMail vM.4.01.03.27 201-229-121-127-20010626) with SMTP id <20011215042636.LZPT403.rwcrmhc52.attbi.com@C1353359A> for ; Sat, 15 Dec 2001 04:26:36 +0000 Reply-To: From: "Mark Thompson" To: Subject: Woodward Hoffman citations Date: Fri, 14 Dec 2001 20:30:06 -0800 Message-ID: <000001c18521$2c6472c0$9e72e40c@attbi.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook IMO, Build 9.0.2416 (9.0.2911.0) Importance: Normal X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4807.1700 Can someone give me the original journal citation(s) to Woodward and Hoffman's Nobel prize winning work? Thanks in advance, Mark ================================= Mark Thompson Planaria Software Seattle, WA. http://www.planaria-software.com ================================= From chemistry-request@server.ccl.net Sat Dec 15 08:37:22 2001 Received: from prserv.net ([32.97.166.31]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFDbMi26438 for ; Sat, 15 Dec 2001 08:37:22 -0500 Received: from attglobal.net (slip-12-64-6-95.mis.prserv.net[12.64.6.95]) by prserv.net (out1) with SMTP id <2001121513370220102on4kme>; Sat, 15 Dec 2001 13:37:03 +0000 Message-ID: <3C1B0E1D.2215B29D@attglobal.net> Date: Sat, 15 Dec 2001 08:47:25 +0000 From: jmmckel@attglobal.net Reply-To: jmmckel@attglobal.net X-Mailer: Mozilla 4.73 [en]C-CCK-MCD NSCPCD473 (WinNT; U) X-Accept-Language: en MIME-Version: 1.0 To: chemistry@ccl.net Subject: F77 to F90/95 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit CCLers, On trying the free Intel Linux compiler, largely for F90/F95, on F77 code the warnings about F77 are quite copious. Is there an F77->F90/95 utility that will help clear up these warnings? Is there a means to de-VAXinate F77 code easily? Thanks! I'll post the results if there is interest. John McKelvey From chemistry-request@server.ccl.net Sat Dec 15 12:11:45 2001 Received: from mail.bancorp.ru (root@[195.239.131.24]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFHBii28522 for ; Sat, 15 Dec 2001 12:11:44 -0500 Received: from 217.107.75.16 ([217.107.75.16]) by mail.bancorp.ru (8.11.3/8.11.3) with ESMTP id fBFHBOd25466 for ; Sat, 15 Dec 2001 20:11:24 +0300 Date: Sat, 15 Dec 2001 20:11:20 +0300 From: Gregory Shamov X-Mailer: The Bat! (v1.45) Educational Reply-To: Gregory Shamov X-Priority: 3 (Normal) Message-ID: <18024905875.20011215201120@bancorp.ru> X-Confirm-Reading-To: Disposition-Notification-To: gas5x@bancorp.ru To: chemistry@ccl.net Subject: disk or diskless beowulf clusters? Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Hello All, Does someone have experience in using quantum chemistry programs like GAMESS or Gaussian on diskless Beowulf Linux clusters? What is better, in general -- to write integral files on local HDD or to send them to fileserver via NFS on Fast or Gigabit Ethernet? How many diskless nodes could possibly be served with common x86 PC-based fileserver like a dual P3 or Athlon with IDE or SCSI RAID ? Thank you in advance. I will summarize the results. -- Best regards, Gregory Shamov, Dept. Phys. Chem, Kazan State University Kazan, Russian Federation mailto:gas5x@bancorp.ru From chemistry-request@server.ccl.net Sat Dec 15 14:04:38 2001 Received: from gandalf.cber.nih.gov ([128.231.52.5]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFJ4ci29755 for ; Sat, 15 Dec 2001 14:04:38 -0500 Received: from localhost (rvenable@localhost) by gandalf.cber.nih.gov (980427.SGI.8.8.8/980728.SGI.AUTOCF) via ESMTP id NAA71558 for ; Sat, 15 Dec 2001 13:58:39 -0500 (EST) Date: Sat, 15 Dec 2001 13:58:39 -0500 From: Rick Venable To: CHEMISTRY@ccl.net Subject: Linux kernel 2.4 parallel scaling Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Myself and others here at NIH have been benchmarking CHARMM and other parallel chemistry codes on Linux clusters using MPICH libs with an ethernet network. Tests on 2.4.12 kernel nodes with 2, 4, and 8 procs showed terrible parallel scaling, compared to the same tests run with the 2.2.16-tcpfix kernel. I'd expect that other people using Linux clusters for computational chemistry may have encountered this already, and may have looked into possible solutions. For instance, are there some kernel parameters in /proc that can be used to tune the communications performance? Are there any other software solutions to this problem? Suggestions appreciated; I will summarize. =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= Rick Venable FDA/CBER/OVRR Biophysics Lab 1401 Rockville Pike HFM-419 Rockville, MD 20852-1448 U.S.A. (301) 496-1905 Rick_Venable@nih.gov ALT email: rvenable@speakeasy.org =+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+=+= From chemistry-request@server.ccl.net Sat Dec 15 15:35:32 2001 Received: from socrates.cgl.ucsf.edu ([128.218.27.3]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFKZWi30856 for ; Sat, 15 Dec 2001 15:35:32 -0500 Received: (from ross@localhost) by socrates.cgl.ucsf.edu (8.9.3/8.9.3/GSC1.7) id MAA308229 for CHEMISTRY@ccl.net; Sat, 15 Dec 2001 12:35:12 -0800 (PST) Date: Sat, 15 Dec 2001 12:35:12 -0800 (PST) From: Bill Ross Message-Id: <200112152035.MAA308229@socrates.cgl.ucsf.edu> To: CHEMISTRY@ccl.net Subject: Re: CCL:Linux kernel 2.4 parallel scaling Myself and others here at NIH have been benchmarking CHARMM and other parallel chemistry codes on Linux clusters using MPICH libs with an ethernet network. Tests on 2.4.12 kernel nodes with 2, 4, and 8 procs showed terrible parallel scaling, compared to the same tests run with the 2.2.16-tcpfix kernel. ...are there some kernel parameters in /proc that can be used to tune the communications performance? Are there any other software solutions to this problem? In another context and programming in java, I noticed that turning off Nagle's algorithm on the application side by setting sockets to 'nodelay' improved network performance considerably. Watching the low-level TCP/IP traffic, we saw that for a send of an application-level msg, the protocol went from 4 sequential packets with acks to a blast of 4 packets with a single ack coming back. When there is any network delay, the sequential mode amplifies it, more than I would have expected. I'm not sure how this maps to e.g. C - I couldn't find anything in 'man setsockopt' for example. Looking up "Nagle's algorithm" & linux on the web, http://mosquitonet.stanford.edu/~laik/projects/nagle_test/nagle_test.html "This program is useful because some TCP/IP stacks (in particular Linux pre-2.0.30) do not correctly implement Nagle's Algorithm. " So there is some precedent for problems in this area, and the other web references in such a search look useful. I wonder if Nagle's algorithm may be better adapted to network conditions that are more lossy than usually seen these days, or for slower computers. If so, I could imagine that it might be disabled by default in the kernel, at least sometimes, which could explain the variable results w/ different kernels. Bill Ross From chemistry-request@server.ccl.net Sat Dec 15 16:16:19 2001 Received: from socrates.cgl.ucsf.edu ([128.218.27.3]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFLGJi31279 for ; Sat, 15 Dec 2001 16:16:19 -0500 Received: from socrates.cgl.ucsf.edu (localhost [127.0.0.1]) by socrates.cgl.ucsf.edu (8.9.3/8.9.3/GSC1.7) with ESMTP id NAA309717; Sat, 15 Dec 2001 13:15:59 -0800 (PST) Message-Id: <200112152115.NAA309717@socrates.cgl.ucsf.edu> To: Bill Ross cc: CHEMISTRY@ccl.net, dek@cgl.ucsf.EDU Subject: Re: CCL:Linux kernel 2.4 parallel scaling In-Reply-To: Your message of "Sat, 15 Dec 2001 12:35:12 PST." <200112152035.MAA308229@socrates.cgl.ucsf.edu> Date: Sat, 15 Dec 2001 13:15:59 -0800 From: David Konerding Bill Ross writes: > Myself and others here at NIH have been benchmarking CHARMM and other > parallel chemistry codes on Linux clusters using MPICH libs with an > ethernet network. Tests on 2.4.12 kernel nodes with 2, 4, and 8 procs > showed terrible parallel scaling, compared to the same tests run with > the 2.2.16-tcpfix kernel. > > ...are there some kernel parameters in /proc that can be used > to tune the communications performance? > > Are there any other software solutions to this problem? > >In another context and programming in java, I noticed that turning >off Nagle's algorithm on the application side by setting sockets >to 'nodelay' improved network performance considerably. Watching >the low-level TCP/IP traffic, we saw that for a send of an >application-level msg, the protocol went from 4 sequential packets >with acks to a blast of 4 packets with a single ack coming back. >When there is any network delay, the sequential mode amplifies it, >more than I would have expected. > >I'm not sure how this maps to e.g. C - I couldn't find anything >in 'man setsockopt' for example. Looking up "Nagle's algorithm" >& linux on the web, > > http://mosquitonet.stanford.edu/~laik/projects/nagle_test/nagle_test.html > > "This program is useful because some TCP/IP stacks (in particular > Linux pre-2.0.30) do not correctly implement Nagle's Algorithm. " > >So there is some precedent for problems in this area, and the >other web references in such a search look useful. Actually, MPICH already disables Nagle's Algorithm, which was designed to reduce the number of packets going over the wire by agglomeration. This was very useful in long-haul links and hub-based TCP/IP from about 5-10 years ago, but it is somewhat less useful now. Also, 2.0.30 is so old as to be completely irrelevant. Even 2.2 had Nagle bugs, see: http://www.icase.edu/coral/LinuxTCP.html (their site or name server seems to be down). 2.4 should have a much better stack. But, realistically, programs like AMBER and CHARMM aren't going to scale very well over TCP/IP due to the high latency and low bandwidth of 100BaseT (and even gigabit) ethernet. The best I've been able to get is about 2X for 4 cpus and 3X for 6 cpus. However, since the cpus are cheap and so is the networking hardware, I don't particularly care that the scaling is poor. Dave From chemistry-request@server.ccl.net Sat Dec 15 16:46:00 2001 Received: from socrates.cgl.ucsf.edu ([128.218.27.3]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFLk0i31488 for ; Sat, 15 Dec 2001 16:46:00 -0500 Received: from socrates.cgl.ucsf.edu (localhost [127.0.0.1]) by socrates.cgl.ucsf.edu (8.9.3/8.9.3/GSC1.7) with ESMTP id NAA311159; Sat, 15 Dec 2001 13:45:36 -0800 (PST) Message-Id: <200112152145.NAA311159@socrates.cgl.ucsf.edu> To: Rick Venable cc: CHEMISTRY@ccl.net Subject: Re: CCL:Linux kernel 2.4 parallel scaling In-Reply-To: Your message of "Sat, 15 Dec 2001 13:58:39 EST." Date: Sat, 15 Dec 2001 13:45:36 -0800 From: David Konerding Rick Venable writes: >Myself and others here at NIH have been benchmarking CHARMM and other >parallel chemistry codes on Linux clusters using MPICH libs with an >ethernet network. Tests on 2.4.12 kernel nodes with 2, 4, and 8 procs >showed terrible parallel scaling, compared to the same tests run with >the 2.2.16-tcpfix kernel. > >I'd expect that other people using Linux clusters for computational >chemistry may have encountered this already, and may have looked into >possible solutions. > >For instance, are there some kernel parameters in /proc that can be used >to tune the communications performance? > >Are there any other software solutions to this problem? Can you please let us know what sort of network card you are using, whether zero-copy networking is enabled, and how many errors you see on your link (run ifconfig). I missed the fact earlier that you got better scaling with 2.2.16-tcpfix. This suggests to me that you need to look into how you configured the kernel and what other variables changed (or didn't change). How do simple programs like FTP compare between the two kernels? That's the first thing I'd check. Dave From chemistry-request@server.ccl.net Sat Dec 15 17:29:52 2001 Received: from bute.st-andrews.ac.uk (im17@[138.251.12.1]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBFMTqi31935 for ; Sat, 15 Dec 2001 17:29:52 -0500 Received: from localhost (im17@localhost) by bute.st-andrews.ac.uk (8.9.1a/8.9.1) with SMTP id WAA02981 for ; Sat, 15 Dec 2001 22:29:32 GMT Date: Sat, 15 Dec 2001 22:29:32 +0000 (GMT) From: Ibrahim Moustafa X-Sender: im17@st-andrews.ac.uk To: chemistry@ccl.net Subject: Molecular dynamics with implicit solvent Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear ccl users, I'm doing MD on my protein structure without explicit solvents because of the limiting computing power. I use Tripos FF, in SYBYL6.6; cutoff distance 15 A, distance dependent dielectric function with dielectric constant 2. First, I heat the system to 300 K in steps ( 10, 50, 100, 200, 300 K), 2 ps for each step. Comparing the resultant configuration of the heating step with the starting configuration gives ( rmsd 3.79 A ). Of course the absence of solvents is responsible for this nonsense movements. Now I'm thinking to do the simulation at low temperature ( 50 K ) to supress these large movements. I want to discuss this with ccl users before doing so. What I want to ask about is: suppose i conduct this simulation at 50 K instead of 300 k, non physiological condition, ( I have seen some people doing that " Journal of molecular structure (theochem) 368;1996: 205-212") Can we rely on the output results of that kind of simulation? can we realy trust the answer for the question under investigation? -Another question: in absence of solvents do we need to equilebrate the system before the production phase in the MD? if so, WHY? - Is it better to start the simulation from the average structure of the heating step ( or the equilebration step ) or start from the last configuration of each step?? I'd be really appreciated if someone answered these questions. I promise I'll sumarize the answer to the CCL users. Many thanks, Ibrahim Name :Ibrahim M.Moustafa Mail address :Center for Biomolecular Science, BIOMOLECULAR SCIENCE BUILDING, North Haugh,St-Andrews, Fife,KY16 9ST Scotland,U.K. Tel :+44(0)1334-467257 E-mail :im17@st-andrews.ac.uk From chemistry-request@server.ccl.net Sat Dec 15 20:19:45 2001 Received: from epyon.INS.cwru.edu (root@[129.22.9.21]) by server.ccl.net (8.11.6/8.11.0) with ESMTP id fBG1Jji05162 for ; Sat, 15 Dec 2001 20:19:45 -0500 Received: from gavin (chem51286.CHEM.CWRU.Edu [129.22.129.217]) by epyon.INS.cwru.edu with SMTP (8.11.6+cwru/CWRU-3.8) id fBG1JQv16485; Sat, 15 Dec 2001 20:19:26 -0500 (EST) (from hxt10@po.cwru.edu for ) Message-ID: <000d01c185d0$71249840$d9811681@cwru.edu> Reply-To: "Hui-Hsu \(Gavin\) Tsai" From: "Hui-Hsu \(Gavin\) Tsai" To: Subject: advise for protein folding and molecular electronics Date: Sat, 15 Dec 2001 20:24:43 -0500 MIME-Version: 1.0 Content-Type: multipart/alternative; boundary="----=_NextPart_000_000A_01C185A6.8823D6C0" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 6.00.2479.0006 X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2479.0006 This is a multi-part message in MIME format. ------=_NextPart_000_000A_01C185A6.8823D6C0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable Hi Everyone, I will start new projects in studying the protein folding and = molecular electronics (or nanotech) by utilizing computational methods. I have some questions which need your important comments. My questions are as below: 1. Which research groups did pioneer work in understanding the protein = folding and molecular electronics by using computational methods? 2. Is there any significant progress recently in understanding and = predicting protein folding? 3. What's the most difficulty in these fields (protein folding and = molecular electronics)? 4. What are the current research directions of these fields (protein = folding and molecular electronics)? Any comment is welcomed. I will summarize the results. Thanks Gavin =20 ------=_NextPart_000_000A_01C185A6.8823D6C0 Content-Type: text/html; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable
Hi Everyone,
 
    I will start new = projects in=20 studying the protein folding and molecular electronics (or = nanotech) by=20 utilizing computational methods.
I have some questions which need your = important=20 comments.
My questions are as below:
 
1. Which research groups = did pioneer work in=20 understanding the protein folding and molecular electronics by using=20 computational methods?
 
2. Is there any significant = progress recently=20 in understanding and predicting protein folding?
 
3. What's the most difficulty in these = fields=20 (protein folding and molecular electronics)?
 
4. What are the current research = directions of=20 these fields (protein folding and molecular electronics)?
 
 
Any comment is welcomed.
 
I will summarize the = results.
 
Thanks
 
Gavin  
------=_NextPart_000_000A_01C185A6.8823D6C0--