From: John Stone (
Date: Mon Mar 13 2006 - 10:31:24 CST

  I don't know what the ++server-port flag is, but I don't think
it has anything to do with NAMD, I believe that must be a Charm++ flag,
and it certainly won't help you run IMD. The most likely source of trouble
here is that your SP4 is blocking incoming connections to the compute nodes.
This is a common problem with doing IMD on supercomputers being run at the
various centers. If that's not the cause, then the next most likely cause
is that you're attempting to connect to the wrong node of the big machine.
For IMD, you need to connect to the root node of the NAMD job, which is
where the IMD communication occurs. If you try to connect to any of the
other nodes, it'll fail. You might try running a very simpel 1 processor
NAMD job (so there's only one node involved) and see if you can successfully
connect to that before continuing on with larger jobs. If you can't get that
working, then there's definitely something amiss. The IMD code should be safe
from 64-bit/32-bit issues as we've been using it in mixed word size client
server combinations for quite a few years. In any case, even if there is
a word size issue, you should still get an initial connection (without
that unusual ++server-port flag AFAIK). Try the suggestions I've given
above and get back to me and we'll figure out what's going on.

  John Stone

On Mon, Mar 13, 2006 at 02:29:36PM +0100, Marc Baaden wrote:
> Hi,
> we have some trouble making IMD work between a visualisation/interaction
> client running VMD under Linux and a NAMD calculation on an IBM SP4. I
> wonder whether anybody has successfully used such a combination? The
> underlying question is whether there is a specific problem with our setup
> or a general problem with these architectures.
> For more details see below (a yet unanswered detailed description sent to
> the NAMD mailing list by a colleague).
> Thank you very much in advance,
> Marc Baaden
> Subject: Problem to run IMD simulation
> From: Laurent Leger <>
> Date: Fri, 24 Feb 2006 11:29:29 +0100
> Hi All,
> I'm trying to run an IMD simulation with VMD (1.8.4b6 for Linux workstation)
> and with NAMD (2.6b1 for IBM-SP) but without success.
> I downloaded the binaries of each software on the respective web sites.
> My problem is that I'm not able to connect two softwares on a TCP/IP port.
> My chosen IP port (54321) isn't a restricted port and I have no restriction
> to access to the IBM SP4 from the Linux workstation.
> So I don't see where the problem is.
> I have made somes investigations :
> a/ I tried to connect VMD (on the same Linux wokstation) with NAMD on a SGI
> computer (Onyx 3800). I had no problem : IMD simulation runs perfectly.
> So I thought that 1.8.4b6 version of VMD for Linux is OK and I looked on
> the side of NAMD on my IBM SP4.
> b/ In first, I saw that NAMD didn't listen on TCP/IP port even if I have
> theses lines in my config file :
> IMDon yes
> IMDport 54321
> IMDfreq 1
> IMDwait on
> I have found a workaround by setting "++server-port" flag of NAMD.
> That seems strange to me to have to use this flag. I have the same
> problem if I run NAMD under LoadLeveler control or directly from the
> prompt. Does anyone have the same problem with an IBM-SP host ?
> c/ So now, I run NAMD on my IBM SP4 with the command :
> namd ++server-port 54321 config-file
> Like this, I have a unix process listenning on 54321 IP port. Then, when
> I run VMD (on Linux workstation), I can see a connection established
> between two hosts on this IP port. But VMD returns this error message :
> ERROR) Unable to ascertain relative endianness of remote machine
> Unable to connect to host zahir on port 54321
> The connection is immediatly closed (because VMD is unable to ascertain
> endianness). But, on IBM-SP, the IP port has the CLOSE_WAIT status (thus
> something has been exchanged between two hosts on IP port).
> In more, NAMD doesn't detect the connection and stay in wait mode. It's
> seem to me that NAMD is unable to use the IP port.
> Does anyone can explain me this problem ?
> For information, I specifie that my IBM SP4 uses 64 bits adressing mode and
> downloaded binaries are designed to run in 32 bits adressing mode.
> Is the reason of my problem ?
> I tried to built my own version of NAMD by setting "-q64" flag of xlC compiler.
> But builted executable doesn't run. I have a "Memory fault" message.
> Is there someone who compile NAMD with 64 bit adressing mode ?
> Have you an idea to help me ?
> Thanks for your help.
> Laurent Leger
> --
> Dr. Marc Baaden - Institut de Biologie Physico-Chimique, Paris
> -
> FAX: +33 15841 5026 - Tel: +33 15841 5176 ou +33 609 843217

NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
Email:                 Phone: 217-244-3349
  WWW:      Fax: 217-244-6078