VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Thu Jun 06 2002 - 01:41:32 CDT

Will,

On Thu, Jun 06, 2002 at 02:27:25AM -0400, Will Sheffler wrote:
> John,
>
> Thanks you for another prompt and thourough response. We've already
> got a VRPN setup for our CAVE, so I think I'm in good shape. Is there
> a 32 processor cap? We've got a 48 proc cluster; the more the merrier.

There isn't really a "cap", but as with any parallel program, NAMD
exhibits a gradual decrease in performance per added processor, which
depends greatly on the size of the molecule being simulated. If you
simulate a 66 atom molecule for example, there isn't enough work to
keep 66 processors busy, so it would actually run _slower_ to run such
a tiny simulation on a large cluster. I mention 32 processors because
that's more than enough for many IMD simulations we've run locally.
It depends a lot on the type of interconnect in your cluster. Your
mileage may vary, I'd suggest running NAMD with a varying number of
processors and see what the benchmark time it reports is, and choose
the number of processors that gives you the best performance on the
molecule you plan to simulate. On a very small molecule 8 processors
might actually do better than 32.

> We're having some trouble getting the CAVE dispay to work. Do you
> happen to know what version of cavelib is required by VMD 1.8?

Are you building your own, or using our binaries? Our binaries were built
using the NCSA cavelib, version 2.6g. I know Todd Furlong has been building
his own VMD binaries for his company's CAVE, and I believe they are using a
much newer version than we are, and it works ok for him. How many walls does
your CAVE have? The VMD 1.8 code currently assumes that you are running
_exactly_ one display process per display wall. This is important as our
code performs a barrier synchronization inside of the VMD rendering code
which forces all of the slave drawing processes to wait until the master
process allows them to draw, and then makes the master process wait to
update display data until they have finished rendering a frame.

> Do you recommend using the beta 2.5 namd?

That's what we're running locally, it should work just fine
as far as I know.

> The VMD/NAMD combination is amazing. I can't believe it's not all over
> our bio dept yet.

Thanks for the compliment. Feel free to show it to as many people as
you like. We're always happy to get more people using our programs! :)

Let us know if you need more help getting things going.

Thanks,
  John Stone
  vmd_at_ks.uiuc.edu

> On Thu, Jun 06, 2002 at 12:24:19AM -0500, John Stone wrote:
> |
> | Will,
> | Generally speaking its mostly an issue of getting a VRPN server
> | running for your Phantom. Getting VMD and NAMD up and connecting to
> | each other is actually quite easy once you've got a VRPN server on
> | the haptic device. In order to do IMD really well, you'll need a
> | minimum of 3 machines: one to run a VRPN server, one to run VMD,
> | and one (or several) to run NAMD. In our laboratory, we use PC's
> | running NT to drive the Phantom, and run the VRPN server. Most any
> | machine will do, we actually use a cheap laptop to run one of them,
> | so that particular Phantom can be carried to different rooms
> | (or to off-site demos) relatively easily. For VMD, you'll just need
> | a decent viz box, be it a workstation like a Sun/SGI, or a Linux box
> | with a GeForce board. For NAMD, you'll need the fastest box or boxes
> | you can get your hands on, we normally recommend using a PC cluster if
> | you've got one available locally. For best results something on the order
> | of 8 1GHz Athlons works quite well for many IMD simulations, though you
> | can use as many as 32 with good results, the performance has a lot to do
> | with the size of the simulated molecular system, and performance
> | characteristics of the cluster's network. You can of course run NAMD
> | on supercomputers or large shared memory machines if you happen to have
> | such machines available, they work well also.
> |
> | The best way to start, if you haven't ever done this before, is to get
> | VRPN from UNC and get a VRPN server up and running. Once you've got that
> | going, then download and install VMD on one of your machines. From there,
> | you should be able to setup your .vmdsensors file specifying your local
> | VRPN server, and attach to it with VMD, even before trying a real simulation.
> | After that, you'll want to setup NAMD, and try connecting to a simulation.
> | One of the VMD developers (Paul Grayson) has written an easy-to-use script
> | for setting up and running IMD simulations, and this will make running the
> | simulations very easy once you've got each of the individual components
> | up and running.
> |
> | Please let us know if you have any other questions. It may sound a bit
> | daunting at first, but its actually not too hard. Getting the Phantom
> | to cooperate with VRPN is often the biggest problem (since there are various
> | incompatible versions of the Phantom drivers and the Ghost toolkit...),
> | the rest of the process is typically quite smooth.
> |
> | Thanks,
> | John Stone
> | vmd_at_ks.uiuc.edu
> |
> | On Thu, Jun 06, 2002 at 01:11:34AM -0400, Will Sheffler wrote:
> | > VMD people,
> | >
> | > How difficult is it to get a PHANToM working with a VMD/NAMD
> | > interactive sim? Does anybody have practical experience doing this?
> | > We've got an PHANToM around somewhere and I'm thinking about playing
> | > with it...
> | >
> | > -Will Sheffler
> | > Graphics Group
> | > Brown University
> |
> | --
> | 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: johns_at_ks.uiuc.edu Phone: 217-244-3349
> | WWW: http://www.ks.uiuc.edu/~johns/ Fax: 217-244-6078
>
>
> -Will Sheffler
> -------------------
> $console> cd /pub
> $console> more beer
> [_]3
> beer (END) 

-- 
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: johns_at_ks.uiuc.edu                 Phone: 217-244-3349              
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078