VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed Nov 30 2011 - 09:22:56 CST

Hi,
  Just FYI, NVIDIA has released a new 290.10 driver package and the new
driver seems to be free of the problems I described below previously.
I'd be curious to know if the VRPN issues are also resolved with the
290.10 driver.

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu
  

On Fri, Nov 18, 2011 at 01:49:18AM -0600, John Stone wrote:
> Hi,
> I wanted to add some recent observations regarding the possibility of
> problems with certain NVIDIA driver versions. I just discovered that
> some of the most recent driver versions (285.xxx) seem to trigger various
> VMD stability problems that are related to multithreaded code and the
> NVIDIA tls library. I discovered this when immediately
> after updating from my older 280.13 driver to 285.xxx I began having
> crashes that were not explained by several hours of tests within valgrind
> or other debugging tools. Reverting back to the older 280.13 drivers
> cured the problem in my case. If I can reproduce these problems more
> broadly on several machines I will likely file a bug report with NVIDIA.
> If others have observed any problem that sounds like this, let me know.
>
> Cheers,
> John Stone
> vmd_at_ks.uiuc.edu
>
> On Thu, Nov 17, 2011 at 11:04:38AM +1000, Tristan Croll wrote:
> > Hi Axel,
> >
> > An update on this, which first needs a bit more detail:
> >
> > The machine I'm running is an HP Z800, with two Xeon Nehalem CPUs. I recently purchased a Tesla C2050 and Quadro 4000 in order to upgrade it to the point where I can run small interactive simulations with 3D visualisation. Unfortunately we were short one power cable for the Tesla, so we just stuck the Quadro in as a CUDA compute board while we were waiting for that to arrive.
> >
> > That's when we ran into the troubles with the Falcon - which was, as I said, resolved by rolling back the NVidia driver.
> >
> > After the missing power cable arrived we added the Tesla, and shifted the Quadro over to running the displays. That's all we did - no changes were made to any of the drivers. And suddenly the Falcon wasn't working again. Different etiology this time: no error messages in connecting VMD to the vrpn server, but it simply wouldn't register any movement of the Falcon (the Falcon test program ran fine).
> >
> > Anyway, I downloaded the alpha version of VMD you suggested, and that's working. Now we're planning to compile version 1.9 with the latest vrpn and CUDA packages, and see how that goes. So it seems at least part of the problem with the older version is a conflict with NVidia devices, but I'm at a loss to say what that conflict is.
> >
> > Incidentally, I noticed another problem along the way: when using the Falcon in VMD, occasionally the system would dramatically slow down (to a few screen refreshes per second). This would persist for a couple of minutes after I closed both vmd and the vrpn server. It seems (fingers crossed) that this problem was resolved by turning off the X11 Composite extension in xorg.conf.
> >
> > Cheers,
> >
> > Tristan
> >
> > -----Original Message-----
> > From: Axel Kohlmeyer [mailto:akohlmey_at_gmail.com]
> > Sent: Wednesday, 2 November 2011 10:08 AM
> > To: Tristan Croll
> > Cc: vmd-l_at_ks.uiuc.edu
> > Subject: Re: vmd-l: VRPN-ICMS conflict with NVidia driver
> >
> > On Tue, Nov 1, 2011 at 8:03 PM, Tristan Croll <tristan.croll_at_qut.edu.au> wrote:
> > > Hi Axel,
> > >
> > > Yes, it was completely unexpected to us as well.  It only came to light when he ran VMD remotely off my machine using his laptop, and things magically started working.  To answer your specific question below, though, I tried that - downloaded vmd-1.9.1a12 off Biocore yesterday.  It didn't help.
> >
> > no, that is not the version that i was talking about, there should be
> > a special (older) version that has the newer VRPN client, but this
> > is in a subdirectory. check this out:
> >
> > ls /home/akohlmey/biocore/VMD\ \(Public\)/alpha/VRPN-ICMS
> >
> > README.VRPN-ICMS
> > vmd-1.8.8a21.VRPN-ICMS.LINUX.tar.gz
> > vmd-1.8.8a28-ICMS.bin.LINUX.tar.gz
> > vmd-1.8.8a21.VRPN-ICMS.LINUXAMD64.tar.gz
> > vmd-1.8.8a28-ICMS.bin.LINUXAMD64.tar.gz
> >
> > cheers,
> > axel.
> >
> > >
> > > Cheers,
> > >
> > > Tristan
> > >
> > > -----Original Message-----
> > > From: Axel Kohlmeyer [mailto:akohlmey_at_gmail.com]
> > > Sent: Tuesday, 1 November 2011 11:16 PM
> > > To: Tristan Croll
> > > Cc: vmd-l_at_ks.uiuc.edu
> > > Subject: Re: vmd-l: VRPN-ICMS conflict with NVidia driver
> > >
> > > hi tristan,
> > >
> > > On Tue, Nov 1, 2011 at 3:16 AM, Tristan Croll <tristan.croll_at_qut.edu.au> wrote:
> > >> Hi,
> > >>
> > >> My local IT guru and I have just spent the better part of a day banging our
> > >> heads against a problem that has the same appearance as a previous one
> > >> mentioned in the VMD list (starting at
> > >> http://www.ks.uiuc.edu/Research/vmd/mailing_list/vmd-l/16277.html).  We'd
> > >
> > > hmm... this problem has been tracked down to an incompatibility
> > > in the behavior of the (rather old) VRPN client code in VMD and
> > > the newer server code in VRPN-ICMS. so using a VMD binary
> > > linked against a newer VRPN library resolved that for real.
> > >
> > >> connected a Novint Falcon, used the demo program to confirm it was working
> > >> properly, and used the vrpn test programs to confirm that vrpn was
> > >> communicating OK.  Yet when we tried to connect through VMD, however, it
> > >> refused to connect, with error messages identical to those described in the
> > >> message I linked above.
> > >>
> > >> Anyway, the problem was eventually (and rather bizarrely) solved by rolling
> > >> our NVidia drivers back from 285.05.09-285.05.09-139.el6.x86_64 to
> > >> 280.13-280.13-138.el6.x86_64.
> > >
> > > that is unusual and unexpected. i have no idea how this would
> > > correlate to the connection issues, but thanks for letting us know.
> > >
> > >> Thought it was worth a mention, for weirdness if nothing else.
> > >
> > > since you spent already some time on this, would you mind
> > > doing one more test? and use the contributed alpha test
> > > binaries of VMD that include the new VRPN client code?
> > > those should allow you to work with the newer nvidia drivers.
> > >
> > > thanks,
> > >    axel.
> > >
> > >>
> > >>
> > >> Cheers,
> > >>
> > >>
> > >>
> > >> Tristan
> > >
> > >
> > >
> > > --
> > > Dr. Axel Kohlmeyer
> > > akohlmey_at_gmail.com  http://goo.gl/1wk0
> > >
> > > College of Science and Technology
> > > Temple University, Philadelphia PA, USA.
> > >
> >
> >
> >
> > --
> > Dr. Axel Kohlmeyer
> > akohlmey_at_gmail.com  http://goo.gl/1wk0
> >
> > College of Science and Technology
> > Temple University, Philadelphia PA, USA.
>
> --
> NIH Resource for Macromolecular Modeling and Bioinformatics
> Beckman Institute for Advanced Science and Technology
> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
> http://www.ks.uiuc.edu/~johns/ Phone: 217-244-3349
> http://www.ks.uiuc.edu/Research/vmd/ Fax: 217-244-6078 

-- 
NIH Resource for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
http://www.ks.uiuc.edu/~johns/           Phone: 217-244-3349
http://www.ks.uiuc.edu/Research/vmd/       Fax: 217-244-6078