VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Thu Jun 11 2009 - 21:30:22 CDT

Marius, Lucie,
  NVIDIA engineers were able to reproduce the hanging problem today
and they provided us with a suggestion for a workaround which worked
when Axel and I tried it on one of this machines here at U. Penn
(I'm visiting Axel this week).

Try adding this extra config line to the device section
of your /etc/X11/xorg.conf config file:
  Option "ConstantFrameRateHint" "True"

This apparently disables some performance optimizations, but
results in stable behavior in the case of the problem you're
experiencing with the most recent linux distros and the new
NVIDIA drivers. Let us know if this solves the problem for you
on your machines. This is what the NVIDIA engineers suggested
we try, and it worked for Axel and I on his test box.

NVIDIA has a bug filed on this and whenever a production driver is
available that fixes the problem, I'll post a note to VMD-L.
For now, I suggest adding the config line above and use it as
a short-term remedy, so that you can still use CUDA and other
advanced features of the latest NVIDIA drivers.

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Thu, Jun 11, 2009 at 08:59:54AM -0500, John Stone wrote:
>
> Marius, Lucie,
> Axel and I did some testing (I'm at U. Penn. visiting his lab this week)
> and it seems that the Fedora "legacy" NVIDIA 173.xx driver package
> works fine also. I am in contact with the driver team at NVIDIA and
> they are working on reproducing the problem in their lab currently.
> If they can reproduce the problem in their lab, then they should be
> able to tell us more about what's going on here. In the short term,
> using the 173.xx driver series looks like it may be a viable workaround
> until we have a better solution.
>
> Cheers,
> John Stone
> vmd_at_ks.uiuc.edu
>
> On Thu, Jun 11, 2009 at 09:43:05AM +0200, Marius Retegan wrote:
> > Yesterday I also did a simple test. While rebooting with an older kernel for
> > which
> > I have only the nouveau drivers I didn't noticed any problems while running
> > VMD.
> > So as you said it's possible that this whole thing is not related to VMD.
> > Anyway thank you.
> > Marius
> >
> > On Wed, Jun 10, 2009 at 6:47 PM, John Stone <johns_at_ks.uiuc.edu> wrote:
> >
> > >
> > > Marius,
> > > Axel and I have been looking into this using a Fedora 11 machine
> > > he just installed yesterday, and there's definitely something fishy
> > > going on where the VMD process gets a signal, a signal handler runs
> > > (apparrently in the OpenGL driver), and the OpenGL driver library
> > > function calls malloc, and the whole process gets stuck in a glibc
> > > malloc mutex call. My theory at present is that either
> > > glibc has a malloc double locking bug, or the NVIDIA driver
> > > should not be calling malloc in a signal handler. We won't know
> > > more until we do some VMD compiles on Fedora 11 itself and see if
> > > the same problem occurs when compiled against that libc/libstdc++ etc.
> > >
> > > So far I see nothing to indicate that this is a problem in VMD itself.
> > > We'll let you know what we find out after we do more testing.
> > >
> > > Cheers,
> > > John Stone
> > > vmd_at_ks.uiuc.edu
> > >
> > > On Tue, Jun 09, 2009 at 06:30:43PM +0200, Marius Retegan wrote:
> > > > They release it today :)
> > > > I've updated my system to the last stable repos and I've disabled rawhide
> > > > repositories.
> > > > Sure I can run the tests.
> > > > Marius
> > > >
> > > > On Tue, Jun 9, 2009 at 6:26 PM, John Stone <johns_at_ks.uiuc.edu> wrote:
> > > >
> > > > >
> > > > > Marius,
> > > > > Which Fedora 11 beta release is this? They are expected to announce
> > > > > the final version imminently, but last I checked they haven't released
> > > the
> > > > > final version yet:
> > > > > http://fedoraproject.org/wiki/Releases/11/Schedule
> > > > >
> > > > > If this is something specific to bleeding edge Fedora or other
> > > > > distros, it may be a libc, libstc++, or Xlib binary compatibility
> > > > > issue of some kind. If so, we could do builds specifically for
> > > > > these platforms, but I'd prefer to find a solution that would allow
> > > > > you to run the same binary that works everywhere else. If you're
> > > willing
> > > > > to do some more testing for me, I might have a few ideas to try that
> > > > > involve copying in various shared libraries to see if they have an
> > > effect.
> > > > >
> > > > > Cheers,
> > > > > John Stone
> > > > > vmd_at_ks.uiuc.edu
> > > > >
> > > > > On Tue, Jun 09, 2009 at 06:07:25PM +0200, Marius Retegan wrote:
> > > > > > Hi
> > > > > >
> > > > > > The OpenGL window always starts, but then VMD hangs. The main window
> > > (all
> > > > > > the menus windows :) ) starts only if I go in debug mode.
> > > > > > I've tried with and without Composit extension. Same result.
> > > > > > Compiz is not enabled.
> > > > > >
> > > > > > On Tue, Jun 9, 2009 at 4:48 PM, John Stone <johns_at_ks.uiuc.edu>
> > > wrote:
> > > > > >
> > > > > > >
> > > > > > > Marius,
> > > > > > > Interesting. Does the program hang, crash, or appears to start up
> > > > > > > but you never see the OpenGL window? Do you have Compiz or the
> > > > > Composite
> > > > > > > extension enabled on this machine? (use xdpyinfo to query for
> > > this)
> > > > > > >
> > > > > > > Cheers,
> > > > > > > John
> > > > > > >
> > > > > > > On Tue, Jun 09, 2009 at 09:08:37AM +0200, Marius Retegan wrote:
> > > > > > > > I've tried the 185.18.14 driver. Strangely the main window
> > > doesn't
> > > > > even
> > > > > > > come
> > > > > > > > up. VMD in debug mode works again.
> > > > > > > > I can post any details about my system if necessary.
> > > > > > > > Marius
> > > > >
> > > > > --
> > > > > 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/><
> > > http://www.ks.uiuc.edu/%7Ejohns/>
> > > > > 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
> > > Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> > > WWW: http://www.ks.uiuc.edu/~johns/ <http://www.ks.uiuc.edu/%7Ejohns/>
> > > 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
> Email: johns_at_ks.uiuc.edu Phone: 217-244-3349
> WWW: http://www.ks.uiuc.edu/~johns/ 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
Email: johns_at_ks.uiuc.edu                 Phone: 217-244-3349
  WWW: http://www.ks.uiuc.edu/~johns/      Fax: 217-244-6078