VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed May 18 2005 - 11:15:59 CDT

Sabuj,
  I just posted a followup to your ATI bug report suggesting that
people having that particular ATI driver problem use a different
method for disabling the use of GLSL rather than the VMDSIMPLEGRAPHICS
method, which also turns off a bunch of other features:
  setenv VMD_EXCL_GL_EXTENSIONS GL_ARB_shader_objects

The traceback that I just got this morning has the appearance of ATI's
shader compiler getting caught in some very deep recursion, perhaps
resulting in a stack overflow. The traceback I saw was well over 1,500
function calls deep, so that's almost certainly some bogosity in their
driver.

Hopefully we'll hear back from the ATI guys on this soon.

  John Stone
  vmd_at_ks.uiuc.edu

On Tue, May 17, 2005 at 02:15:47PM -0500, Sabuj Pattanayek wrote:
> Hi,
>
> I posted this to ati's unofficial bugzilla some time ago in the hopes
> that someone from ati would find it (I've heard they check in from time
> to time). Someone did respond with a gdb output (don't know if he's from
> ati). The bug is at: http://ati.cchtml.com/show_bug.cgi?id=123 .
>
> In case the website goes down again:
>
> ------- Additional Comment #2 From Aric Cyr 2005-05-11 03:56 EST
> [reply] -------
>
> There is a workaround: set then environment variable VMDSIMPLEGRAPHICS
> to 1 before running vmd. This will disable using shaders in VMD and
> allows the program to otherwise run fine.
>
> Here is a backtrace from the crash... clearly the segfault is in the driver:
>
> Program received signal SIGSEGV, Segmentation fault.
> 0xb7c13ebe in s6241 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> (gdb) bt
> #0 0xb7c13ebe in s6241 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> #1 0xb7cee285 in s4693 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> #2 0xb7c15762 in s3715 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> #3 0xb7c0f771 in s9456 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> #4 0xb7876444 in s4062 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> #5 0xb79167c0 in s9013 () from /usr/X11R6/lib/modules/dri/fglrx_dri.so
> #6 0x0809296c in OpenGLShader::CompileShaders (this=0x81df270,
> vertexShader=0x83c96f8 "/", '*' <repeats 75 times>, "\n *cr\n *cr", ' '
> <repeats 12 times>, "(C) Copyright 1995-2005 The Board of Trustees of the\n
> *cr", ' ' <repeats 24 times>, "University of Illinoi"...,
> fragmentShader=0x83ca5b0 "/", '*' <repeats 75 times>, "\n *cr\n
> *cr", ' '
> <repeats 12 times>, "(C) Copyright 1995-2005 The Board of Trustees of the\n
> *cr", ' ' <repeats 24 times>, "University of Illinoi"...) at
> OpenGLShader.C:142
> #7 0x08092631 in OpenGLShader::LoadShader (this=0x81df270,
> shaderpath=0x83bdea0 "/tmp/vmd/lib/vmd/shaders/vmd") at
> OpenGLShader.C:56
> #8 0x0808b8af in OpenGLRenderer::setup_initial_opengl_state
> (this=0x81dea40)
> at OpenGLRenderer.C:368
> #9 0x0808964d in OpenGLDisplayDevice::open_window (this=0x81dea40,
> nm=0x81df218 "VMD 1.8.3 OpenGL Display", size=0xbffff220, loc=0x0,
> argc=1,
> argv=0xbffff314) at OpenGLDisplayDevice.C:558
> #10 0x08088aa7 in OpenGLDisplayDevice::init (this=0x81dea40, argc=1,
> argv=0xbffff314, size=0xbffff220, loc=0x0) at OpenGLDisplayDevice.C:357
> #11 0x08125e29 in VMDApp::VMDinit (this=0x81de108, argc=1, argv=0xbffff314,
> displaytype=0x819a400 "WIN", displayLoc=0x81d423c,
> displaySize=0x81d4234)
> at VMDApp.C:198
> #12 0x0813853f in main (argc=1, argv=0xbffff314) at vmd.C:247
>
> -------------------------------------------------------------
>
> The thing is I don't have these problems with vmd 1.8.2 and the fglrx
> drivers. John, I know you're busy, but when you get a chance would it be
> possible to post a 1.8.3 32bit linux version that doesn't do the
> OpenGLSL checks on startup or that doesn't do anything more than what
> vmd 1.8.2 does with openGL calls?
>
> Thanks,
> Sabuj Pattanayek
>
> John Stone wrote:
> > Hi,
> > Can you try doing the following steps for me, and let me know what
> > results you get?
> > start an xterm, then in that xterm do:
> > /bin/csh (use csh so the other commands I list will work)
> > source /usr/local/bin/vmd (use the startup script's env vars)
> > (then immediately quit that VMD)
> > gdb /usr/local/lib/vmd/vmd_LINUX
> > (then issue these commands to gdb)
> > run
> > (then load a structure that normally crashes your VMD)
> > (... then try to crash it)
> > where (this shows where VMD was when the crash occured)
> >
> > If you can email the output to me, I should be able to determine if
> > this is another ATI driver problem, or something else.
> >
> > John Stone
> > vmd_at_ks.uiuc.edu
> >
> > On Sun, May 15, 2005 at 05:16:08PM +0200, Gyimesi Gergely wrote:
> >
> >>Helo,
> >>
> >> Just for your info: I'm having the same problem, with the fglrx-8.10.19 and
> >>fglrx-8.12.10 drivers also, with a Radeon 8500 card. I've tried setting
> >>VMDSIMPLEGRAPHICS, but it didn't help.. The segfault only happens when
> >>loading large molecules (eg. proteins), it works OK for smaller ones. It also
> >>works OK in text mode.
> >>Looking forward to anything you can do, after May 20. :-)
> >>
> >>Gergely Gyimesi
> >>Ph.D student
> >>Institute of Enzymology,
> >>Hungarian Academy of Sciences
> >>
> >>
> >>>Hi,
> >>>Try setting this environment variable before starting VMD, it will
> >>>disable all of the advanced OpenGL features that VMD uses, in favor of
> >>>a more minimalistic and simpler OpenGL 1.1-based functionality:
> >>>setenv VMDSIMPLEGRAPHICS 1
> >>>
> >>>Let me know if this gets you up and running.
> >>>
> >>>John Stone
> >>>vmd_at_ks.uiuc.edu
> >>>
> >>>
> >>>
> >>>
> >>>>Is there any way to work around the fglrx-8.10.19 driver crash? I've
> >>>>seen the posts and bug report but would prefer not to downgrade my
> >>>>driver. Is there any way to use MESA or some other software rendering
> >>>>instead?
> >>>>
> >>>>
> >>>
> >>>
> >>>
> > 

-- 
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