Re: namd2 CUDA errors for non-superusers in Debian

From: Michael Mazack (michael_at_mazack.org)
Date: Fri Feb 15 2013 - 08:47:47 CST

Does adding the user who needs to use CUDA to the "video" group in
/etc/group have any affect?

Michael

On Fri, Feb 15, 2013 at 8:42 AM, Blake Mertz <blake.mertz_at_mail.wvu.edu>wrote:

> Aron,
>
> Sorry for the confusion -- I knew I should have mentioned that I had
> the correct LD_LIBRARY_PATH in my previous email (they are identical
> for myself and the other user). That was actually my first
> inclination, was that it was an improperly referenced variable. The
> gist of the situation is this:
>
> 1) When the second user logs on and tries to run namd, she gets the CUDA
> error.
>
> 2) When the second user logs on and runs namd as a superuser, namd
> will properly run.
>
> So there is some disconnect with Debian recognizing this second user
> as having the correct privileges to run namd. I've got the namd binary
> soft-linked to the source directory.....I'm sure this is a linux
> issue, and not namd. I was hoping that someone had run into this issue
> before, but it looks like I'll need to troubleshoot it myself. Thanks
> for putting in your two cents, Aron. I really appreciate it, and
> didn't get the chance to thank you last fall -- you did help me on
> another CUDA-related issue, just forgot to reply back then.
>
> Blake
>
> On Thu, Feb 14, 2013 at 6:31 PM, Aron Broom <broomsday_at_gmail.com> wrote:
> > I'm a bit confused.
> >
> > 1) I don't see any errors
> >
> > 2) You are saying the SU gets errors? but your subject line says the
> non-SU
> > gets errors?
> >
> > Either way, I can imagine it might simply be a problem with the library
> > path. You should be running NAMD using the cuda library that is shipped
> > with it, rather than the one that is part of the cuda libs. So if one of
> > you has set your LD_LIBRARY_PATH to be the NAMD install directory, and
> the
> > other hasn't, this kind of problem could occur. That's my best guess at
> > this point.
> >
> > ~Aron
> >
> > On Thu, Feb 14, 2013 at 5:28 PM, Blake Mertz <blake.mertz_at_mail.wvu.edu>
> > wrote:
> >>
> >> Hello,
> >>
> >> I realize this may be an Nvidia-related issue, but I wanted to see if
> >> any namd users had run up against this problem as well. I'm running
> >> Debian wheezy with native Nvidia drivers and cuda libraries, and can
> >> run CUDA-accelerated pre-compiled binaries with the apoa1 benchmark
> >> system myself. I have another user who has superuser privileges. She
> >> gets the following error when running namd after ssh'ing in:
> >>
> >> namd2 +idlepoll +p4 +devices 0 apoa1.namd
> >
> >
> >
> >
> > --
> > Aron Broom M.Sc
> > PhD Student
> > Department of Chemistry
> > University of Waterloo
>
>
>
> --
> Assistant Professor
> C. Eugene Bennett Department of Chemistry
> (304) 293-9166
> http://community.wvu.edu/~jbm085/
>
> "Life is not easy for any of us. But what of that? We must have
> perseverance and above all confidence in ourselves. We must believe
> that we are gifted for something and that this thing must be
> attained." Marie Curie
>
> "Start by doing what's necessary; then do what's possible; and
> suddenly you are doing the impossible." St. Francis of Assissi
>
>

-- 
Michael J. M. Mazack
Scientific Computation Program
University of Minnesota
Minneapolis, MN 55455
E-mail: michael_at_mazack.org
Web: http://mazack.org

This archive was generated by hypermail 2.1.6 : Tue Dec 31 2013 - 23:22:57 CST