VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed May 26 2010 - 10:39:53 CDT

Hi,
  I've implemented new code that will entirely prevent VMD from touching the
GPUS when the VMDNOCUDA environment variable is set. To give more detailed
control, I've also added a new VMDCUDADEVICEMASK environment variable that
can be set with a hexadecimal number indicating which GPUs are allowed to
be used and which ones are not (similar to CPU thread affinity masks).
The new version is posted in the alpha test area on BioCoRE if you'd like
to try it out.

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Wed, May 12, 2010 at 07:48:46AM +0200, Joakim Swedberg wrote:
> Hi, thanks for the help - I'm looking forward for that new feature in VMD
>
> Joakim Swedberg
> BSc (Honours)
> PhD student
>
> ----------------------------------------------------------------------
>
> Institute of Health and Biomedical Innovation (IHBI)
> 60 Musk Avenue
> Kelvin Grove
> Queensland 4059
> Australia
>
> http://www.ihbi.qut.edu.au/
>
> Phone +61 (0)7 3138 6222
> Mobile +61 (0)413 165 805
> Email j.swedberg_at_student.qut.edu.au
> joakim_swedberg_at_hotmail.com
>
> ----------------------------------------------------------------------
>
>
>
> > Date: Tue, 11 May 2010 22:07:03 -0500
> > From: johns_at_ks.uiuc.edu
> > To: akohlmey_at_gmail.com
> > CC: joakim_swedberg_at_hotmail.com; vmd-l_at_ks.uiuc.edu
> > Subject: Re: vmd-l: Selecting CUDA devices in VMD 1.8.7
> >
> >
> > Hi,
> > Yes, I know (compute-exclusive), but since he didn't mention having
> > set the devices to anything but normal mode, I doubt this is the reason
> > he's encountering the problem.
> >
> > I'm guessing that what is happening is that ACEMD is
> > getting a failed memory allocation if it is running at the moment
> > VMD goes through its startup memory allocation/clear/deallocation.
> >
> > If it's just an issue of compute-exclusive mode, there are a number
> > of ways of dealing with it.
> >
> > I haven't added code to VMD to allow a "gpu subset" to be specified thus
> > far, but it wouldn't be too hard to do so for people that want this
> > capability. I could probably implement such a feature in one of the
> > alpha VMD builds in a few days. I'm traveling off and on from
> > Wednesday until Monday, but I could probably have a new test build
> > with such a feature available shortly thereafter.
> >
> > Cheers,
> > John
> >
> > On Tue, May 11, 2010 at 09:08:25PM -0400, Axel Kohlmeyer wrote:
> > > On Tue, 2010-05-11 at 19:23 -0500, John Stone wrote:
> > >
> > > john,
> > >
> > > if the GPUs are configured for "compute-exclusive" mode,
> > > then you cannot access them with any other code for as
> > > long as VMD is running. i have seen this behavior on our
> > > 6-GPU node for example with the latest drivers, VMD cvs v
> > > ersion, CUDA-3.0 etc.
> > >
> > > cheers,
> > > axel.
> > >
> > > > Hi,
> > > > VMD doesn't "lock" any of the devices. When it starts up it does
> > > > clear the GPU memory and do a brief allocation/clear/deallocation
> cycle
> > > > designed to catch any unusual GPU behavior during startup. Does
> ACEMD fail
> > > > while VMD is starting up? What happens if you do a "kill -STOP" on
> ACEMD,
> > > > then start VMD, than do a "kill -CONT" on ACEMD after VMD prints the
> startup
> > > > messages?
> > > >
> > > > Cheers,
> > > > John Stone
> > > > vmd_at_ks.uiuc.edu
> > > >
> > > > On Wed, May 12, 2010 at 12:49:54AM +0200, Joakim Swedberg wrote:
> > > > > Hi everybody,
> > > > >
> > > > > I have a problem where I need to run another CUDA accelerated
> software's
> > > > > at the same time as VMD 1.8.7. The other software ACEMD lets me
> specify
> > > > > which devices I want to use but I have not been able to find a way
> to do
> > > > > that in VMD and it locks all devices, As a result the other
> software
> > > > > always crashes if I run VMD 1.8.7.
> > > > >
> > > > > Any help would be useful
> > > > >
> > > > > Regards
> > > > >
> > > > > Joakim Swedberg
> > > > > BSc (Honours)
> > > > > PhD student
> > > > >
> > > > >
> ----------------------------------------------------------------------
> > > > >
> > > > > Institute of Health and Biomedical Innovation (IHBI)
> > > > > 60 Musk Avenue
> > > > > Kelvin Grove
> > > > > Queensland 4059
> > > > > Australia
> > > > >
> > > > > http://www.ihbi.qut.edu.au/
> > > > >
> > > > > Phone +61 (0)7 3138 6222
> > > > > Mobile +61 (0)413 165 805
> > > > > Email j.swedberg_at_student.qut.edu.au
> > > > > joakim_swedberg_at_hotmail.com
> > > > >
> > > > >
> ----------------------------------------------------------------------
> > > > >
> > > > >
> > > >
> > >
> > > --
> > > Dr. Axel Kohlmeyer akohlmey_at_gmail.com
> > > http://sites.google.com/site/akohlmey/
> > >
> > > Institute for Computational Molecular Science
> > > 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
> > 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