VMD-L Mailing List

From: John Stone (johns_at_ks.uiuc.edu)
Date: Wed Nov 29 2006 - 14:25:37 CST

Christian,
  VMD performance in the CAVE is very dependent on how you have
the system setup. If you're running in stereo, that will already
be a factor of 2x slower than what you see running VMD in non-CAVE
mode. Another key issue when running in the CAVE is whether you've
properly configured your CAVE system to run on multiple processors.
If you've got a single processor machine driving several displays,
this would be another reason for a drop in performance. VMD is designed
to take advantage of multiple processors when running in CAVE mode,
each processor drives a separate graphics board, and so the performance
of the program should remain nearly constant so long as you have at least
one CPU per display, and one CPU for the main VMD process. For example,
the NCSA CAVE runs on a 12 processor machine and drives 6 walls, more
than enough processors for the amount of computation and display work
that VMD performs. There's a small amount of synchronization overhead
among the rendering processes and the main VMD process, but we use an
efficient reader/writer lock strategy that performs very
well in our tests here, so the performance running on a one-walled
CAVE versus a six-walled CAVE should be similar assuming that you
scale the number of processors and graphics accelerators appropriately.

Another minor issue with running VMD in the CAVE is that not all
of the same graphics features can be used in CAVE mode as we can use
in normal OpenGL, so you may lose a small amount of speed there
when compared to a non-CAVE VMD run.

Regarding trackers, in order to use the CAVE wand/trackers
with VMD you need to create a .vmdsensors file. If your CAVE
coordinate system is setup the same way that the NCSA CAVE is,
then it will work "out of the box" with the example .vmdsensors
file we provide with VMD. If not, you may need to either edit the
.caverc file you're using, or else edit the .vmdsensors file to apply
the appropriate translation, rotation, and scaling operations to the
tracker coordinates and orientation. Let me know if you need help
with this.

  John Stone
  vmd_at_ks.uiuc.edu

On Tue, Nov 28, 2006 at 03:16:48PM +0100, Christian Wohlschlager wrote:
> ** Reply Requested When Convenient **
>
> Dear John !
>
> Thank you
> The Help CAVELib is now running the Tracker doesn't realy do much ( in
> fact nothing) but i think that a configuration problem becouse i can see
> the Head sensor i moving around the molekul.
>
> The options now working is LINUXIA64 OPENGL CAVE FLTK TK NETCDF TCL
> PTHREADS SILENT SHARED
>
> I can see its quite slow (should be 10 x Faster) maybe some parameters
> for the cave graphik are not really good)
>
> maybe you can give me some hints .
>
>
> thank you
>
> christian 

-- 
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