VMD-L Mailing List

From: Axel Kohlmeyer (akohlmey_at_gmail.com)
Date: Fri Jan 07 2011 - 09:01:48 CST

On Fri, Jan 7, 2011 at 9:40 AM, Nima Nouri <nouri.f.nima_at_gmail.com> wrote:
> Thanks a lot
> Axel
> what a super fast answer.

please always keep the list in cc: so that people see
how your problem is resolved.

> Really Thank you.
>
> I just need it for visualizing the output of my MD program.
> the 32bit version works well for small models but when I increase the size
> of my model (lets say 2 Million Particles) after a short period (30min) VMD
> hangs. Although the simulation is still going forward. I am using the mol2
> and dcd output file formats + some commands for graphical visualization of

i don't quite get what you are doing. if you have that many particles and want
to keep everything in memory, then you have to have an insane amount of
memory in your machine. i am not sure if a "normal" windows can even handle
that well. i always found the memory management on windows inferior to
macos and linux.

> the atoms with VDW visualization type. I am also using PBCtools for showing
> the Periodic box size change during the simulation. BTW I use imd for online
> visualization of the results.

well, none of that uses CUDA acceleration, so compiling in CUDA support
won't help much.

> the last week I recognized that the slowness was because of the lack
> of memory (when VMD allocates about 3.4GB of memory it hangs which is a
> normal thing to a 32bit application). I have tried to decrease the number of
> trajectory outputs send to imd but it does not change the situation.

if you enable the GLSL rendering, then VMD will take advantage
of the processing power of your GPU, but visualizing millions of
particles with VDW is pretty pointless. you just don't have enough
pixels and you can get a better (and much faster) representation
of your system with "Points" in combination with depth cueing.
that is what i do in these cases.

> I have already installed a linux version but for me as a windows user it is
> really time consuming to switch between two os everytime I want to run Sth.

well, there always is the option to run two system images at the same
time through virtualization. i personally have long ago abandoned using
windows, since it is so incredibly inconvenient for scientific applications.
i found myself installing all kinds of unix utilities on my windows machine,
so that i then came to the conclusion it would be better to have the
real thing.

> Another Question is How can I remove the bonds croosing the PBC from
> display?

you either have to output your trajectory already in unwrapped coordinates
or use a combination of pbc join/unwrap/wrap to keep molecules together.

for the most part VMD is not at all trying to be smart, but rather shows
the data that you feed it, just the way it is.

in general, if you are working with large data sets, you have to be a
bit smarter about how you go about it. i often strip of the solvent or
save a coarse grained representation of the molecules for visualization
of the full data set and otherwise operate on parts or use offline
processing through the bigdcd script.

cheers,
   axel.

> At the end sorry for the bad english and thanks a lot for the great
> software.

>
> Regards
> N.Nouri 

-- 
Dr. Axel Kohlmeyer
akohlmey_at_gmail.com  http://goo.gl/1wk0
Institute for Computational Molecular Science
Temple University, Philadelphia PA, USA.