VMD-L Mailing List

From: Andrews, Casey (casey-andrews_at_uiowa.edu)
Date: Wed Nov 27 2013 - 12:28:22 CST

Hi John,

Thank you very much for your quick response and your suggestions, both of which were really helpful.

Best wishes,
Casey Andrews
________________________________________
From: John Stone [johns_at_ks.uiuc.edu]
Sent: Tuesday, November 26, 2013 7:02 PM
To: Andrews, Casey
Cc: vmd-l_at_ks.uiuc.edu
Subject: Re: vmd-l: Speeding up the VDW drawing method and Tachyon rendering process

Hi,
  I'm traveling (using a tethered cell phone) so I will have to keep
this brief:
  1) enable Display|Rendermode|GLSL, then try using the "Points"
     rep and crank up the size to taste, this is extremely fast, but
     not as nice looking as the normal VDW rep. Be sure you're using
     GLSL or they will not look like spheres and will instead look like
     circles...

  2) Yes, parsing files is a single-threaded process, and not something
     that's likely to change. You'll note that this is the case with
     nearly all renderers. However, this is completely avoidable:
     Don't use "Tachyon" which writes the scene to a file. Use the
     "TachyonInternal" renderer instead, which renders straight out of
     the VMD in-memory data structures...

Try thse suggestions and let me know if you need further help.
Hopefully I'll have a better network link by then...

Cheers,
  John Stone
  vmd_at_ks.uiuc.edu

On Tue, Nov 26, 2013 at 11:53:02PM +0000, Andrews, Casey wrote:
> Hello,
>
>
>
> Currently we are trying to make visual images of a large system that
> contains a few millions of atoms. We would like to represent the molecules
> in each image as VDW. However, this takes a very long time to complete and
> seems to be only using one CPU core out of a possible 12. Additionally,
> the VDW representation does not seem to be taking advantage of the GPU we
> have installed (Nvidia Telsa 20C), but we know the GPU works because the
> Quicksurf representation uses it. We were curious if there was a way in
> which creating the VDW representation for millions of atoms could be made
> faster or if there was a faster alternative to VDW that would have the
> same visualization effect. We have tried to switch the render mode to GLSL
> but there wasn*t a huge speed increase in the VDW rendering process.
>
>
>
> An additional question that we have is about the speed of the Tachyon
> rendering process on our large system. It appears that the scene parsing
> phase of the rendering process is only using one thread and it takes a
> very long time for each image. However once the scene parsing phase
> is completed, then the process is parallelized and happens quite quickly.
> Again we were wondering if there was a way in which the earlier stage of
> the process could be sped up?
>
>
>
> For the rendering of these large systems, we have tried both VMD 1.9.1 and
> VMD 1.9.2. The OS we are using is Ubuntu version 13.10. Our Nvidia driver
> for the GPU is version 319.60 and our CUDA complier is release 5.0. Any
> help or guidance concerning the above issues would be greatly appreciated.
>
>
>
> Thank you very much for your work on an amazing program,
>
> Casey Andrews
>
> 

--
NIH Center for Macromolecular Modeling and Bioinformatics
Beckman Institute for Advanced Science and Technology
University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
http://www.ks.uiuc.edu/~johns/           Phone: 217-244-3349
http://www.ks.uiuc.edu/Research/vmd/