VMD-L Mailing List

From: Ajasja Ljubetič (ajasja.ljubetic_at_gmail.com)
Date: Wed Jan 18 2012 - 07:17:32 CST

Hi,

The new CUDA-calculated representation is called QuickSurf. Under windows
all you have to do to get CUDA to work is to install CUDA compatible
drivers. Version 285.86 downloaded from
http://developer.nvidia.com/cuda-toolkit-41 works for me just fine.
If you like you can send me the problem PDB and I'll try to crash my VMD :)

Best regards,
Ajasja

On Wed, Jan 18, 2012 at 13:12, FX <fxcoudert_at_gmail.com> wrote:

> Hi all,
>
> I have a strange problem with the "Surf" representation (molecular
> surfaces). I have a PDB file of a porous material, which when loaded into
> VMD and switched to "surf" representation:
>
> -- on my Mac OS 10.7.2, with VMD 1.9 (not the latest beta), works fine
> -- on our x86_64 linux boxes (without CUDA), alternatively does not work
> (display seemingly random triangles) or crashes VMD as a whole, depending
> on probe size used; this happens with both versions 1.9 and 1.9.1
> -- on our Windows 7 64-bit box (which has CUDA accelerators), same thing
> as Linux, both with 1.9 and 1.9.1
>
> I cannot make any sense of that. I know that the "surf" code is not the
> most stable algorithm, but that's very impressive crashing rate. Moreover,
> I thought there was a new CUDA-based code for calculation of molecular
> surfaces (the "surf" representation) in 1.9.1, and so I expected it to work
> fine on the Windows system (which detects the graphic card fine, and states
> that CUDA accelerators are present).
>
> So, how do I debug this? Are any of my assumptions wrong? How could I get
> the CUDA version of the code to run on Windows (hoping it is more stable
> that the external SURF program)?
>
> Many thanks in advance,
> FX Coudert
>