[gmx-users] Gromacs on GPU

[Matthew Lardy]

I should add that while we had runs that completed, all we got were
NaN for the coordinates and energies.

Personally, I have attributed our inability to execute the GPU code
cleanly to our C2050 Teslas [they are double precision].  (As I had
the GPUized code working with an earlier card, although the options
were pretty limited)

Matt

On Sat, Jan 28, 2012 at 11:32 AM, Benjamin Hall
<benjamin.a.hall at ucl.ac.uk> wrote:
> Justin A. Lemkul wrote:
>
>
> We get junk output with 4.5.4 if the box size is set to zero, for some odd
> reason, though in principle it should be ignored with "pbc = no" in the
> .mdp
> file.  Using any non-zero box results in usable results.  Our OpenMM version
> is
> the same, but Cuda is in the 3.x series (can't remember off the top of my
> head,
> sorry).
>
>
> I've tried setting the box size in the gro file but that doesn't seem to
> work (100x100x100nm). Which compiler did you use? I realise that this is an
> unlikely source of trouble but it's another variable to eliminate in
> testing.
>
> I'll update issue 568 on redmine, which has been closed but appears to be
> this problem.
>
> Thanks & best wishes
>
> Ben
>
>
> --
> Dr Benjamin A Hall
> Centre for Computational Science, Department of Chemistry, UCL
> benjamin.a.hall at ucl.ac.uk
>
>
>
>
>
> --
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/Search before posting!
> Please don't post (un)subscribe requests to the list. Use the
> www interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/Support/Mailing_Lists