[gmx-users] Question
Mark Abraham
mark.j.abraham at gmail.com
Wed Aug 12 22:38:43 CEST 2015
Hi,
On Wed, Aug 12, 2015 at 9:47 PM Eric Smoll <ericsmoll at gmail.com> wrote:
> Hello gromacs users,
>
> I am running a set of MD simulations where we are only interested in
> logging the coordinates and kinetic energy of a specific group in the
> system. I want to write the coordinates and energy of this group to file at
> every timestep.
>
That's not an operation that mdrun supports. You can get all the
coordinates and velocities, or a subset of the coordinates at a user-chosen
precision, but that's all.
Supplying an index file with the group of interest to grompp and specifying
> this group in the mdp file under "energygrps" doesn't seem to impact the
> edr file. The edr file still outputs the total system energy as a function
> of time.
This is a limitation of the current GPU implementation - energygrps is not
supported (and your md.log file will say this).
Is there someway I can request that gromacs only compute the
> energy of one group in my system?
Only when you postprocess the trajectory on just the CPU, e.g. with mdrun
-rerun traj -nb cpu
Will this result in a speedup?
>
No. The bulk of the work is in the forces, so any savings from the energy
calculation on only part of the interactions would be offset by the need to
run a separate GPU kernel and do more book-keeping, so probably no gain.
But as above, none of this is implemented.
> Are there other ways I can increase the speed of my simulation? My systems
> have 20,000-40,000 atoms with PME and 1.5 nm cutoffs. Are these simulations
> well suited for GPU acceleration or is this unlikely to produce a speedup?
>
Speedup on a single GPU with a comparable CPU, yes.
Mark
>
> Best,
> Eric
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.
>
More information about the gromacs.org_gmx-users
mailing list