Multiple energy groups is not implemented for GPUs

User discussions
1

GROMACS version:2020.4
i was run this command

gmx mdrun -deffnm nvt -nb gpu

and got in

nvt.log

GPU info:
Number of GPUs detected: 1

Multiple energy groups is not implemented for GPUs, falling back to the CPU. For better performance, run on the GPU without energy groups and then do gmx mdrun -rerun option on the trajectory with an energy group .tpr file.

does energy group .tpr file is nvt.tpr ?

2

Hi,
energy group are defined in the parameter file (mdp file)
using the option

  • “energygrps”: group(s) for which to write to write short-ranged non-bonded potential energies to the energy file (not supported on GPUs)

You can modify the mdp file (removing the energy groups), re-generate an tpr file, perform the simulation using GPU, and then re-run following the instruction given in the error .

For mdp file option see

Best regards
Alessandra

3

thanks it’s done

4

Hello, I had the same issue for my mdrun, so thank you for this thread!

I ran on the GPU without the energygrps like Alessandra suggested but now I am a little stuck on what to do next.

I know I have to create the .tpr file with the energygrps and then run ‘gmx mdrun -rerun’ with that tpr file, like the error says but I am not sure how to do this?
Is it the same as running a normal md run? So, does it take the same amount of time and memory? As in should I have the same GPU memory and time allocated as to the original simulation?

And would the command be something like:
gmx mdrun -s new.tpr -rerun md_rerun.xtc

5

No, the recalculation will be much faster. You do not need to worry about GPU memory or anything as the rerun calculation must be performed on CPU. This is the only way to use energygrps.

Add -nb cpu and you’ve got it.

6

Amazing, thank you very much!!!