Hi Yong,
I don't see any obvious reason why what you are writing should not work. Can you try it with the CPU code and see if that works please?
All the best
Ross
> On Nov 25, 2016, at 5:54 AM, Yong Wang <wyongciac.gmail.com> wrote:
>
> Hi,
>
> We tried to run a pH-REMD simulation with 8 replicas on a GPU node (having
> 8 CPUs and 4 GPUs). The script is pasted as below:
>
> unset CUDA_VISIBLE_DEVICES
>
> export CUDA_VISIBLE_DEVICES=0,1,2,3
>
> mpirun -np 8 $AMBERHOME/bin/pmemd.cuda.MPI -ng 8 -groupfile groupfile -rem
> 4 -remlog pHremd.log
>
> But the results suggested we'd better to have 12 or 16 replicas. So we
> tried to submit 16 replicas in the same node, using:
>
> mpirun -np 16 $AMBERHOME/bin/pmemd.cuda.MPI -ng 16 -groupfile groupfile
> -rem 4 -remlog pHremd.log
>
> But it failed with an error of "MPI size is not a multiple of -ng". We
> couldn't understand why it works for 8 replicas but not 16. Or is there any
> trick to make it work? Thanks.
>
> Regards,
>
> Yong
>
> --
>
> ==========================
> Yong Wang
> PhD Student, Structural Biology and NMR Laboratory (SBiNLab)
> Department of Biology, University of Copenhagen, Denmark
> Ole Maaloes Vej 5, DK-2200 Copenhagen N
> https://www.researchgate.net/profile/Yong_Wang15
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Nov 25 2016 - 08:00:04 PST
