Right, thank you so much, David!
just one question related to the compilation of the pmemd.cuda
I noticed several versions have been produced as the results of the
installation:
pmemd pmemd.cuda_DPFP.MPI pmemd.cuda_SPFP.MPI
pmemd.cuda pmemd.cuda.MPI pmemd.MPI
pmemd.cuda_DPFP pmemd.cuda_SPFP
In what case do I need to run a simulation using a double precision
version pmemd.cuda_DPFP. ?
Would it be much slower compared to the pmemd.cuda?
Cheers,
Enrico
Il giorno ven 5 ago 2022 alle ore 14:30 David A Case
<david.case.rutgers.edu> ha scritto:
>
> On Fri, Aug 05, 2022, Enrico Martinez wrote:
>
> >So after compiling and testing the MPI version, I have not noticed a
> >difference between 1 or 2 gpus for my system.
>
> I never use pmemd.cuda.MPI myself, preferring to run separate simulations in
> parallel, each on a single CPU. You might run some short tests to see if
> you get much parallel speedup with 2 GPUs. It depends a lot on your system.
>
> > nstlim=250000000, dt=0.002,
>
> Your comments make is sound like you are rather new to running simulations
> on GPUs. Consider gaining some quicker experience with some simulations
> that are 100 times shorter, to test things like timing, being sure that you
> know how to run mmpbsa_py, etc. You can go back to longer simulations
> later.
>
> >P.S. do I need &end after &ewald?
>
> You need an ending sigil. "/" is correct one; "&end" is obsolete, but still
> OK on most systems, apparently including yours.
>
> ....dac
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Aug 05 2022 - 06:30:03 PDT
