On Wed, Feb 18, 2015 at 4:02 PM, Kenneth Huang <kennethneltharion.gmail.com>
wrote:
> Dear all,
>
> I'm trying to run MMPBSA.py.MPI on a local cluster, and my system contains
> a K+ that I've added into mdread.F90 as this entry-
>
> else if (atype == 'K+') then
> ! K+ ionic radius is 152 pm to 1.52 angstroms
> ! See archive.ambermd.org/201101/0220.html for more info
> x(L165-1+i) = 1.52d0 + 1.4d0
>
> I've used the same mdread.F90 file on several different local machines with
> the same system, and it works without a problem. MMPBSA works fine for
> other systems, like the one in the tutorial.
>
> However, when I try to run MMPBSA on my system with the cluster, I get a
> bad atom type: K+ error. So I had my new entry added to the mdread.F90
> file, the cluster admin recompiled AmberTools14 (only the parallel version)
>
That's your problem. It may sound confusing, but you actually need the
*serial* version of Amber to run the parallel version of MMPBSA.py. The
reason is that the parallelism is implemented on the Python side, so there
are multiple Python threads that each run a serial sander calculation.
That results in *much* better scaling on average than splitting a single
frame amongst processors.
Reconfigure and recompile in serial, and it should work.
HTH,
Jason
--
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Feb 18 2015 - 14:30:03 PST
