I believe that is a sander-only flag, not pmemd.
adrian
On 11/17/13 6:17 PM, yunshi11 . wrote:
> Hi everyone,
>
> I am testing pmemd.MPI (of amber12) on parallel machines (e.g. 72 CPUs),
> and trying to tweak parameters to improve parallel efficiency.
>
> My .in is like:
>
> Plain MD
> &cntrl
> ntx=5, irest=1,
> ntpr=500, ntwx=5000, ntwe=5000, ntwr=5000,
> ntc=2, ntf=2, ntb=2, igb=0, cut=10, nsnb=10, skinnb=2, column_fft=1,
> nstlim=50000000, dt=0.002, t=0,
> ntt=1, tautp=1, temp0=298.0,
> ntp=1, taup=2.0, pres0=1,
> &end
>
> But I got something like "amber invalid reference to variable in NAMELIST
> input, unit 5, ...", pointing to my setting of column_fft=1, by which I
> thought could make parallel runs faster. The Manual does say "The column
> method can be faster for larger processor counts since..."
>
>
> So does anyone know why did I get such error message?
>
> And which parameters should I play with to improve parallel efficiency?
>
> Best,
> Yun
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
--
Dr. Adrian E. Roitberg
Colonel Allan R. and Margaret G. Crow Term Professor.
Quantum Theory Project, Department of Chemistry
University of Florida
roitberg.ufl.edu
352-392-6972
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Nov 17 2013 - 15:30:03 PST
