Dave's reply is exactly on target. Now, the reason. If you do shake with
ntc 3, then you shake all bonds I believe, not just the ones containing H.
Thus, when you partition the problem in parallel, coordinate partitioning
has to be molecule-based. So one can in theory fragment the molecules and
have the owners of the fragments communicate, but since shake is an
iterative process that effects the coordinates, this gets very very nasty
from a performance perspective. I can think of other approaches; I just
have not been had sufficient motivation - no one is telling me that H-only
shake is a major problem in MD simulations (this is an invitation for folks
to correct any misconceptions I may have in this area :-)). Also, as Dave
says, sander has this exact same limitation.
Regards - Bob Duke
----- Original Message -----
From: "oguz gurbulak" <gurbulakoguz.gmail.com>
To: <amber.ambermd.org>
Sent: Monday, February 02, 2009 8:32 AM
Subject: [AMBER] md simulation on parallel PMEMD
> Dear All,
>
> I'm trying to run a md simulation on parallel PMEMD. I have 6
> equilibration
> file with rising the temperature for 50 K ( 0-273 K ). My first
> equilibration file is in below. I used ntc=3, ntf=3, and I got this error
> in
> output file :
>
> 3. ATOMIC COORDINATES AND VELOCITIES
> --------------------------------------------------------------------------------
>
>
>
> begin time read from input coords = 0.000 ps
>
> Number of triangulated 3-point waters found: 0
> this parallel version only works for ntc < 3
>
> Isn't it possible to use ntc=3, ntf=3, on parallel PMEMD ? Or I should
> just
> use ntc=2, ntf=2, on parallel PMEMD . or I can only use sander with
> ntc=3,
> ntf=3, ? Could you help me to slove this problem ?
>
> Thanks in advance.
>
>
> MD equilibration
> &cntrl
> imin=0, irest=0, nstlim=5000000, dt=0.002, ntx=1,
> ntc=3, ntf=3, ntpr=2000, ntwx=2000, ntwr=2000, ntwe=2000, ntwv=2000,
> cut=9.825, ntb=1, ntp=0, ntr=0, ntt=1,tautp=2.0,
> tempi=0.0, temp0=50.0,
> /
> _______________________________________________
> 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 Wed Feb 04 2009 - 01:18:20 PST
