AW: AW: AW: Periodic cell has become too small for original patch grid

From: Norman Geist (norman.geist_at_uni-greifswald.de)
Date: Wed Aug 15 2018 - 09:52:13 CDT

Cause NAMD was there initialized with the same smaller box at all replicas. And as I stated, the box may just not become much smaller than the initial box. ;)

> -----Ursprüngliche Nachricht-----
> Von: Rohit Modee [mailto:rohit.modee_at_research.iiit.ac.in]
> Gesendet: Mittwoch, 15. August 2018 13:06
> An: Norman Geist <norman.geist_at_uni-greifswald.de>
> Cc: namd-l_at_ks.uiuc.edu
> Betreff: Re: AW: AW: namd-l: Periodic cell has become too small for original
> patch grid
>
> what happens when in the first run the replica at higher temperature
> exchanges with lower temperature. box will shrink in this case too so why is it
> not giving error.
>
> Regards,
> Rohit
>
> ----- Original Message -----
> From: "Norman Geist" <norman.geist_at_uni-greifswald.de>
> To: namd-l_at_ks.uiuc.edu, "Rohit Modee" <rohit.modee_at_research.iiit.ac.in>
> Sent: Wednesday, August 15, 2018 4:22:38 PM
> Subject: AW: AW: namd-l: Periodic cell has become too small for original
> patch grid
>
> The error message you see is basically related to NAMDs domain
> decomposition. The box is departed into patches, which are then distributed
> across workers in order to accelerate the computation based on
> parallelization. This decomposition is only done at NAMDs startup. So it's not
> a problem if the box expands after this initialization, but if it shrinks too
> much.
>
> Assuming you are using a usual temperature range for REMD, what do you
> think will happen to your cell at temperatures above the boiling point of
> water? Yes, the box size expands to a very large number. At the 1st run this is
> not a problem for NAMD to continue since the initialization was done before,
> although it doesn't make physical sense for your REMD, IMHO. At the 2nd
> run, NAMD initializes very different domain decompositions across the
> replicas, since some of them already have the huge box. If such a replica now
> exchanges to a lower temperature, the box shrinks a lot and NAMD aborts
> with the respective message.
>
> How to solve this now? Well, I personally used a temperature dependent
> target pressure computed from the ANTOINE equation multiplied by three.
> The supplements include my the REMD code with the respective changes
> https://pubs.acs.org/doi/10.1021/acs.jpcb.8b05178
>
> Another option is: 1. Pre-equilibrate the cell to the base temperature and
> disable NPT during the REMD. Depending on the dimensions of the expected
> conformational changes of your solute, using NPT might be required or not. A
> small solute will probably work without NPT just fine, since conformational
> changes will not suppress too much water so that the overall density in your
> cell will not rise significantly.
>
> Best of luck
> Norman Geist
>
> > -----Ursprüngliche Nachricht-----
> > Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im
> > Auftrag von Rohit Modee
> > Gesendet: Mittwoch, 15. August 2018 11:35
> > An: Norman Geist <norman.geist_at_uni-greifswald.de>
> > Cc: namd-l_at_ks.uiuc.edu
> > Betreff: Re: AW: namd-l: Periodic cell has become too small for original
> patch
> > grid
> >
> > yes, I am doing NPT REMD.
> >
> > langevinPistonTarget 1.01325
> >
> > dimesnsions of box
> > # Periodic Boundary Conditions
> > cellBasisVector1 51.579999923706055 0 0
> > cellBasisVector2 0 51.4640007019043 0
> > cellBasisVector3 0 0 51.75699996948242
> > cellOrigin 0.225006103515625 0.10952535271644592 -0.10641514509916306
> >
> > MARGIN 1.0
> > PMEGridSpacing 1.0
> >
> > ----- Original Message -----
> > From: "Norman Geist" <norman.geist_at_uni-greifswald.de>
> > To: namd-l_at_ks.uiuc.edu, "Rohit Modee" <rohit.modee_at_research.iiit.ac.in>
> > Sent: Tuesday, August 14, 2018 11:58:06 AM
> > Subject: AW: namd-l: Periodic cell has become too small for original patch
> grid
> >
> > Are you doing NPT remd?
> > What target pressure are you using?
> > What are the box dimensions?
> >
> > > -----Ursprüngliche Nachricht-----
> > > Von: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] Im
> > > Auftrag von Rohit Modee
> > > Gesendet: Montag, 13. August 2018 10:26
> > > An: namd-l_at_ks.uiuc.edu
> > > Betreff: namd-l: Periodic cell has become too small for original patch grid
> > >
> > > Dear All,
> > >
> > > I am running REMD simulations. have divided 100 ns into 5 jobs of 20 ns
> > each.
> > > my job0 of 20 ns is complete but when i start job1 from checkpoint files
> > from
> > > job0 it gives me following error at 22ns
> > > Reason: REPLICA 9 FATAL ERROR: Periodic cell has become too small for
> > > original patch grid!
> > > Possible solutions are to restart from a recent checkpoint,
> > > increase margin, or disable useFlexibleCell for liquid simulation.
> > >
> > > can some one please explain what this error technically means.
> > >
> > > i am already using margin of 1 A my water box is cubic with l=50 A
> > >
> > > in log file i get this value "PME GRID DIMENSIONS 54 54 54" why is this
> > 54
> > > instead of 50.
> > >
> > > what effect will it have if
> > > 1. I restart job1 with margin as more than 1.
> > >
> > > regards,
> > > Rohit

This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:21:21 CST