From: Yinglong Miao (yimiao_at_indiana.edu)
Date: Tue Nov 20 2007 - 12:46:10 CST
I have done a lot of testing on the timestep for the slow variables on
several systems. When I use small enough timesteps, the simulations can
keep going without being crashed. But it doesn't speed up the MD much,
which is one of my main goals. Assuming I run short MD for dt and
advance the structure for dT by projecting slow variables, I want to
make dT/dt as large as possible. So I want to handle the global
exclusion count error from MD run to make dT adaptive.
Quoting Victor Ovchinnikov <ovchinnv_at_MIT.EDU>:
> I think that when you integrate the slow variables you should use a much
> smaller timestep than you must be doing. If you want to use a large
> timestep, then you can do inner iterations -- e.g. advance for dt/10,
> run short MD to equilibrate, advance for another dt/10, equilibrate,
> etc., until you get to t+dt.
> I cannot think of any method that will tell you whether your coarse time
> step is too large
> On Tue, 2007-11-20 at 11:10 -0500, Yinglong Miao wrote:
>> Hi, Peter,
>> Thanks for your reply!
>> I read both the trouble shooting webpage you referred to and the user
>> guide regarding the errors. I also knew what's wrong when I looked at
>> the ouput trajectory of my simulations. There were unreasonable
>> stretched bonds and angles in the structure. This cannot be avoided in
>> my simulations, because I am implementing a method to advance a
>> macromolecule over long time by using the values of some slow variables
>> I define and their evolution from short MD runs. As implied by my
>> method, the slow variables projected over long time may lead to an
>> unreasonable atomic structure. MY QUESTION IS HOW TO DETERMINE THERE
>> WILL BE "BAD GLOBAL EXCLUSION COUNT" ERROR FROM THE UNREASONABLE
>> STRUCTURE IN NAMD. If I know how to do this, I can go back to reduce
>> the projection time for slow variables until getting a reasonable
>> structure to continue the simulation with MD run, instead of going
>> crashed (the default pathway provided by NAMD).
>> I am also thinking of checking the forces and energies calculated from
>> the structure, but it seems only forces on several atoms can be
>> reasonably loaded and I don't know how to load the energies. Do you
>> have any ideas about this?
>> Quoting Peter Freddolino <petefred_at_ks.uiuc.edu>:
>> > Hi Long,
>> > you can get a feel for what causes these errors at
>> > http://www.ks.uiuc.edu/Research/namd/wiki/?NamdTroubleshooting.
>> > The two most important things to do if you're encountering this error
>> > are make sure your system was sufficiently minimized, and look at your
>> > trajectory in vmd (possibly writing to the dcd file at every timestep)
>> > so see if you can see where the instability is occurring.
>> > Best,
>> > Peter
>> > Yinglong Miao wrote:
>> >> Hi, NAMD developers and users,
>> >> In some of my simulations, I keep getting error messages like "FATAL
>> >> ERROR: Bad global exclusion count!" and "atoms moving too fast". I am
>> >> wondering whether there are flag values returned when executing
>> >> "minimize" and "run". Let's say, when these commands are executed
>> >> correctly, a flag value of 0 is returned; when above error messages
>> >> come up, a flag value of -1 is returned so that I can do something to
>> >> "guide" the simulation according to the flag value instead of just
>> >> going crashed.
>> >> If no such flag values returned in NAMD, could anybody tell me how is
>> >> the error "FATAL ERROR: Bad global exclusion count!" determined? I
>> >> know it could be due to stretched bonds, too close atoms or other
>> >> unphysical structure, but how does it work in computation? And how to
>> >> analyze and evolve the atomic structure to avoid above errors?
>> >> Any ideas and suggestions will be greatly appreciated!
>> >> Thanks,
>> >> Long
>> >> --
>> >> Yinglong Miao
>> >> Ph.D. Candidate
>> >> Center for Cell and Virus Theory
>> >> Chemistry Department, Indiana University
>> >> 800 E Kirkwood Ave Room C203A, Bloomington, IN 47405
>> >> 1-812-856-0981(office); http://ylmiao.dict.cn/mypage/
>> Yinglong Miao
>> Ph.D. Candidate
>> Center for Cell and Virus Theory
>> Chemistry Department, Indiana University
>> 800 E Kirkwood Ave Room C203A, Bloomington, IN 47405
>> 1-812-856-0981(office); http://ylmiao.dict.cn/mypage/
-- Yinglong Miao Ph.D. Candidate Center for Cell and Virus Theory Chemistry Department, Indiana University 800 E Kirkwood Ave Room C203A, Bloomington, IN 47405 1-812-856-0981(office); http://ylmiao.dict.cn/mypage/
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:45:34 CST