Antw: Re: Antw: [gmx-users] tpbconv for following runs

[Anton Feenstra]

Alan Wilter Sousa da Silva wrote:
> Ok, I messed some informations but how t coupling is affecting this since
> she is continuing a simulation?  And what if one is recovering a crash
> simulation?

Recovery and continuation are identical procedures, using identical
restart information (i.e. an X+V frame from .trr).

> Is all about bUncStart=TRUE or FALSE?

No, it's about x at t, v at t+dt/2 (which come from the trr) and v at
t-dt/2 (which you don't have, dt is your integration timestep of 1 or
2 fs, not your frame interval of 1 ps or so). So, you don't have the
information necessary to calculate the average v (v at t), which you
need for accurate Ekin calculation (which goes into temp, and pres).

It does seem that for many people this is a real obstacle, so I'd guess
we should spend some time in finding a good solution for it and
implementing that. One quick and dirty hack would be to always write
two consecutive x+v frames to the .trr frame. That needs probably only
a trivial modification to mdrun.c. I'll try and have a quick look at
it, and post the additional line or two of code, so people who need
it now can fix it themselves.

However, IMHO, it *really* *doesn't* *matter*! MD is a chaotic process,
and any small change (hardware, code, compiler, T-coupling, precision)
will amplify itself in time. If the change is small enough not to be an
actuall error in its own right, you should be able to live with it.
So, perhaps to motivate my finding a quick solution, could someone
explain why in certain cases it might be so important to reproduce the
exact same run? (This is not an attempt at being cynical ;-)


-- 
Groetjes,

Anton
  _____________ _______________________________________________________
|             |                                                       |
|  _   _  ___,| K. Anton Feenstra                                     |
| / \ / \'| | | Dept. of Pharmacochem. - Vrije Universiteit Amsterdam |
|(   |   )| | | De Boelelaan 1083 - 1081 HV Amsterdam - Netherlands   |
| \_/ \_/ | | | Tel: +31 20 44 47608 - Fax: +31 20 44 47610           |
|             | Feenstra at chem.vu.nl - www.chem.vu.nl/~feenstra/       |
|             | "If You See Me Getting High, Knock Me Down"           |
|             | (Red Hot Chili Peppers)                               |
|_____________|_______________________________________________________|