[MMTK] Auto-discard notification

Konrad Hinsen hinsen at cnrs-orleans.fr
Fri Jun 23 18:03:59 CEST 2006


On Jun 22, 2006, at 11:51, Sanzo Miyazawa wrote:

> Although I cannot understand why the order of improper DH depends  
> on a computer,
> if so, it can be explained why different runs on different computers
> and even on the same computer yield completely different MD  
> trajaectories.

The order depends on details of memory allocation inside the Python  
interpreter. This does not depend on the hardware of the computer,  
but it can well depend on subtle aspects of its configuration. In  
principle, the different machine name could be sufficient. You can  
also get different results in multiple runs on a single machine under  
some conditions.

> However, evern though the difference is small, any difference on the
> energy gradient can lead the system to the completely different and
> distant local minimum.

Indeed. This is the well-known fact that Molecular Dynamics  
trajectories are very sensitive to initial conditions, round-off  
errors, etc.

> By the way, if you have a patch for MMForceField.py to do so,  
> please tell me.

Not at the moment, but that would not be much work. Perhaps I'll add  
it as an option to the Amber force field.

Konrad.
--
---------------------------------------------------------------------
Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: hinsen at cnrs-orleans.fr
---------------------------------------------------------------------





More information about the mmtk mailing list