[gmx-developers] domain decomposition

[Jason DeJoannis]

Hi All, 
To get an idea how the domain decomposition works I simulated 500   
Argon/Krypton atoms at STP on 2 nodes with a boxlength of about 13nm.  
I plotted (see xmgrace attachment) the x-coordinate vs step  
for the middle particle (i=250). The position evolves smoothly  
and strays far from the middle of the box indicating that it  
must be the same particle the whole time. I conclude that the  
domain decomposition into X-coordinate-based slabs is only  
done at the begining of the simulation. Thereafter the 'slabs'  
can be far from physical domains. That raises the question of  
how is the grid-based neighbor search done. It is not sufficient  
to communicate cells along the slab boundaries. What is going on here?  
  
thanks,  
  
  -Jason  
  
  
  
---  
Jason de Joannis, Ph.D.  
Chemistry Department, Emory University  
1515 Pierce Dr. NE, Atlanta, GA 30322  
Phone: (404) 712-2983  
Email: jdejoan at emory.edu  
http://userwww.service.emory.edu/~jdejoan  
  
  
  
  
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: grompp.mdp
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20031001/2092b1e1/attachment.ksh>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: trajectory.xmgr
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-developers/attachments/20031001/2092b1e1/attachment-0001.ksh>