[gmx-users] parallel version of gromacs

[Maira d'Alessandro]

Hallo! 
I made some tests about the performances of GROMACS when changing the number of
processors, in the parallel version.
I found different results in the energy from 1 to 10 processors.
Do you know if it is well known? 
Is it due to the implementation of gromacs parallel version or is it intrinsec
with the hardware of clusters?
The communications among processors is in single or double precision (I use
always the double precision gromacs package)?
If somebody knows, please give me some references about this problem.
thank you to everybody.

-- 
Maira D'Alessandro   
e-mail : m.dalessandro at caspur.it  
  
Post-Doc researcher   
University of Rome "Tor Vergata"    
Dep. of Chemical Science and Technology  
Viale della Ricerca Scientifica, 00133 Roma   
tel 0039(0)672594905



------------------------------------------------
This mail sent through Caspur Webmail powered by
IMP (http://horde.org/imp/)