[gmx-users] Re: Hi, May I ask you a question about the free energy calculation tutorial?

[Justin A. Lemkul]

yywang wrote:
> 
> 
> 
> 
> At 2011-12-31 04:42:12,"Justin A. Lemkul" <jalemkul at vt.edu <mailto:jalemkul at vt.edu>> wrote:
>>
>>
>>yywang wrote:
>>>  Dear Justin Lemkul,
>>>          I find a tutorial about the free energy calculation by g_bar in 
>>> Gromacs.  (  
>>> http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin/gmx-tutorials/free_energy/index.html)
>>>          I have a question when I use this tutorail. That is I cannot 
>>> use muti-cores in cluster. When I try to use mpirun -np 2 or more, an 
>>> error occured (There is no domain decomposition for 2 nodes that is 
>>> compatible with the given box and a minimum cell size of 6.46036 nm 228 
>>> Change the number of nodes or mdrun option -rdd)
>>>  But when i turn the mpirun off ,eg mpirun -np 1 mdrun -nt 2 ,  
>>> everything is OK.
>>
>>I do not know the basis for why MPI vs. threading would lead to different 
>>behavior.  I use threading by default on a dual-core machine since the tutorial 
>>system is small.
> 
>>>          Since my model is a little big, the speed of use 1 core and 2 
>>> threads is too slow, is there any method to change this situation, or do 
>>> you know any paramater that control this?
>>>         
>>
>>Are we still talking about the tutorial system?  It is quite small.  The DD cell 
>>size you've posted above seems to be from a different system.  I cannot offer 
>>personalized advice for individual systems that I've not worked with.  The best 
>>information is online:
> 
>  >http://www.gromacs.org/Documentation/Errors#There_is_no_domain_decomposition_for_n_nodes_that_is_compatible_with_the_given_box_and_a_minimum_cell_size_of_x_nm
>>
>>-Justin
> 
> Sorry I haven`t express my idea well. I use your tutorial to calculation the free energy of a big system of my own,which is a big system of 24,0000 atoms, so this speed is too slow for me and I am trying to run it on big clusters.
> Thanks any way.

The simple decoupling procedure in the tutorial is likely not suitable for very 
large systems; by changing many degrees of freedom simultaneously in this way, 
convergence is exceptionally hard to achieve.  I suspect the very large size of 
the DD cells is related to the couple-intramol parameter, which for large 
systems results in explicit pair interactions, which occur within the same DD 
cell.  Refer to the manual for a more suitable setting and the reasons why you 
might want to adjust certain parameters.  Good luck.

-Justin

-- 
========================================

Justin A. Lemkul
Ph.D. Candidate
ICTAS Doctoral Scholar
MILES-IGERT Trainee
Department of Biochemistry
Virginia Tech
Blacksburg, VA
jalemkul[at]vt.edu | (540) 231-9080
http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin

========================================