[gmx-users] particle decomposition requires preliminary trjconv -pbc mol input

[chris.neale at utoronto.ca]

Thanks Berk,

increasing -rdd from 1.0 to 2.5 has solved my problem. Note that the  
actual distance between the two atoms in the angle restraint never  
exceeds 0.9 nm, although this is a rather complex set of virtual atoms  
based on the positions of other virtual atoms, each virtual atom in  
its own charge group, so there could be other distances involved here.  
In any event, increasing -rdd to 2.5 is fine for this system on 8  
cores. This would inhibit me from going too much beyond 8 cores, but I  
had not planned to do that in any event. It is also quite possible  
that some value of -rdd that is greater than 1.0 and less than 2.5  
would also work. I simply noted that my speed did not decrease by more  
than 5% when going to -rdd 2.5 and so I accepted this.

This is great as the -pd option reduced my speed on 8 cores to 65% of  
what I can get with -dd.

-- original message --

Hi,

Have you actually tried increasing -rdd to the distance the maximum distance
that can occur between the two atoms in the angle restraint?

Berk

> Date: Sun, 3 Jan 2010 15:53:36 -0500
> From: chris.neale at utoronto.ca
> To: gmx-users at gromacs.org
> Subject: [gmx-users] particle decomposition requires preliminary  
> trjconv	-pbc mol input
>
> Hello,
>
> I recently used mdrun -pd while attempting to solve my problems with  
>  angle_restraints_z   
> (http://lists.gromacs.org/pipermail/gmx-users/2010-January/047785.html)  
> since  I discovered that -noddcheck is not a valid solution to the  
> missing  interactions problems with angle restraints.
>
> I received immediate errors with particle decomposition (mdrun -pd),  
>  although no errors with domain decomposition (mdrun -dd).
>
> After a couple of days of changing parameters, I discovered that  
> mdrun  -pd requires an input .gro file in which all molecules are  
> whole (as  the used to be written out by gromacs mdrun version 3).
>
> My basic problem was that I had run parallel EM using domain   
> decomposition and then since my molecules were broken, parallel MD   
> using particle decomposition failed.
>
> The solution:
> grompp -f one.gro -o my.tpr
> trjconv -s my.tpr -pbc mol -o two.gro
> grompp -f two.gro -o my.tpr
>
> I'm posting so that this can be found by a mailing list search for  
> the  next time somebody runs into this as it is, as far as I can  
> tell, an  undocumented requirement.
>
> Chris.
>
> -- 
> gmx-users mailing list    gmx-users at gromacs.org
> http://lists.gromacs.org/mailman/listinfo/gmx-users
> Please search the archive at http://www.gromacs.org/search before posting!
> Please don't post (un)subscribe requests to the list. Use the www  
> interface or send it to gmx-users-request at gromacs.org.
> Can't post? Read http://www.gromacs.org/mailing_lists/users.php

_________________________________________________________________
New Windows 7: Simplify what you do everyday. Find the right PC for you.
http://windows.microsoft.com/shop
-------------- next part --------------
An HTML attachment was scrubbed...
URL:  
http://lists.gromacs.org/pipermail/gmx-users/attachments/20100108/2bc72ff5/attachment.html