[gmx-users] angle_restraints_z and -ddcheck

[chris.neale at utoronto.ca]

I have determined that the proper solution here is to use particle  
decomposition. -noddcheck simply delays the inevetable fatal LINCS  
error. Nevertheless, mdrun -pd requires that one maintains whole  
molecules in the starting conformation:  
http://lists.gromacs.org/pipermail/gmx-users/2010-January/047802.html

My particle decomposition simulations with angle_restratins_z have now  
completed > 10 ns * hundreds of systems with no crashes or warnings.

Chris.

-- original message --

I have just started working with the [ angle_restraints_z ] portion of
the code. My initial tests indicate that it works in my system.
However, in some cases I am getting a fatal error that a pair
interaction is too far apart to be calculated within domain
decomposition. Although the error message refers to a bonded
interaction,  in actuality this atomic pair is the two atoms that
define my vector whose angle to the z-axis is restrained. I guess that
the same code that requires "bonded interactions have been properly
assigned to the domain decomposition cells" also requires this of the
vector termini for angle_restraints_z.

I suspect that my solution is to use -noddcheck:

"For pair interactions and tabulated bonds that do not generate
exclusions, this check can be turned off with the option -noddcheck."

although it seems zealous to turn off a useful global error detector
in order to avoid a crash based on the spurious detection of a known
group of only two atoms.

I'm interested first to see if anybody else has encountered this,
second to ask if applying -noddcheck seems like a good idea, and third
to suggest that angle_restraints_z not spawn this fatal error --
unless it is indeed required (in which case I should not be turning
off my ddcheck).

error message follows, wherein atoms 138 and 293 are those that define
my angle_restraints_z, which looks like this:

[ angle_restraints_z ]
138 293 1 40 500 1

Thank you,
Chris.

###############

Error messages from the .log file:

Not all bonded interactions have been properly assigned to the domain
decomposition cells

A list of missing interactions:
         Angle Rest. Z of      1 missing      1

Molecule type 'Protein'
the first 10 missing interactions, except for exclusions:
         Angle Rest. Z atoms  138  293           global   138   293

-------------------------------------------------------
Program mdrun_openmpi, VERSION 4.0.5
Source code file: domdec_top.c, line: 341

Fatal error:
1 of the 110951 bonded interactions could not be calculated because
some atoms involved moved further apart than the multi-body cut-off
distance (1 nm) or the two-body cut-off distance (1 nm), see option
-rdd, for pairs and tabulated bonds also see option -ddcheck
-------------------------------------------------------