[gmx-users] Volume Contraction Issue

[tca1]

I'm currently trying to run a semi-isotropic NPT simulation where one  
dimension of the box changes in size over a fairly wide range  
throughout the course of the multiple successive trajectories run on a  
given node. One issue I keep running into is that after a certain  
point, the box has changed dimensions enough that the initial  
processor decomposition used in the first trajectory no longer works  
for the reduced box size, and GROMACS terminates with an error about  
not being able to use the assigned processors to construct a domain  
decomposition when it tries to create the tpr file for the next phase  
of the simulation.

This is a frustrating error to deal with manually, and it would make  
sense to have the system adapt to its changing size by changing the  
number of processors used. Is there a way to tell GROMACS to handle  
extra processors more gracefully - for example by ignoring some  
available cores in order to obtain a correct decomposition? The error  
messages only suggest ways to change the desired domain size for the  
simulation and it's not clear to me a priori what size would work  
throughout an entire simulation, or if such a decomposition exists  
without changing the number of processors available.

Thanks,
Tom