[gmx-users] Crash of DD with periodic_molecules

Semen yesint3 at ukr.net
Thu Jun 11 11:23:30 CEST 2015


Dear All, 
We are stuck with an error in domain decomposition for graphene layer, which is periodic in XY plain (periodic_molecules=yes). The system runs perfectly on one node, but when we try to use more than 1 MPI process it crashes regardless of what we do: 

Using 2 MPI processes 
Using 4 OpenMP threads per MPI process 
... 
A list of missing interactions: 
          exclusions of 199319 missing   -274 

------------------------------------------------------- 
Program mdrun_mpi, VERSION 5.0.4 
Source code file: /Data/Users/glaville/WORK/TMP/gromacs-5.0.4/src/gromacs/mdlib/domdec_top.c, line: 390 

Software inconsistency error: 
One or more interactions were multiple assigned in the domain decompostion 
For more information and tips for troubleshooting, please check the GROMACS 
website at http://www.gromacs.org/Documentation/Errors 
------------------------------------------------------- 


This only happens with the periodic bonds - all other systems runs with no problems on the same cluster. 
I played with rdd, dd and dds parameters with no result - it gives the same error. 
Is it a bug or we can overcome this somehow? 

Any help is appreciated! 

Sincerely, 
Semen 


More information about the gromacs.org_gmx-users mailing list