[gmx-users] Free energy, domain decomposition and gmx 465 v 505
Gmx QA
gmxquestions at gmail.com
Thu Jun 25 09:21:33 CEST 2015
Hi all
I am running some free energy tutorial calculations to learn more about how
those are set up in gromacs, but have run into a problem when using
different versions.
I am following the free energy tutorial on the Gromacs website
(fe-tutorial-4.6), so not Justin's tutorial.
With vmx 4.6.5 everything goes according to the tutorial, and I can run
with four thread on my laptop.
Please find what I think is the relevant section of the log file below:
$/usr/local/gmx465/bin/mdrun -v -deffnm topol_465
Initializing Domain Decomposition on 4 nodes
Dynamic load balancing: auto
Will sort the charge groups at every domain (re)decomposition
Initial maximum inter charge-group distances:
two-body bonded interactions: 0.239 nm, LJC-14 q, atoms 4 7
multi-body bonded interactions: 0.239 nm, Angle, atoms 2 6
Minimum cell size due to bonded interactions: 0.263 nm
Maximum distance for 5 constraints, at 120 deg. angles, all-trans: 0.497 nm
Estimated maximum distance required for P-LINCS: 0.497 nm
This distance will limit the DD cell size, you can override this with -rcon
Using 0 separate PME nodes, as there are too few total
nodes for efficient splitting
Scaling the initial minimum size with 1/0.8 (option -dds) = 1.25
Optimizing the DD grid for 4 cells with a minimum initial size of 0.622 nm
The maximum allowed number of cells is: X 3 Y 3 Z 2
Domain decomposition grid 2 x 2 x 1, separate PME nodes 0
PME domain decomposition: 2 x 2 x 1
Domain decomposition nodeid 0, coordinates 0 0 0
Using 4 MPI threads
While on the other hand, when I am using v 5.0.5 I cannot start simulations
using more than one thread, otherwise the domain decomposition complains,
even though the excerpt from the log file from that run is (almost)
identical
$/usr/local/gmx505/bin/mdrun -v -deffnm topol_505
Initializing Domain Decomposition on 4 ranks
Dynamic load balancing: auto
Will sort the charge groups at every domain (re)decomposition
Initial maximum inter charge-group distances:
two-body bonded interactions: 0.239 nm, LJC-14 q, atoms 4 7
multi-body bonded interactions: 0.239 nm, Angle, atoms 2 6
Minimum cell size due to bonded interactions: 0.263 nm
Maximum distance for 5 constraints, at 120 deg. angles, all-trans: 0.497 nm
Estimated maximum distance required for P-LINCS: 0.497 nm
This distance will limit the DD cell size, you can override this with -rcon
Using 0 separate PME ranks, as there are too few total
ranks for efficient splitting
Scaling the initial minimum size with 1/0.8 (option -dds) = 1.25
Optimizing the DD grid for 4 cells with a minimum initial size of 0.622 nm
The maximum allowed number of cells is: X 3 Y 3 Z 2
-------------------------------------------------------
Program gmx, VERSION 5.0.5
Source code file:
/Users/pk/source/gromacs-5.0.5/src/gromacs/mdlib/domdec.c, line: 6902
Fatal error:
There is no domain decomposition for 4 ranks that is compatible with the
given box and a minimum cell size of 0.62175 nm
Change the number of ranks or mdrun option -rcon or -dds or your LINCS
settings
Look in the log file for details on the domain decomposition
For more information and tips for troubleshooting, please check the GROMACS
website at http://www.gromacs.org/Documentation/Errors
-------------------------------------------------------
Is there something that I have overlooked in v 5 that needs to be taken
into consideration? I am not that familiar with the new parallelization
scheme of v5, maybe there is something there?
Thanks
/PK
More information about the gromacs.org_gmx-users
mailing list