[gmx-users] Re: reasons for slow computation?

Yun Shi yunshi09 at gmail.com
Mon Oct 10 22:22:03 CEST 2011


And another difference I noticed from .log files are:
..............
Initial maximum inter charge-group distances:
    two-body bonded interactions: 0.451 nm, LJ-14, atoms 3586 4808
  multi-body bonded interactions: 0.451 nm, Proper Dih., atoms 3586 4808
Minimum cell size due to bonded interactions: 0.496 nm
................
Optimizing the DD grid for 48 cells with a minimum initial size of 0.620 nm
The maximum allowed number of cells is: X 12 Y 12 Z 12
Domain decomposition grid 8 x 3 x 2, separate PME nodes 24
PME domain decomposition: 8 x 3 x 1
Interleaving PP and PME nodes
This is a particle-particle only node

Domain decomposition nodeid 0, coordinates 0 0 0

Using two step summing over 21 groups of on average 2.3 processes
...................

for the slow one, while for the fast one:
................
Initial maximum inter charge-group distances:
    two-body bonded interactions: 0.450 nm, LJ-14, atoms 2064 3014
  multi-body bonded interactions: 0.450 nm, Proper Dih., atoms 2064 3014
Minimum cell size due to bonded interactions: 0.494 nm
................

Optimizing the DD grid for 48 cells with a minimum initial size of 0.618 nm
The maximum allowed number of cells is: X 12 Y 12 Z 12
Domain decomposition grid 8 x 3 x 2, separate PME nodes 24
PME domain decomposition: 8 x 3 x 1
Interleaving PP and PME nodes
This is a particle-particle only node

Domain decomposition nodeid 0, coordinates 0 0 0

Using two step summing over 18 groups of on average 2.7 processes
................


Are these differences likely to generate huge performance difference?

Thanks,
Yun
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20111010/bede3563/attachment.html>


More information about the gromacs.org_gmx-users mailing list