[CPMD-list] CPMD parallel scalability
Alexandr Isayev
alex at ccmsi.us
Mon Apr 14 16:31:15 CET 2008
Dear Maurice,
In addition to other suggestions...
I'm not sure about Altix 4700, however on our Altix 3700
there is a "dplace" utility from SGI ProPack that can possibly
improve scalability (assuming no hardware/software problems :)
of your job under heavily loaded machine.
it should be run something like:
mpirun -np $NCPUS dplace -c1-$NCPUS -s1 <your app is here>
Greetings from Jackson,
Alexandr
Monday, April 14, 2008, 8:57:43 AM, you wrote:
MdK> Hi all,
MdK> I´m running CPMD on an Altix 4700 system with 44 CPU´s and 88 Gb of RAM
MdK> memory.
MdK> At the moment I´m running a CP MD run of a cell containing 96 water
MdK> molecules using the
MdK> BLYP functional at 300 K. I noticed that the scalability is not very
MdK> good. If I run on more than
MdK> about 16 CPU´s, the time per MD step starts even increases, such that
MdK> the total time starts growing with the
MdK> number of CPU´s. Is there anything I can do about this?
MdK> Below is a part of the input script
MdK> Cheers,
MdK> Maurice
MdK> &CPMD
MdK> MOLECULAR DYNAMICS CP
MdK> RESTART WAVEFUNCTION COORDINATES VELOCITIES
MdK> QUENCH BO ELECTRONS
MdK> TIMESTEP
MdK> 4
MdK> TEMPCONTROL IONS
MdK> 300.0 40
MdK> MAXSTEP
MdK> 20000
MdK> TRAJECTORY XYZ
MdK> EMASS
MdK> 600
MdK> STORE
MdK> 100
MdK> &END
MdK> &SYSTEM
MdK> SYMMETRY
MdK> 8
MdK> SCALE
MdK> ANGSTROM
MdK> CELL ABSOLUTE
MdK> 13.1493536801862057 15.1976493759670106 14.3247149284289197 0 0 0
MdK> CUTOFF
MdK> 70
MdK> &END
MdK> &DFT
MdK> FUNCTIONAL BLYP
MdK> &END
MdK> _______________________________________________
MdK> CPMD-list mailing list
MdK> CPMD-list at cpmd.org
MdK> http://cpmd.org/mailman/listinfo/cpmd-list
-------------------------------------------------------
Alexandr Isayev,
Graduate Research Assistant, and System Administrator
@ Computational Center for Molecular Structure
and Interactions (CCMSI),
Jackson State University,
Jackson, MS USA
Tel: +(601) 979-1134
e-mail: alex(at)ccmsi.us
Web: http://www.ccmsi.us/alex
--------------------------------------------------------
More information about the CPMD-list
mailing list