Homec4science
Diffusion lammps (lammps-icms) Browse

lammps/doc/loadbalance.html
8eb36be09280lammps-icms

Last Change
Raw File

loadbalance.html
View Options

<HTML>
<CENTER><A HREF = "http://lammps.sandia.gov">LAMMPS WWW Site</A> - <A HREF = "Manual.html">LAMMPS Documentation</A> - <A HREF = "Section_commands.html#comm">LAMMPS Commands</A>
</CENTER>
<HR>
<H3>loadbalance command
</H3>
<P><B>Syntax:</B>
</P>
<PRE>loadbalance style keywords
</PRE>
<UL><LI>style = load balancing style
<LI>keywords = load balancing style specific flags
</UL>
<P><B>Examples:</B>
</P>
<PRE>loadbalance nlocal/simple ntry 4
loadbalance off
</PRE>
<P><B>Description:</B>
</P>
<P>Automatically adjust runtime parameters to try balancing
the load between multiple MPI processors for improved
overall performance. At the moment on a simple particle
bases load balancing strategy is implemented.
</P>
<P>The <I>style</I> argument selects the algorithm and strategy
that is used of automatic load balancing. The style <I>off</I>
disables load balancing.
</P>
<P>The <I>nlocal/simple</I> style adjusts dimensions of individual
sub-domains during domain decomposition to have an equal
number of (local) atoms. This load balancing strategy
assumes that the computational effort on each processor
will be approximately proportional to the number of atoms.
The required <I>ntry</I> flag specifies how many iterations
to redistribute the atoms are attempted. The higher this number,
the faster the load balancer can react to changes in the
particle distribution, but also the more overhead through
communication is caused.
</P>
<HR>
<P><B>Restrictions:</B>
</P>
<P>The <I>nlocal/simple</I> load balancing style cannot be used for
systems with triclinic boxes or with calculations using the
r-RESPA run style.
</P>
<P><B>Related commands:</B>
</P>
<P><A HREF = "processors.html">processors</A>, <A HREF = "run_style.html">run_style</A>
</P>
<P><B>Default:</B> none
</P>
</HTML>
c4science · Help