Homec4science
Diffusion lammps (efficient_neuronet) Browse

lammps/doc/doc2/timer.html
c2352b946fdcefficient_neuronet

Last Change
Raw File

timer.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>timer command
</H3>
<P><B>Syntax:</B>
</P>
<PRE>timer args
</PRE>
<LI><I>args</I> = one or more of <I>off</I> or <I>loop</I> or <I>normal</I> or <I>full</I> or <I>sync</I> or <I>nosync</I>
<PRE> <I>off</I> = do not collect and print timing information
<I>loop</I> = collect only the total time for the simulation loop
<I>normal</I> = collect timer information broken down in sections (default)
<I>full</I> = like <I>normal</I> but also include CPU and thread utilzation
<I>sync</I> = explicitly synchronize MPI tasks between sections
<I>nosync</I> = do not synchronize MPI tasks when collecting timer info (default)
</PRE>
<P><B>Examples:</B>
</P>
<PRE>timer full sync
timer loop
</PRE>
<P><B>Description:</B>
</P>
<P>Select to which level of detail LAMMPS is performing internal profiling.
</P>
<P>During regular runs LAMMPS will collect information about how much
time is spent in different sections of the code and thus can provide
valuable information for determining performance and load imbalance
problems. This can be done at different levels of detail and
accuracy. For more information about the timing output, please have a
look at the <A HREF = "Section_start.html#start_8">discussion of screen output</A>.
</P>
<P>The <I>off</I> setting will turn all time measurements off. The <I>loop</I>
setting will only measure the total time of run loop and not collect
any detailed per section information. With the <I>normal</I> setting,
timing information for individual sections of the code are collected
and also information about load imbalances inside those sections
presented. The <I>full</I> setting adds information about CPU utilization
and thread utilization, when multi-threading is enabled.
</P>
<P>With the <I>sync</I> setting, all MPI tasks are synchronized at each timer
call and thus allowing to study load imbalance more accuractly, but
this usually has some performance impact. Using the <I>nosync</I> setting
this can be turned off (which is the default).
</P>
<P>Multiple keywords can be provided and for keywords that are mutually
exclusive, the last one in that group is taking effect.
</P>
<P>IMPORTANT NOTE: Using the <I>full</I> and <I>sync</I> options provides the most
detailed and accurate timing information, but also can have a
significant negative performance impact due to the overhead of the
many required system calls. It is thus recommended to use these
settings only when making tests to identify the performance. For
calculations with few atoms or a very large number of performance,
even using the <I>normal</I> setting can have a measurable performance
impact. In those cases just use the <I>loop</I> or <I>off</I> setting.
</P>
<P><B>Restrictions:</B> none
</P>
<P><B>Related commands:</B>
<A HREF = "run.html">run post no</A>, <A HREF = "kspace_modify.html">kspace_modify fftbench</A>
</P>
<P><B>Default:</B>
</P>
<P>timer normal nosync
</P>
</HTML>
c4science · Help