> > Ultimately you should be looking at the total time reported
> at the end of
> > the output file. This is actually Wall Clock Time (NOT CPU
> TIME) so will
> > give you a real understanding of 'time to solution' on a
> specific machine
> > setup.
>
> If you share a single CPU with other users, particularly if
> your time is regulated, I would argue that CPU time is more
> useful. In that case, wallclock would vary according to other
> jobs on the machine. Wallclock is definitely what you need
> for a cluster in which you have nodes to yourself.
Just to clarify what I meant by this statement. Often when a program reports
cputime used it reports the time actually used by the process it does not
include operating system overhead associated with things such as handling
TCP/IP headers etc. The actual cpu usage by a simulation is actually the
process cputime + all the overhead. Most large computer clusters bill on
walltime anyway so this really is the important value.
Really the thing to consider is really wallclocktime x cpus = actual
wallclock usage per cpu. (this is what you would be billed for by a
supercomputer center etc.)
All the best
Ross
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Wed Aug 31 2005 - 22:53:01 PDT
