[gmx-users] Parallel Timings

Velocet math at velocet.ca
Wed Dec 12 17:57:44 CET 2001


On Wed, Dec 12, 2001 at 04:06:20PM +0100, K.A.Feenstra's all...
> "Jithesh P.V." wrote:
> > 
> > Hello,
> > 
> > The logfile from mdrun on multiprocessors reports the
> > various timings...Could you please tell me what these
> > values indicate? I tried to locate the explanation in
> > the manual, but could not.
> > 
> >                NODE (s)   Real (s)      (%)
> >        Time:    508.000    508.000    100.0
> >                        8:28
> >         (Mnbf/s)  (MFlops) (ps/NODE hour)
> > (NODEhour/ns)
> > Performance:9.330  293.136   70.866       14.111
> 
> I thought these were pretty obvious....
> 
> Time:
> cpu-time (derived from number of CPU-cycles used) in seconds
> wall-time (as your watch or wall-clock indicates) in seconds
> ratio of cpu vs wall time in percent
> 
> Performance:
> number of Mega non-bonded forces per second
> number of Mega floating point operation per second
> picosecond of simulation time per hour
> hours per nanosecond of simulation time

The comparative values presented on the parallel benchmarks on the webpage
are in ps per computation-day AFAICT. Took a while to figure that out, but
I think it does say that somewhere on the page.

/kc

> 
> -- 
> Groetjes,
> 
> Anton
>  _____________ _____________________________________________________
> |             | Anton Feenstra       Dept. of Biophysical Chemistry |
> |  _   _  ___,| K.A.Feenstra at chem.rug.nl    University of Groningen |
> | / \ / \'| | | Tel +31 50 363 4327                    Nijenborgh 4 |
> |(   |   )| | | Fax +31 50 363 4800               9747 AG Groningen |
> | \_/ \_/ | | | http://md.chem.rug.nl/~anton        The Netherlands |
> |             | "Make the Floor Burn" (2 Unlimited)                 |
> |_____________|_____________________________________________________|
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.

-- 
Ken Chase, math at velocet.ca  *  Velocet Communications Inc.  *  Toronto, CANADA 



More information about the gromacs.org_gmx-users mailing list