Re: [AMBER] Question about mdinfo file.

From: Ross Walker <ross.rosswalker.co.uk>
Date: Tue, 17 Aug 2010 11:56:31 -0700

> | Current Timing Info
> | -------------------
> | Total steps : 100000 | Completed : 1000 | Remaining :
> 99000
> |
> | Average timings for last 1000 steps:
> | Elapsed(s) = 2.2 Per Step(ms) = 2.2
> | ns/day = 79.3 seconds/ns = 1090.1
> |
> | Average timings for all steps:
> | Elapsed(s) = 2.2 Per Step(ms) = 2.2
> | ns/day = 79.3 seconds/ns = 1090.1
> |
> |
> | Estimated time remaining: 3.6 minutes.
> ----------------------------------------------------------------------
> --------
> Does this mean that the simulation was < 60 secs. If yes, why does it
> have
> steps remaining. Sorry if I am missing something obvious.
>
> But if I increase the number of steps to 1000000 and hence, simulation
> time,
> I get ns/day in compliance with the results on website.

Do you have a mdout file to go with this? The mdinfo gets flushed
approximately every 60 seconds although it may get flushed sooner than that
depending on your system. I'd have to check the code carefully to be sure.
What you are showing here looks the mdinfo file within the first 60 seconds
but the job is still running no? What does the mdinfo file look like once
the calculation has actually finished. Clearly here it has flushed the
mdinfo file after just 2.2 seconds although I suspect this is because it
spent a lot of time in setup. Short answer is the timings after just 2.2s of
elapsed time will not be enough for reliable statistics. Hence let the
calculation run to the end, do the 100,000 steps you requested and then see
what the mdinfo file says. You should then have much more reliable
statistics in the mdinfo file.

You should probably always a couple of minutes into a run before looking at
the mdinfo file for timing data anyway. Maybe I'll also update the code to
write the same timing data to mdout at the end of the run since it could be
useful.

All the best
Ross


/\
\/
|\oss Walker

---------------------------------------------------------
| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| http://www.rosswalker.co.uk | http://www.wmd-lab.org/ |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
---------------------------------------------------------

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.






_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Aug 17 2010 - 12:00:05 PDT
Custom Search