[gmx-users] wierd behavior of mdrun

[Mark Abraham]

Paymon Pirzadeh wrote:
> One more point:
> when I run the job interactively, although it works but date and time
> are wrong!
> 
> step 0
> step 100, will finish Wed Sep 16 16:16:00 2009
> step 200, will finish Wed Sep 16 14:32:34 2009
> step 300, will finish Wed Sep 16 13:57:51 2009
> step 400, will finish Wed Sep 16 11:35:46 2009
> step 500, will finish Wed Sep 16 11:50:12 2009
> step 600, will finish Wed Sep 16 10:36:38 2009
> step 700, will finish Wed Sep 16 09:44:04 2009
> step 800, will finish Wed Sep 16 10:07:02 2009
> step 900, will finish Wed Sep 16 09:29:25 2009
> step 1000, will finish Wed Sep 16 08:59:19 2009
> step 1100, will finish Wed Sep 16 09:20:06 2009
> step 1200, will finish Wed Sep 16 08:55:47 2009
> step 1300, will finish Wed Sep 16 09:13:38 2009
> step 1400, will finish Wed Sep 16 08:53:15 2009
> 
> Is it a cluster related or GROMACS related issue? Is it important at
> all?

That's just an estimate based on the elapsed wall time. If the source of 
elapsed wall time is unreliable or wrong, or the sample of wall time is 
too short, then the estimate will be wrong or vary widely accordingly.

This prediction should stabilize over the course of the run.

Mark