Re: [AMBER] premed.MPI warning

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 12 Jan 2016 11:54:58 -0500

On Tue, Jan 12, 2016 at 11:34 AM, Irfan Alibay <
irfan.alibay.manchester.ac.uk> wrote:

> Hi,
>
> Similarly to the results posted by George earlier, the "Final Performance
> Info" was printed in the mdinfo (see below) and the mdout's (see below)
> version of the "Final Performance Info" attempts to do an average over 0
> steps (which I assume is because last_step_count gets set to current_step
> when it is called in runfiles before the final call?).
>
> To see what would happen I added a test to prevent an mdinfo flush when
> there is equality between nstep & total_nstlim at the same time as the
> mdinfo_flush_interval check in runfiles and that appears to fix things.
>
> Many thanks,
>
> Irfan Alibay
> PhD Student
> Pharmacy School
> The University of Manchester
>
> mdinfo contents:
>
> NSTEP = 500 TIME(PS) = 20501.000 TEMP(K) = 292.47 PRESS =
> 339.7
> Etot = -13299.1721 EKtot = 3035.0159 EPtot =
> -16334.1880
> BOND = 11.1472 ANGLE = 33.8218 DIHED =
> 44.1738
> 1-4 NB = 13.0001 1-4 EEL = 327.4698 VDWAALS =
> 2442.2472
> EELEC = -19206.0478 EHBOND = 0.0000 RESTRAINT =
> 0.0000
> EKCMT = 1526.3358 VIRIAL = 1143.0184 VOLUME =
> 52257.6272
> Density =
> 0.9954
> Ewald error estimate: 0.1046E-03
>
> | Final Performance Info:
> | -----------------------------------------------------
> | Average timings for last 10 steps:
> | Elapsed(s) = 0.18 Per Step(ms) = 18.22
> | ns/day = 9.48 seconds/ns = 9110.35
> |
> | Average timings for all steps:
> | Elapsed(s) = 9.01 Per Step(ms) = 18.03
> | ns/day = 9.58 seconds/ns = 9014.59
> | -----------------------------------------------------
>
>
> mdout final performance:
>
> | Final Performance Info:
> | -----------------------------------------------------
> | Average timings for last 0 steps:
> | Elapsed(s) = 0.00 Per Step(ms) = Infinity
> | ns/day = 0.00 seconds/ns = Infinity
>

​Looks like a divide-by-zero to me :).

I'll go ahead and take care of this. It is clearly benign and will not
affect results.

Thanks for the report and help hunting this down,
Jason

-- 
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Jan 12 2016 - 09:00:04 PST
Custom Search