Dear Jason, Ross,
It was really the ntxo parameter missing, so both the rst and nc files were written in ascii not in binary, not only that it solve the problem to use ntxo=2 but also speeded up the run by 15%… learning in the tough way….
Thansk!!
Fabian
Dr. Fabian Glaser
Head of the Structural Bioinformatics section
Bioinformatics Knowledge Unit - BKU
The Lorry I. Lokey Interdisciplinary Center for Life Sciences and Engineering
Technion - Israel Institute of Technology, Haifa 32000, ISRAEL
fglaser at technion dot ac dot il
Tel: +972 4 8293701
http://bku.technion.ac.il
> On 12 Nov 2015, at 7:02 PM, Jason Swails <jason.swails.gmail.com> wrote:
>
> On Thu, Nov 12, 2015 at 11:47 AM, Fabian gmail <fabian.glaser.gmail.com <mailto:fabian.glaser.gmail.com>>
> wrote:
>
>> Hi,
>>
>> I got a weird error when running GPU jobs:
>>
>> 1) From the output
>>
>>
>> |---------------- GPU PEER TO PEER INFO -----------------
>> |
>> | Peer to Peer support: ENABLED
>> |
>> |--------------------------------------------------------
>>
>> | ERROR: I could not understand line 49846
>> 199.3676234-169.4028942 339.9971292
>>
>
> This is the wrong line. I have to figure out where the problem is for this
> check.
>
> Look for ***s in your inpcrd file. I suspect that a coordinate has gotten
> too big (or small) to fit in a %12.7F field. So you either need to set
> iwrap=1 or ntxo=2 to wrap coordinates or write NetCDF restarts.
>
> HTH,
> Jason
>
> --
> Jason M. Swails
> BioMaPS,
> Rutgers University
> Postdoctoral Researcher
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org <mailto:AMBER.ambermd.org>
> http://lists.ambermd.org/mailman/listinfo/amber <http://lists.ambermd.org/mailman/listinfo/amber>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Sun Nov 15 2015 - 01:30:03 PST
