Hi Andreas,
Well that does seem like a volume change that could cause pmemd.cuda
problems, but I agree that you should see the error message if you are
running the command directly from your shell (I think it prints to stderr).
One troubleshooting idea: you could try writing the restart file frequently
(say every 250 or 500 steps). After the calculation stops, just continue
from the most recent restart file (as the error message I attached
indicates). If everything works out after a restart, I'd say it points
toward the volume change issue that pmemd.cuda can't handle. But if it
crashes again, then maybe there is some instability or steric problem in
your system that pmemd.cuda cannot handle. Another difference between pmemd
and pmemd.cuda is that the latter does not implement the vlimit keyword and
so things can blow up a little easier.
Also, what version of Amber are you running?
--Niel
On Thu, Dec 15, 2016 at 11:46 PM, Andreas Tosstorff <
andreas.tosstorff.cup.uni-muenchen.de> wrote:
> Thanks for your message Hai,
>
> I'm running on a cluster, but directly from its bash shell, as we have not
> yet installed a queuing system. Shouldn't I see any error directly in the
> shell?
>
>
>
> -----Original Message-----
> From: Hai Nguyen [mailto:nhai.qn.gmail.com]
> Sent: Friday, December 16, 2016 8:32 AM
> To: AMBER Mailing List <amber.ambermd.org>
> Subject: Re: [AMBER] pmemd.cuda incomplete output file
>
> are you running on a cluster? You might want to check the error log
> specified by your script.
>
> (e.g :
> http://stackoverflow.com/questions/9096959/how-to-
> specify-error-log-file-and
> -output-file-in-qsub
> )
>
> Hai
>
> On Thu, Dec 15, 2016 at 11:15 PM, Andreas Tosstorff <
> andreas.tosstorff.cup.uni-muenchen.de> wrote:
>
> > Dear Niel,
> >
> > Thanks for your reply. For both, pmemd and pmemd.cuda, Volume went
> > from
> > ~730000 to ~640000 in 200 ps during equilibration. Pmemd continues
> > working up to 500 ps as specified, pmemd.cuda does not. No error
> > message to be found. I also tried pmemd.cuda with an additional 50ps
> > NVT equilibration before the NPT eq., but no luck there. I also tried
> > the default barostat instead of Monte Carlo.
> >
> > Is there anything else I could try?
> >
> > This is my GPU:
> >
> > |------------------- GPU DEVICE INFO --------------------
> > |
> > | CUDA_VISIBLE_DEVICES: 1
> > | CUDA Capable Devices Detected: 1
> > | CUDA Device ID in use: 0
> > | CUDA Device Name: GeForce GTX TITAN X
> > | CUDA Device Global Mem Size: 12206 MB
> > | CUDA Device Num Multiprocessors: 24
> > | CUDA Device Core Freq: 1.22 GHz
> > |
> > |--------------------------------------------------------
> >
> > -----Original Message-----
> > From: Niel Henriksen [mailto:shireham.gmail.com]
> > Sent: Friday, December 16, 2016 1:40 AM
> > To: AMBER Mailing List <amber.ambermd.org>
> > Subject: Re: [AMBER] pmemd.cuda incomplete output file
> >
> > On Thu, Dec 15, 2016 at 1:31 PM, Andreas Tosstorff <
> > andreas.tosstorff.cup.uni-muenchen.de> wrote:
> >
> > > Hi all,
> > >
> > > I ran the same job with pmemd instead of pmemd.CUDA and it worked
> > > just fine. Any suggestions how I could make it run on pmemd.CUDA?
> > >
> > >
> > Hi Andreas,
> >
> > Just a guess here, but is there any chance the volume is changing
> > significantly during equilibration? If that is the case, you might
> > see the error message I've copied below. Sometimes I've missed that
> > message if stdout/stderr is redirected in a non-standard way (ie,
> batch/queue system).
> >
> > Best,
> > --Niel
> >
> > ERROR: Calculation halted. Periodic box dimensions have changed too
> > much
> > > from their initial values.
> > > Your system density has likely changed by a large amount, probably
> from
> > > starting the simulation from a structure a long way from equilibrium.
> > > [Although this error can also occur if the simulation has blown up
> > > for some reason]
> > > The GPU code does not automatically reorganize grid cells and thus
> you
> > > will need to restart the calculation from the previous restart file.
> > > This will generate new grid cells and allow the calculation to
> > continue.
> > > It may be necessary to repeat this restarting multiple times if
> > > your system
> > > is a long way from an equilibrated density.
> > > Alternatively you can run with the CPU code until the density has
> > > converged
> > > and then switch back to the GPU code.
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> >
> > _______________________________________________
> > AMBER mailing list
> > AMBER.ambermd.org
> > http://lists.ambermd.org/mailman/listinfo/amber
> >
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Dec 16 2016 - 06:00:03 PST
