Hi Ross,
Thank you for your response. We finally got to the bottom of this issue. The segmentation fault appears to be due to differences in job scheduler uses srun (SLURM) and mpirun.
To answer your question on minimization, we are running them on cpus.
Thanks you.
Best,
AJ
----- Original Message -----
From: "Ross Walker" <ross.rosswalker.co.uk>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Tuesday, April 21, 2015 4:06:21 PM
Subject: Re: [AMBER] Amber GPU Segmentation fault
Hi AJ,
Are you trying to run the minimization on multiple GPUs? - This is not supported and should quit with an error message.
All the best
Ross
> On Apr 21, 2015, at 3:29 PM, AJ Venkatakrishnan <ajvenkat.stanford.edu> wrote:
>
> Hi,
>
> We have been running membrane protein simulation on GPUs. These simulations were set up by following the tutorial by Benjamin Madej and Ross Walker and adapting the input files therein: http://ambermd.org/tutorials/advanced/tutorial16/
>
> We have had several simulations running successfully on GPUs by following the instructions in the above-mentioned link. Lately, however, for systems that ran perfectly previously, we are getting segmentation fault errors. The fault occurs right after minimization, before heating with a message like this:
>
> The following have been reloaded with a version change:
> 1) amber/14-intel => amber/14-cuda
>
> srun: error: gpu-9-8: task 0: Segmentation fault
>
> The problem seems to only impact pmemd.cuda.MPI processes, but both pmemd.cuda and pmemd.MPI work perfectly fine.
>
> I would greatly appreciate any help on resolving the segmentation fault issues.
>
> Thank you for your time.
>
> Best,
> AJ
> _______________________________________________
> 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 Apr 24 2015 - 19:30:02 PDT
