[gmx-users] Regarding segmentation fault in GPU_gromacs 4.6.1

[Justin Lemkul]

On 7/17/14, 1:51 AM, DeBaJiT DeY wrote:
> Hello,
>
> I am trying to simulate a peptide-membrane system on GPU using gromacs
> 4.6.1 . However whenever I invoke the mdrun command I get an error as *srun:
> error: coe3: task 0: Segmentation fault (core dumped)*. Can anybody please
> suggest me what might be the problem?
>
> *Script file used to submit the job*:
> #!/bin/bash
> #SBATCH --job-name=GRMCS-GPU
> #SBATCH --nodes=1
> #SBATCH --gres=gpu:1
> #SBATCH --nodelist=coe3
> #SBATCH -e slurm.err
> #SBATCH -v
>
> #cd $SLURM_SUBMIT_DIR
>
> echo "CWD:" $SLURM_SUBMIT_DIR
> echo "NODELIST:" $SLURM_JOB_NODELIST
>
> export PATH=/opt/cuda-5.0/bin:$PATH
> export
> LD_LIBRARY_PATH=/opt/cuda-5.0/lib64:/opt/cuda-5.0/lib:$LD_LIBRARY_PATH
>
> srun -N1 -s /opt/apps/gromacs_GPU/bin/mdrun_mpi -deffnm EM
>
> *.mdp file for energy minimization*:
> title        = EM of B2AR-POPC system    ; Title of the miniization run
> ; Parameters describing the details of the Energy Minimization protocol
> define          =         ; Define any restrain ("-DPOSRES" OR
> "-DSTRONG_POSRES" OR leave blank if there is no restrain)
> integrator    = steep        ; EM Algorithm (steep = steepest descent; cg =
> conjugate gradiant; l-bfgs = Low-memory Broyden-Fletcher-Goldfarb-Shanno
> quasi-Newtonian minimizer)
> emtol        = 100.0      ; Minimization is stopped when the maximum force
> on an atom is less than the given value (kJ/mol/nm)
> emstep          = 0.01          ; Initial step size (nm)
> nsteps        = 20000          ; Maximum number of (energy minimization)
> steps to be performed
>
> ; Parameters describing neighbors searching and details about interaction
> calculations
> nstlist        = 1        ; Neighbor list update frequency (after every
> given number of steps)
> ns_type        = grid        ; Neighbor list search method (simple, grid)
> rlist        = 1.2        ; Neighbor list search cut-off (nm)
> coulombtype    = PME        ; Long range electrostatic interactions
> treatment (cut-off, Ewald, PME)
> rcoulomb    = 1.2        ; Short-range electrostatic cut-off (nm)
> rvdw        = 1.2        ; Short-range van der Waals cut-off (nm)
> pbc        = xyz         ; Direction in which to use Perodic Boundary
> Conditions (xyz, xy, no)
>

Hard to say without any actual information from Gromacs.  Can you tell us:

1. Does the EM work on CPU?
2. Do other systems run on the GPU?
3. Can you post your .log file?

-Justin

-- 
==================================================

Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 601
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441
http://mackerell.umaryland.edu/~jalemkul

==================================================