Re: [AMBER] GPU MPI failed

From: David A Case <david.case.rutgers.edu>
Date: Thu, 11 Aug 2016 08:15:16 -0400

On Thu, Aug 11, 2016, Albert wrote:
>
> I try to run amber GPU MPI job with command line:
>
> mpirun -np 2 pmemd.cuda.MPI -O -i step6.0_minimization.mdin -p
> step5_charmm2amber.parm7 -c step5_charmm2amber.rst7 -o test.mdout -r
> test.rst7 -inf test.mdinfo
>
> however, it failed with messages:
>
> application called MPI_Abort(MPI_COMM_WORLD, 1) - process 0

1. Please look in the test.mdout file to see if errors are reported there;
2. Please check that you job runs OK on a single GPU, and in parallel
     using pmemd.MPI. Note that (generally) minimizations should be
     done on a CPU, since dynamic range limitations in the cuda code
     can cause problems if there are bad contacts in the starting structure.
3. Please confirm that the pmemd.cuda.MPI test cases are passing. (This will
     help in understanding if the problem is with your pmemd.cuda.MPI
     installation, or with the particular problem you are using.)

...dac


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Thu Aug 11 2016 - 05:30:04 PDT
Custom Search