Re: [AMBER] Setting PME cutoff in Crystal Simulations

From: Matthew Guberman-Pfeffer <>
Date: Tue, 5 Jan 2021 22:03:59 -0500

Hi Dr. Case,

I’ve reproduce the error message, which only appears if I try to run on GPUs. The mdout file quits with:

 Number of triangulated 3-point waters found: 7200

     Sum of charges from parm topology file = 0.00001521
     Forcing neutrality...

And the slurm log file (I’m running on an HPC) states:

/var/spool/slurmd/job3989454/slurm_script: line 0: unalias: easy_install-3.6: not found
/var/spool/slurmd/job3989454/slurm_script: line 0: unalias: pip3.6: not found
/var/spool/slurmd/job3989454/slurm_script: line 0: unalias: pip3: not found
/var/spool/slurmd/job3989454/slurm_script: line 0: unalias: python3.6: not found
/var/spool/slurmd/job3989454/slurm_script: line 0: unalias: python3.6m: not found
gpu_neighbor_list_setup :: Small box detected, with <= 2 cells in one or more
                           dimensions. The current GPU code has been deemed
                           unsafe for these situations. Please alter the
                           cutoff to increase the number of hash cells, make
                           use of the CPU code, or (if absolutely necessary)
                           run pmemd.cuda with the -AllowSmallBox flag. This
                           behavior will be corrected in a forthcoming patch.

Can you please explain a bit more about what this error message means? How would you recommend proceeding?

AMBER mailing list
Received on Tue Jan 05 2021 - 19:30:02 PST
Custom Search