Re: [AMBER] ERROR IN SETPAR() upon atom distribution with FF03, 03.r1

From: case <>
Date: Thu, 17 Feb 2011 16:15:33 -0500

On Thu, Feb 17, 2011, Bongkeun Kim wrote:

> And I attached all prmtop, inpcrd, rst and input mdin files with zip.
> You can execute it by the following command.
> sander.MPI -ng 12 -groupfile equilibrate.groupfile

Works for me, if I precede the above command with "mpirun -np 12" or with
"mpirun -np 24". So, you might check what the "ibrun" command in your
"remd.job" script is really doing: how many threads is it actually using?
Note that the number of threads must be a multiple of the number of groups
(in this case, 12).

Sander has an additional requirement that the number of residues must be
greater than the number of threads used for member of the group. If you
are indeed using 192 nodes, you would have 16 threads per group member
(192/12) and an error would occur (since you have only 16 residues). This
is supposed to be checked, but the check may be in error, or the error
message could be getting lost in the multisander run. A quick check is to
see if the job works on 96 threads (or, indeed, any multiple of 12 smaller
than 16.)


AMBER mailing list
Received on Thu Feb 17 2011 - 13:30:02 PST
Custom Search