Re: [AMBER] Regarding the MPI error in parallel running of 3D-RISM

From: George M. Giambasu <>
Date: Thu, 28 Jun 2018 10:03:40 -0400

One reason why this might fail is due to the choice of grid sizes. Here
is what the manual says (Amber18.pdf, pg 101, middle of the page).

"For fixed box sizes, the number of grids points in each dimension must
be divisible by two (a general requirement) and the number of grid
points in the z-axis must be divisible by the number of processes."

The number of grid points along z in your calculation is 90 which is not
divisible by 24, which is the number of mpi threads you require in the
batch script. Try using --buffer option instead of specifying the box
size: this will adapt the box size to the above mentioned requirements
and to the number of available threads on your cluster. (Or, of course
,change the solvation box size & grid size manually).



Dr. George M. Giambasu, Ph.D.
Rutgers, The State University of New Jersey
174 Frelinghuysen Road, Room 308E Piscataway,
NJ 08854, USA
Office: 1-848-445-5223
Cell: 1-612-644-8285
AMBER mailing list
Received on Thu Jun 28 2018 - 07:30:02 PDT
Custom Search