Re: [AMBER] Error compiling pmemd.cuda.MPI

From: Ross Walker <>
Date: Fri, 13 Apr 2012 12:46:04 -0700

Hi Alfredo,

> Ross, the instalation of the sander, sander.MPI, pmemd, pmemd.MPI and
> pmemd.cuda are all ok.
> The output of "which mpif90" is:
> /usr/bin/mpif90
> The output of "echo $MPI_HOME" is:
> /home/aquevedo/Programas/Amber12/amber12/AmberTools/src/mpich2-
> 1.4.1p1/src/env/mpif90
> should mpif90 be called from the $MPI_HOME?

Yeap, that's your problem. These don't match. When you have multiple MPI
installations on your machine you have to be very careful that the correct
one is used. I am assuming that the CPU MPI installations you have built
with the MPI in /usr/bin which means you will use the mpirun from /usr/bin
to run them. Thus you should probably try to build pmemd.cuda.MPI with this
mpi as well. Otherwise you are going to have to keep switching the paths to
use the correct mpirun in each case. The issue is your system MPI might not
support MPI v2 correctly but we'll find that out when you try to compile.

So I am assuming your mpif90, mpicc scripts etc are in /usr/bin and the
corresponding mpi.h is in /usr/include. In this case you should set

And then rerun the configure script for -cuda -mpi and try building again
(do a make clean as well for good measure).

All the best

|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Adjunct Assistant Professor |
| Dept. of Chemistry and Biochemistry |
| University of California San Diego |
| NVIDIA Fellow |
| | |
| Tel: +1 858 822 0854 | EMail:- |

Note: Electronic Mail is not secure, has no guarantee of delivery, may not
be read every day, and should not be used for urgent or sensitive issues.

AMBER mailing list
Received on Fri Apr 13 2012 - 13:00:03 PDT
Custom Search