From: Gengbin Zheng (gzheng_at_uiuc.edu)
Date: Thu Mar 31 2005 - 17:41:20 CST
I don't really have a answer for this, but what I have seen seemed not
make sense when Charm saw the correct number of processors but NAMD
didn't, but NAMD is a Charm program.
My suspicion though is that do you happen to have multiple
installations of NAMD like a net version?
And is the right namd2 really in your PATH? I suggest instead of
mpirun -np 4 namd2 foo.config
specify a path to namd2 and run like
mpirun -np 4 ./namd2 foo.config
Randy Crawford wrote:
>I've successfully compiled NAMD 2.5 and Charm++ 5.4 for our Itanium2
>Linux ethernet cluster using the Intel 8.0 compilers and MPICH-1.2.5.
>But at runtime, NAMD always reports that it's running with only one MPI
>Info: NAMD 2.5 for Linux-ia64-MPI
>Info: Running on 1 processors.
>I have spawned the MPI job using MPICH's mpirun as well as mpiexec, with
>the same outcome. Oddly, when using mpirun, Charm++ reports that it's
>running the *correct* number of processes, but NAMD reports the
>existence of only one. (Charm++ does not report the number of processes
>when I use mpiexec.) I've even tried generating a 'nodelist' file as
>well, but it makes no difference.
>I launch NAMD from PBS using:
> #PBS -l nodes=2:ppn=2,walltime=0:60:00
> mpiexec -comm mpich-p4 namd2 foo.config
>Or I launch interactively using:
> mpirun -np 4 namd2 foo.config
>Any idea of why this is happening or what else I might try?
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:39:17 CST