[gmx-users] PME, MPI and bproc

[Jason DeJoannis]

Dear Gromacs Users,

  We get a "SIGSEGV" crash on each node when
using PME in parallel. This occurs on a 
Beowulf linux cluster which uses "bproc" and
"bpsh" to issue remote processes. The crash
is quite general, i.e. it will even happen
with a box of water at several values of
the mesh spacing and interpolation order.

  We have searched the gromacs archives and 
suspect it has something to do with MPICH
itself. As a test of this hypothesis we have
considered using LAMMPI instead. We have
installed LAMMPI, FFTW and GROMACS on a
workstation and PME works great. However
the installation of LAMMPI on our bproc
Linux cluster fails. From reading the
lam-mpi.org archives it is apparent that 
bproc support is only in an experimental
phase, and moreover does not work with older
versions of bproc.

Our cluster uses:
  bproc-3.1.9
  mpich-1.2.2.3
  fftw-2.1.3-2
  gromacs-3.1.3-1

  Suggestions? Has anyone succeeding in running 
parallel PME simulations on a bproc cluster?


Best regards,

  Jason

---
Jason de Joannis, Ph.D.
Chemistry Department, Emory University
1515 Pierce Dr. NE, Atlanta, GA 30322
Phone: (404) 712-2983
Email: jdejoan at emory.edu
http://userwww.service.emory.edu/~jdejoan