RE: [AMBER-Developers] Parallel Ptraj

From: Ross Walker <ross.rosswalker.co.uk>
Date: Fri, 4 Sep 2009 20:40:59 +0100

Hi Dave,

> >
> > I just want to bring your attention to some recent changes to ptraj
> in the
> > AMBER11 CVS Tree. Thanks to Paul Frybarger, an undergraduate at Hope
> > College, MI who spent the summer working with me there is now a
> parallel
> > version (MPI) of ptraj.
>
> Can I ask what version of MPI you are using? The lam MPI that we ship
> with
> Amber fails at the pnetcdf configure step, with the error message that
> it
> does not support MPI-IO.

Yes you have to support MPI-IO or you are toast. I assumed every
self-respecting MPI implementation would support it since it is part of the
standard. It is possible to build the MPI without ROMIO support which might
be the problem - rather than it not actually supporting MPI-IO full stop.

I use MPICH2-1.0.7 compiled with ifort v10.1.018.

> I'm still fiddling with the configuration
> scripts
> and Makefiles to recover from this (so that one could compile the rest
> of
> AmberTools with MPI turned on, even though ptraj.MPI will fail.

What other AmberTools has MPI right now? When I was looking it looked like
the -mpi flag didn't turn anything on so that the only MPI program ended up
being AmberTools.

> Some thoughts:
>
> 1. Could we have a parallel version of ptraj without netcdf support?
> (i.e.
> make that an option).

Possibly although this defeats most of the purpose since the MPI-IO is used
to read the trajectory in parallel and then each action works on a part of
the trajectory. Paul might be able to answer this better since I haven't
looked for sure. It might be possible to run without MPI-IO support which
would just mean netcdf files would have to be read in serial - I think ascii
files could be converted to just use fileoffsets and multiple open
statements in place of the MPI-IO.

Paul can you comment here?

> 2. Is there an MPI whose source we could supply with Amber that does
> support MPI-IO?

MPICH2 supports it - as does regular MPICH I believe. We should check it
isn't just that we are skipping MPI-IO in the lam build. Normally you just
do a configure with something like --with-romio to enable the MPI-IO
support.

All the best
Ross


/\
\/
|\oss Walker

| Assistant Research Professor |
| San Diego Supercomputer Center |
| Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
| http://www.rosswalker.co.uk | PGP Key available on request |

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-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Mon Sep 14 2009 - 11:52:38 PDT
Custom Search