Re: [Condor-users] Does Condor support the PVM on Windows?

[Erik Paulson <epaulson@xxxxxxxxxxx>]

On Sat, Aug 19, 2006 at 11:34:10AM +0700, Nano Surbakti wrote:
> > However, the 'parallel universe' does work on Windows --- and off the
> > top of my head, I cannot think of any reason why you could not use
> > the parallel universe to launch "regular" pvm jobs.
> 
> Does the "parallel universe" mean MPI ?
> 

It means any type of parallel job. The parallel universe just allocates
machines together. It can be an MPI job, a PVM job, or some sort of 
home-grown parallel library.

The key is to create the launching script. Condor 6.8.0 includes a launching 
script for MPICH 1.2.{4,5,6} on UNIX and LAM MPI. Here's an example submit
file for LAM:

######################################
## Example submit description file
## for LAM MPI
######################################
universe = parallel
executable = lamscript
arguments = my_lam_linked_executable arg1 arg2
machine_count = 4
should_transfer_files = yes
when_to_transfer_output = on_exit
transfer_input_files = my_lam_linked_executable
queue


The "executable" is the lamscript. The lamscript that is included in the
condor distribution runs lamboot, then mpirun, then calls lamhalt. The 
parallel universe doesn't have to anything about how LAM works - all of
that is in the lamscript. (Think of it as a device driver).

Condor 6.8.0 only includes out of the box scripts(device drivers) for 
MPICH 1.2.x on UNIX and LAM MPI. If you want to use the parallel universe
with any other parallel system (like PVM, or FT-MPI, or OpenMPI) you'd
have to write your own script to do so, we haven't written any yet. 
That includes MPICH-NT - we haven't written a script to run an MPI job on
Windows yet. If anyone has written one we'd be glad to include it in 
the Condor releases.

-Erik