[Paraview] Stalls in distributed mode

[Berk Geveci]

Did you try attaching to the paraview process with a debugger to see where
it is stuck? With MPI, specially exotic vendor specific implementation, it
is common to run into problems and not see any error output.
To get more information on what paraview is doing, you can use profiling. I
don't know if you MPI implementation supports it. There are a few flavor of
it. At the simplest level, you link your application to a set of  special
MPI libraries. When you run the application, it produces a log. You can
later use some other tool to look at the result. At the most sophisticated
level, you instrument your code by adding specialized profiling calls.
ParaView doesn't do that.

-Berk

On 7/3/06, Adam M. Smith <amsmith at lanl.gov> wrote:
>
> Hi all,
>
> I am trying to use ParaView on a linux cluster using MVAPICH.  I have
> built the latest ParaView sources with MPI enabled (as far as I know). I
> can launch all of the created binaries with mpirun_rsh and have them work
> fine as long as I only use "-np 1".  Even the simple "mpirun -np 2
> thismachine thismachine /path/to/paraview" stalls.  When I say stalls, I
> mean a process is created, nothing is output to stdout nor do any X
> windows pop up.  CPU utilization remains at 0. Other programs using
> mvapich on these machines work fine.
>
> The (I think) relavent guts of my cmake config are below:
>
> BUILD_SHARED_LIBS=OFF
> MPI_EXTRA_LIBRARY=/usr/local/lib/libibverbs
> MPI_INCLUDE_PATH=/usr/local2/mvapich/include
> MPI_LIBRARY=/usr/local2/mvapich/lib/libmpich.a
> VTK_MAX_NUMPROCS=2 # I didn't set this myself, configure probably got this
> (correctly)
> VTK_USE_MPI=ON
> VTK_MPIRUN_EXE=/usr/local2/mvapich/bin/mpirun
>
> Separate but related question, probaby a simple one at that -- I'm new to
> MPI.  Does anyone know of a nice way to somehow "watch" whats going on
> with MPI to see which hosts are talking to which other hosts when.. in
> which communicators?
> _______________________________________________
> ParaView mailing list
> ParaView at paraview.org
> http://www.paraview.org/mailman/listinfo/paraview
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://public.kitware.com/pipermail/paraview/attachments/20060703/1b482232/attachment.htm