[Paraview] MPI makes Paraview slower?

Berk Geveci berk.geveci at gmail.com
Fri May 5 15:49:02 EDT 2006


What kind of file are you loading? What and how are you rendering? Certain
things will get slower when running with MPI because running in parallel
adds the overhead of interprocess communication. Creating and rendering (in
parallel) a small cone for example will be slower. ParaView tries to
optimize parameters for parallel execution but sometimes the user has to set
certain parameters appropriately.

On 5/5/06, W. Bryan Smith <bryan at ncmir.ucsd.edu> wrote:
>
>
> [apologies if this was posted twice]
>
> hi,
>
> i am (a major newbie) running paraview in
> parallel mode with mpich2 (1.0.3) on a 2
> dual-core cpu machine using this:
>
> mpiexec -n 4 /usr/local/paraview2.5/bin/paraview/
>
> and many things seem slower than just running
> paraview in single cpu mode.  one of the trivial
> things that is noticably MUCH slower is simply
> loading a file: i have a 500MB file that takes
> about 5 seconds to load in single cpu mode, and
> takes about 30-40 seconds when running in MPI.
> also, no matter if i run in single or MPI, the
> slowest process that i have seen so far
> (PainterPolyDataMapper) only uses a single cpu.
> i am sure there is just something i am doing
> wrong.  any hints on how (if!?) i can change
> this behaviour?  i have ordered the book but it
> has not yet arrived...
>
> thanks,
> bryan
>
> W. Bryan Smith
> UCSD/NCMIR
> 9500 Gilman Drive # 0608
> La Jolla, CA, 92093-0608
>
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> The illusion that we are separate from one another
> is an optical delusion of our consciousness.
>                                 -Albert Einstein
> ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> _______________________________________________
> 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/20060505/bea99436/attachment.html


More information about the ParaView mailing list