[vtkusers] streaming and parallel execution

Berk Geveci berk.geveci at kitware.com
Thu May 24 15:07:05 EDT 2007


I recommend looking at the paraview web site (http://paraview.org).
ParaView uses the distributed capabilities of VTK to do distributed
processing and rendering (which would allow you to use multiple
cores/CPUs as well as multiple GPUs). You should also check this paper
http://portal.acm.org/citation.cfm?id=618835 as well as these:
http://www.nvidia.com/object/IO_19962.html
http://www.nvidia.com/object/IO_27539.html
This is what VTK does using MPI.

On the shared-memory, multi-threading side, a large of imaging
algorithms and CPU-based volume rendering algorithms will
automatically use multiple cores/CPUs when available (it is possible
to disable this by setting max number of threads to 1). GPU-based
volume rendering algorithms do not do this.

-berk


On 5/24/07, Steve M. Robbins <steve at sumost.ca> wrote:
> Howdy,
>
> I've been asked to report on the "scalability" of VTK, by which is
> meant what support is available for large data sets.
>
> I have read
> http://www.vtk.org/Wiki/VTK_FAQ#How_to_handle_large_data_sets_in_VTK
> which describes how to reduce memory usage.  I thought there was also
> a "streaming" mode whereby filters would process the data in pieces.
> I see vtkStreamingDemandDrivenPipeline in the doxygen docs, with an
> example program
> (Examples/ParallelProcessing/Generic/Cxx/ParallelIso.cxx) that
> suggests one has to use this explicitly in the pipeline.  Is that the
> case?
>
> On the other end of the spectrum, if you have enough memory and
> multiple CPUs, you might want to process the pipeline in parallel.  I
> have a vague recollection that VTK magically detected the number of
> processors and set up the streaming automatically.  Is that the case?
> How can I tell whether a filter can take advantage of multi-threading?
> I found vtkThreadedImageAlgorithm, which claims to support
> multithreading -- is the set of its subclasses the set of filters that
> support multi-threading?
>
> Is there a way to take advantage of multiple GPUs in a machine -- e.g.
> to speed up volume rendering or to allow volume rendering of a larger
> dataset?
>
>
> Thanks,
> -Steve
>
> -----BEGIN PGP SIGNATURE-----
> Version: GnuPG v1.4.6 (GNU/Linux)
>
> iD8DBQFGVczT0i2bPSHbMcURAu2OAJ976C9KYubhT0kLcYP8SvKOCNnnOACfTZgl
> ovMIGwxhY6DE7CX1ppCT+Mk=
> =Z5AM
> -----END PGP SIGNATURE-----
>
> _______________________________________________
> This is the private VTK discussion list.
> Please keep messages on-topic. Check the FAQ at: http://www.vtk.org/Wiki/VTK_FAQ
> Follow this link to subscribe/unsubscribe:
> http://www.vtk.org/mailman/listinfo/vtkusers
>
>


-- 
 Berk Geveci
 Kitware Inc.
 28 Corporate Drive
 Clifton Park, NY, 12065



More information about the vtkusers mailing list