[Paraview] ParaView in parallel

[Moreland, Kenneth]

Yes.  You should return the same structure for each MPI process.  In fact, you should have the same non-null leaves on each process (that is, the same vtkUnstructuredGrids and vtkPolyDatas) so that parallel iterations over the structure return corresponding data sets.  However, it is perfectly valid to have blocks that contain zero points and cells.

-Ken


On 10/1/09 10:33 AM, "Paul Edwards" <paul.m.edwards at gmail.com> wrote:

Hi,

Thanks for your responses.  My dataset a multi-block structure consisting of zones, each containing a single volume and multiple surfaces, e.g.

 *   zone1 (vtkMultiBlockDataSet)
    *   volume (vtkUnstructuredGrid)
    *
    *   surfaces (vtkMultiBlockDataSet)
       *   surface1 (vtkPolyData)
       *
       *   surface2 (vtkPolyData)
       *
       *   ...
 *   ...
 *
 *   zoneN
    *   volume
    *   surfaces
       *   surface1
       *   surface2
       *   ...

I'm not sure how to keep this structure when writing a parallel reader?  Do I just return exactly the same structure for each MPI process with different data?

Thanks,
Paul


2009/9/23 Moreland, Kenneth <kmorel at sandia.gov>
You said your data is multiblock, but you did not say how many blocks it has.  D3 is going to run its partitioning algorithm on each block independently.  This is bad if you have lots of blocks; you will end up with lots of tiny pieces on all the processes.

To make your reader "parallel", you basically just have to set vtkStreamingDemandDrivenPipeline::MAXIMUM_NUMBER_OF_PIECES() in the output information in the RequestInformation call (usually you set it to -1 to allow the downstream pipeline to set any number of pieces) and then read vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_PIECES() and vtkStreamingDemandDrivenPipeline::UPDATE_PIECE_NUMBER(), which will correspond to the number of processes and local rank in ParaView, in the RequestData call.

If your data typically has lots of blocks, I would design the reader to simply assign blocks to pieces so that each piece has about the same amount of data and then read in the blocks associated with the requested piece.  This should yield the simplest code and the fastest reading times, and the approach (generally) does not require you to worry about ghost cells.

The simplest example of a reader that responds to piece requests is the vtkParticleReader.  That reader breaks up a list of points into the requested pieces, but it is not a stretch to do the same thing with the blocks.

-Ken



On 9/23/09 3:09 AM, "Paul Edwards" <paul.m.edwards at gmail.com <http://paul.m.edwards@gmail.com> > wrote:

Hi,

I've been experimenting with ParaView in parallel without much.  Below I have listed my setup, experiences and questions!

Setup
My setup is a 6 node gigabit cluster (without gfx cards), compiled with OSMesa.  The filesystem is only a RAID on the frontend exported with NFS.  Each node has 8 cores and I experimented running with both 1 core per node and also 8 cores per node

Data
The data I am reading is multiple blocks, where each contains an unstructured grid and a multiblock structure of polydata surfaces.  The data is about 10 million points in total.

Reader
My reader is not written for parallel so I loaded in the mesh and partitioned it with D3.  The partitioning took a long time - approx 2 mins - is this normal?  When the data was partitioned I didn't really notice much difference in speed and the rendering performance was considerably worse than a single node for the same data.  Is this just the result of using OSMesa?  If so, does anyone have any suggestions for the number of nodes to be running on?  I also tried saving my data once it was partitioned but then loading it actually took longer (is this due to a lack of parallel filesystem?)

Filters
I tried to run some of my custom filters but one that calculates scalar variables for pointdata didn't display the variables in the GUI after running the filter.  Do I to do something different for parallel filters?

Finally, is there any documentation for implementing parallel readers (or simple examples) in ParaView?  And, does anyone have any suggestions for how to split up the data?

Thanks,
Paul



   ****      Kenneth Moreland
    ***      Sandia National Laboratories
***********
*** *** ***  email: kmorel at sandia.gov <http://kmorel@sandia.gov>
**  ***  **  phone: (505) 844-8919
    ***      web:   http://www.cs.unm.edu/~kmorel <http://www.cs.unm.edu/%7Ekmorel>





   ****      Kenneth Moreland
    ***      Sandia National Laboratories
***********
*** *** ***  email: kmorel at sandia.gov
**  ***  **  phone: (505) 844-8919
    ***      web:   http://www.cs.unm.edu/~kmorel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20091001/f578fa24/attachment-0001.htm>