[Paraview] parallel run filter operate on plane source

[Bam Ting]

Do not worry! I am not using it as an example of how to!

Rather I like to know how ParaView work in the case of the distribute stream
tracer, it's documnet say in effect - all seed must be the same on all
process. I try something like that my self, and find it do not work. I hope
you can still answer my last question. I won't use it in genral. I only try
to make sense of how distributed stream tracer works in paraview with a
plane source, and I spend already a day on it, I will like in the end to
have the correct understaning.



On Mon, Aug 10, 2009 at 11:38 AM, Moreland, Kenneth <kmorel at sandia.gov>wrote:

>  Please don’t look at the plane source as an example for parallel
> processing.  The plane source has no parallel processing.  The plane source
> is designed to output a small amount of geometry.  Trying to process this in
> parallel would just add overhead rather than provide any real benefit.  If
> you are going to look at anything, please look at components that are
> actually designed to work with distributed partitioned data.  Most of the
> sources are not designed this way because they typically generate small
> geometries.  Many of the readers are designed for parallel processing.
>
> -Ken
>
>
>
> On 8/10/09 12:34 PM, "Bam Ting" <bampingting at gmail.com> wrote:
>
>  I see no logic for this in the plane source. So this is something that
> happen behind the scene in pipeline execute? I guess it works like this: If
> a source say by way of key "I produce piece 0 of 1" and he is run in
> parallel if he is really piece n not 0 of m piece then RequestData is not
> called on him? I this correct?
>
> Thank you for the help! Appreciate much!
>
> On Mon, Aug 10, 2009 at 6:04 AM, Utkarsh Ayachit <
> utkarsh.ayachit at kitware.com> wrote:
>
> You are almost correct. The keys don't cause the data to be transfered
> from process 0, but it causes all processes to produce the full plane
> data as if they were being run on a single processes. Trypically,
> UPDATE_NUMBER_OF_PIECES corresponds to the number of processes, while
> UPDATE_PIECE_NUMBER corresponds to the process index. By setting them
> to 1 and 0 respectively, we are telling the source that we want the
> entire dataset, not a piece of it.
>
> Utkarsh
>
> On Fri, Aug 7, 2009 at 4:47 PM, Bam Ting<bampingting at gmail.com> wrote:
> > I look more at how the distribute stream tracer work and learnt that it
> work
> > by setting: UPDATE_PIECE_NUMBER() to 0 and UPDATE_NUMBER_OF_PIECES to 1.
> > Apparently this cause data from process 0 to be transfer across network
> to
> > other process. I make a guess here, is that correct?
> >
> >
> >
> > On Fri, Aug 7, 2009 at 12:45 PM, Bam Ting <bampingting at gmail.com> wrote:
> >>
> >> Forgive my misconception before! I am baffelt by how this work!
> >>
> >> I looked closer. I see plane source on my imput is empty! bounds 1 -1
> ...
> >> on all process except 0... so that answer something! before this I
> suppose
> >> that the source must be duplicate on all process, because the class doc
> for
> >> vtkPStreamTracer say thaat this must be so. Now I will like to know how
> does
> >> filter such as distributed stream tracer work in this case (plane source
> is
> >> empty on all imput except 0)? Web doc day seed points must be the same
> >> across on all process? Can we explain how paraview works here?
> >>
> >>
> >>
> >>
> >> On Fri, Aug 7, 2009 at 11:22 AM, Bam Ting <bampingting at gmail.com>
> wrote:
> >>>
> >>> Have some issue with a filter that operate on plane source when run
> with
> >>> more than a single process. Plane source on imput seem to be duplicated
> >>> identical on all processes (no problem), then my filter run and make a
> >>> shallow copy of the source and pass to output (problem start!). In the
> >>> spreadsheet view I see more points than I should in the output, + I add
> >>> point data I see more point data as well, I explore with spreadsheet
> view
> >>> show that unexpected point data all duplicate of expected.
> >>>
> >>> My question, how paraview handle source duplicated on all process? How
> >>> paraview can know that the data is duplicate for plane source? How does
> it
> >>> render such duplicate data? I seek to add some point data to plane
> source
> >>> and provide on the output. What I need to do in the filter to avoid the
> >>> issue?
> >>>
> >>
> >
> >
> > _______________________________________________
> > Powered by www.kitware.com <http://www.kitware.com>
> >
> > Visit other Kitware open-source projects at
> > http://www.kitware.com/opensource/opensource.html
> >
> > Please keep messages on-topic and check the ParaView Wiki at:
> > http://paraview.org/Wiki/ParaView
> >
> > Follow this link to subscribe/unsubscribe:
> > http://www.paraview.org/mailman/listinfo/paraview
> >
> >
>
>
>
>
>
>    ****      Kenneth Moreland
>     ***      Sandia National Laboratories
> ***********
> *** *** ***  email: kmorel at sandia.gov
> **  ***  **  phone: (505) 844-8919
>     ***      web:   http://www.cs.unm.edu/~kmorel<http://www.cs.unm.edu/%7Ekmorel>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.paraview.org/pipermail/paraview/attachments/20090810/5b98121f/attachment.htm>