<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
</head>
<body dir="auto">
<div>Yes. In fact that is the whole basic point of the pipeline is that you can call update just once at the bottom of the pipeline and it will pull all necessary data to get the result.</div>
<div><br>
</div>
<div>Calling update on every filter in the pipeline is at best unnecessary. VTK will remember what the last result of each filter execution is, so when you call update at the bottom it will simply reuse previous updates. However, if a filter does something
like request ghost cells, then the results from your update on upstream filters becomes invalid and has to be re-executed. Therefore, it is not only convenient to call update once at the bottom, it is generally more efficient.</div>
<div><br>
<div>-Ken</div>
<div><br>
</div>
Sent from my iPad so blame autocorrect.</div>
<div><br>
On Nov 8, 2012, at 2:36 AM, "Andrew Parker" <<a href="mailto:andy.john.parker@googlemail.com">andy.john.parker@googlemail.com</a>> wrote:<br>
<br>
</div>
<blockquote type="cite">
<div>Hi,
<div><br>
</div>
<div>Many thanks. Based on that can you confirm that I have been making a mistake then? I normally call update() on each filter in turn, basically because I didn't know not to, I presume given what you've said plus the call stack in totalview that I should
only call Update() on the very last filter in the ancestry/furtherest downstream??</div>
<div><br>
</div>
<div>Cheers,</div>
<div>Andy<br>
<br>
<div class="gmail_quote">On 7 November 2012 17:01, Moreland, Kenneth <span dir="ltr">
<<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div style="font-size:14px;font-family:Calibri,sans-serif;word-wrap:break-word">
<div>
<div>
<div>By "downstream" I mean a separate filter that is connected to the output of vtkDistributedDataFilter or further down the output ancestry. In your original example you had the chain of filters vtkDistributedDataFilter -> vtkPieceScalars -> vtkDataSetSurfaceFilter.
vtkPieceScalars and vtkDataSetSurfaceFilter are "downstream" from vtkDistributedDataFilter in this example.</div>
<div><br>
</div>
<div>-Ken</div>
<div>
<div><br>
</div>
</div>
</div>
</div>
<span>
<div style="border-right:medium none;padding-right:0in;padding-left:0in;padding-top:3pt;text-align:left;font-size:11pt;border-bottom:medium none;font-family:Calibri;border-top:#b5c4df 1pt solid;padding-bottom:0in;border-left:medium none">
<span style="font-weight:bold">From: </span>Andrew Parker <<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>><br>
<span style="font-weight:bold">Date: </span>Wednesday, November 7, 2012 9:57 AM<br>
<span style="font-weight:bold">To: </span>Kenneth Moreland <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>><br>
<span style="font-weight:bold">Cc: </span>George Zagaris <<a href="mailto:george.zagaris@kitware.com" target="_blank">george.zagaris@kitware.com</a>>, "<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>" <<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>>,
"<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>" <<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>>
<div>
<div class="h5"><br>
<span style="font-weight:bold">Subject: </span>[EXTERNAL] Re: [vtkusers] [Paraview] vtkDistributedDataFilter and ghost cells - PROGRESS<br>
</div>
</div>
</div>
<div>
<div class="h5">
<div><br>
</div>
<blockquote style="BORDER-LEFT:#b5c4df 5 solid;PADDING:0 0 0 5;MARGIN:0 0 0 5">
<div>
<div>Ok thanks again for the advice. Think it's really clear I'm confused between the "downstream" nomenclature. Do mean just a class that constructs or takes a D3 as input that I write from scratch outside of vtk? Or do you mean something totally different?
Clearly my understanding of "it was to inherit" was totally wrong. I would prefer to stick with the downstream filter if that's more robust.
<div><br>
</div>
<div>Could you give me a quick explanation in this context? In addition could you point me towards an example of using a downstream filter in this context?</div>
<div><br>
</div>
<div>Sorry again for the confusion,</div>
<div>Andy<br>
<br>
<div class="gmail_quote">On 7 November 2012 16:48, Moreland, Kenneth <span dir="ltr">
<<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div style="font-size:14px;word-wrap:break-word">
<div style="font-family:Calibri,sans-serif">
<div>
<div>The instructions I gave were for a downstream filter separate from D3. If you want to modify D3 itself (which sounds like a reasonable good idea for your purposes) then you will have change the value for UPDATE_NUMBER_OF_GHOST_LEVELS that the RequestData
reads. If you want to go with this approach, first blow away your RequestUpdateExtent method. Let D3 do its regular thing. Then implement a RequestData method like the following:</div>
</div>
</div>
<div style="font-family:Calibri,sans-serif"><br>
</div>
<blockquote style="margin-top:0px;margin-right:0px;margin-bottom:0px;margin-left:40px;border-top-style:none;border-right-style:none;border-bottom-style:none;border-left-style:none;border-width:initial;border-color:initial;padding-top:0px;padding-right:0px;padding-bottom:0px;padding-left:0px">
<div><font face="Consolas">int MyD3::RequestData(</font></div>
<div><font face="Consolas">
<div> vtkInformation *request,</div>
<div> vtkInformationVector **inputVector,</div>
<div> vtkInformationVector *outputVector)</div>
<div>
<div>{</div>
<div>
<div> vtkInformation *outInfo = outputVector->GetInformationObject(0);</div>
</div>
<div><br>
</div>
</div>
<div> int requestedGhostLevel = outInfo->Get(</div>
<div>
<div> vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_GHOST_LEVELS());</div>
</div>
<div> outInfo->Set(</div>
</font><span style="font-family:Consolas">
<div>
<div> vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_GHOST_LEVELS(),</div>
</div>
<div> requestedGhostLevel + 1);</div>
<div><br>
</div>
<div> return this->Superclass::RequestData(request, inputVector, outputVector);</div>
<div>}</div>
<div><br>
</div>
</span></div>
</blockquote>
<div style="font-family:Calibri,sans-serif">I don't know if the executive is going to get upset about you changing UPDATE_NUMBER_OF_GHOST_LEVELS in RequestData, but I think it will be all right.</div>
<div style="font-family:Calibri,sans-serif"><br>
</div>
<div style="font-family:Calibri,sans-serif">-Ken</div>
<div style="font-family:Calibri,sans-serif"><br>
</div>
<span style="font-family:Calibri,sans-serif">
<div style="border-right:medium none;padding-right:0in;padding-left:0in;padding-top:3pt;text-align:left;font-size:11pt;border-bottom:medium none;font-family:Calibri;border-top:#b5c4df 1pt solid;padding-bottom:0in;border-left:medium none">
<span style="font-weight:bold">From: </span>Andrew Parker <<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>><br>
<span style="font-weight:bold">Date: </span>Wednesday, November 7, 2012 9:31 AM<br>
<span style="font-weight:bold">To: </span>George Zagaris <<a href="mailto:george.zagaris@kitware.com" target="_blank">george.zagaris@kitware.com</a>><br>
<span style="font-weight:bold">Cc: </span>Kenneth Moreland <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>>, "<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>" <<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>>,
"<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>" <<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>><br>
<span style="font-weight:bold">Subject: </span>[EXTERNAL] Re: [vtkusers] [Paraview] vtkDistributedDataFilter and ghost cells - PROGRESS<br>
</div>
<div>
<div>
<div><br>
</div>
<blockquote style="BORDER-LEFT:#b5c4df 5 solid;PADDING:0 0 0 5;MARGIN:0 0 0 5">
<div>
<div>So I've added this in an .cpp file and complied it:
<div><br>
</div>
<div>
<div>vtkStandardNewMacro(MyD3)</div>
<div><br>
</div>
<div>MyD3::MyD3()</div>
<div>: vtkDistributedDataFilter()</div>
<div>{</div>
<div>}</div>
<div><br>
</div>
<div>void MyD3::PrintSelf(ostream& os, vtkIndent indent)</div>
<div>{ </div>
<div> this->Superclass::PrintSelf(os,indent);</div>
<div>}</div>
<div><br>
</div>
<div>int MyD3::RequestUpdateExtent(vtkInformation *vtkNotUsed(request),</div>
<div><span style="white-space:pre-wrap"></span> vtkInformationVector **inputVector,</div>
<div><span style="white-space:pre-wrap"></span> vtkInformationVector *outputVector)</div>
<div>{</div>
<div> // get the info objects</div>
<div> vtkInformation *inInfo = inputVector[0]->GetInformationObject(0);</div>
<div> vtkInformation *outInfo = outputVector->GetInformationObject(0);</div>
<div> </div>
<div> int piece, numPieces, ghostLevels;</div>
<div> </div>
<div> // We require an extra layer of ghost cells from upstream.</div>
<div> </div>
<div> piece = outInfo->Get(vtkStreamingDemandDrivenPipeline::UPDATE_PIECE_NUMBER());</div>
<div> numPieces = outInfo->Get(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_PIECES());</div>
<div> ghostLevels = outInfo->Get(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_GHOST_LEVELS()) + 1; //##### New bit, modified from Ken's notes</div>
<div> </div>
<div> inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_PIECE_NUMBER(), piece);</div>
<div> inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_PIECES(), numPieces);</div>
<div> inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_GHOST_LEVELS(), ghostLevels);</div>
<div> </div>
<div> return 1;</div>
<div>}</div>
<div><br>
</div>
<div>And I add this to a .h file:</div>
<div><br>
</div>
<div>
<div>class MyD3</div>
<div> : public vtkDistributedDataFilter</div>
<div>{</div>
<div> vtkTypeMacro(MyD3, vtkDistributedDataFilter);</div>
<div> </div>
<div>public:</div>
<div> static MyD3 *New();</div>
<div> void PrintSelf(ostream& os, vtkIndent indent);</div>
<div> int RequestUpdateExtent(vtkInformation *vtkNotUsed(request),</div>
<div><span style="white-space:pre-wrap"></span> vtkInformationVector **inputVector,</div>
<div><span style="white-space:pre-wrap"></span> vtkInformationVector *outputVector);</div>
<div>protected:</div>
<div> MyD3();</div>
<div> </div>
<div>private:</div>
<div> // MyD3(const MyD3&); // Not implemented.</div>
<div> // void operator=(const MyD3&); // Not implemented.</div>
<div>};</div>
</div>
<div><br>
</div>
<div>I then invoke it in parallel 3 way, as I did before, but now using:</div>
<div><br>
</div>
<div>
<div> vtkSmartPointer<MyD3> dd = vtkSmartPointer<MyD3>::New();</div>
<div> </div>
<div> // here mesh is a vtkUnstructuredGrid on each processor, obtained from other (non-vtk) means, but represents a valid and checked subset of the entire mesh.</div>
<div> dd->SetInput(mesh);</div>
<div> dd->SetController(getVtkController());</div>
<div> dd->SetBoundaryModeToSplitBoundaryCells(); // clipping</div>
<div> dd->UseMinimalMemoryOff();</div>
<div> dd->Update();</div>
<div> </div>
<div> vtkSmartPointer<vtkUnstructuredGrid> ddUmesh(vtkUnstructuredGrid::SafeDownCast(dd->GetOutput()));</div>
<div> </div>
<div> auto ids = ddUmesh->GetPointData()->GetGlobalIds();</div>
<div> auto cids = ddUmesh->GetCellData()->GetGlobalIds();</div>
<div><br>
</div>
</div>
<div>########################</div>
<div>Notes:</div>
<div><br>
</div>
<div>1) Both ids, and cids remain null, although I now understand from Ken that cids will always be null as they're not created.</div>
<div>2) Using totalview and stopping the code in parallel 3 way on line 670 of vtkDistributedDataFilter.cxx, the member this->GhostLevel is always zero and never executes codes branches dependent on it, however, the above code successfully increments the ghostLevels
to one.</div>
<div>3) So basically moved forward a little bit, but still not working. Any thoughts??</div>
<div>4) If I had created the ghost cells, how would I know, is my next question. How can I visualise them to check?</div>
<div><br>
</div>
<div>Any thoughts really welcome.</div>
<div><br>
</div>
<div>Cheers again,</div>
<div>Andy</div>
<div><br>
</div>
<div><br>
</div>
<br>
<div class="gmail_quote">On 7 November 2012 14:50, George Zagaris <span dir="ltr">
<<a href="mailto:george.zagaris@kitware.com" target="_blank">george.zagaris@kitware.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div>On Wed, Nov 7, 2012 at 9:34 AM, Andrew Parker<br>
<<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>> wrote:<br>
> Many thanks for that. I perform some threshold stuff early on, so basically<br>
><br>
<br>
</div>
Perhaps an easy solution is to append a "GlobalCellID" and<br>
"GlobalNodeID" before thresholding. To the best of my knowledge,<br>
threshold will extract vtkUnstructuredGrid instance from whatever goes<br>
in the input and not automatically retain this type of global<br>
information.<br>
<br>
Hope this helps.<br>
<br>
Best,<br>
George<br>
<div>
<div><br>
> On 7 November 2012 13:00, George Zagaris <<a href="mailto:george.zagaris@kitware.com" target="_blank">george.zagaris@kitware.com</a>> wrote:<br>
>><br>
>> Hello,<br>
>><br>
>> Since the data is essentially structured (rectilinear grids), I think<br>
>> the best way to get the global/local mapping is to use extent<br>
>> information. Specifically, you need the whole extent of the grid and<br>
>> the global extent of each partition. Then from the global IJK of a<br>
>> point or cell, and the whole extent you can use methods from<br>
>> vtkStructuredData to compute the corresponding global linear index for<br>
>> a point or cell. Do you have extent information?<br>
>><br>
>> Also, I have done some work on ghost generation in VTK for structured<br>
>> (and AMR) data. Unfortunately, currently it doesn't work outside the<br>
>> box with rectilinear grids. This functionality can still be used to<br>
>> create ghost layers and exchange the fields (node and/or cell data) on<br>
>> the grid but, it doesn't communicate the coordindates on the<br>
>> rectilinear grid. Alternatively, you may be able to converge to a<br>
>> solution faster if you convert the rectilinear grid to a<br>
>> vtkStructuredGrid and use vtkPStructuredGridConnectivity or<br>
>> vtkPStructuredGridGhostDataGenerator to generate ghost layers. There<br>
>> are tests that illustrate how to use these classes and I can provide<br>
>> more information if you are interested.<br>
>><br>
>> Best,<br>
>> George<br>
>><br>
>> On Wed, Nov 7, 2012 at 7:22 AM, Moreland, Kenneth <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>><br>
>> wrote:<br>
>> > Hmmm. I thought global point IDs were always created. Maybe it only<br>
>> > happens<br>
>> > when ghost cells are created. Try getting that working first.<br>
>> ><br>
>> > As far as global cell IDs are concerned, you are out of luck.<br>
>> > vtkDistributedDataFilter does not create them. But they are not all that<br>
>> > useful anyway since, unlike points, once you remove the out cells there<br>
>> > is<br>
>> > no overlap. Usually you can get all the information you want from the<br>
>> > ghost<br>
>> > level.<br>
>> ><br>
>> > If you really need global cell IDs, I'm sure you would have no trouble<br>
>> > making them yourself. I would do it before running<br>
>> > vtkDistributedDataFilter.<br>
>> > Assuming when you first read in your data you have no duplicate cells,<br>
>> > just<br>
>> > label those on process 0 as 0..N_0, those on process 1 as N_0+1..N_1,<br>
>> > and so<br>
>> > on.<br>
>> ><br>
>> > -Ken<br>
>> ><br>
>> > Sent from my iPad so blame autocorrect.<br>
>> ><br>
>> > On Nov 7, 2012, at 2:38 AM, "Andrew Parker"<br>
>> > <<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>> wrote:<br>
>> ><br>
>> > So that's working now in terms of the cast, forgot GetOutput() inside<br>
>> > the<br>
>> > cast operator! The returned vtkUGrid is now fully populated and<br>
>> > contains<br>
>> > sensible information.<br>
>> ><br>
>> > However, both getPointData()->GetGlobalIds() and<br>
>> > getCellData()->GetGlobalIds() return null pointers. Any thoughts?<br>
>> ><br>
>> > Also, should I be using CellData since I want the cell global to local<br>
>> > mapping for cells not the nodes, at the moment at least?<br>
>> ><br>
>> > On 6 November 2012 19:05, Moreland, Kenneth <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>> wrote:<br>
>> >><br>
>> >> Perhaps it is outputting a composite data set of some type? Try<br>
>> >> running<br>
>> >> GetClassName() to see what type of data object it really is.<br>
>> >><br>
>> >> -Ken<br>
>> >><br>
>> >> From: Andrew Parker <<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>><br>
>> >> Date: Tuesday, November 6, 2012 9:50 AM<br>
>> >><br>
>> >> To: Kenneth Moreland <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>><br>
>> >> Cc: "<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>" <<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>>, "<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>"<br>
>> >> <<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>><br>
>> >> Subject: [EXTERNAL] Re: [Paraview] vtkDistributedDataFilter and ghost<br>
>> >> cells<br>
>> >><br>
>> >> Thanks on both accounts. Any thoughts why the downcast called after<br>
>> >> dd->Update() on distributedDataFilter is a null pointer? As in, dd is<br>
>> >> working perfectly properly, but I don't seem to be able to extract a<br>
>> >> valid<br>
>> >> unstructuredgrid. For a follow up question I assume<br>
>> >> getPointData()->GetGlobalIds() gives the local to global for the mesh<br>
>> >> nodes,<br>
>> >> and I should use getCellData()->GetGlobalIds() to get the local to<br>
>> >> global<br>
>> >> for the cells? Once I get a valid pointer that is....<br>
>> >><br>
>> >> Cheers again,<br>
>> >> Andy<br>
>> >><br>
>> >> On 6 November 2012 16:40, Moreland, Kenneth <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>> wrote:<br>
>> >>><br>
>> >>> You should be able to do a vtkUnstructuredGrid::SafeDownCast() to the<br>
>> >>> data to get the unstructured mesh (and access to the point data).<br>
>> >>><br>
>> >>> -Ken<br>
>> >>><br>
>> >>> From: Andrew Parker <<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>><br>
>> >>> Date: Tuesday, November 6, 2012 9:32 AM<br>
>> >>> To: Kenneth Moreland <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>><br>
>> >>> Cc: "<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>" <<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>>, "<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>"<br>
>> >>> <<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>><br>
>> >>> Subject: [EXTERNAL] Re: [Paraview] vtkDistributedDataFilter and ghost<br>
>> >>> cells<br>
>> >>><br>
>> >>> Another question which I'd forgotten, how do I get to a<br>
>> >>> vtkUnstructuredGrid per processor from the vtkDistributedDataFilter.<br>
>> >>><br>
>> >>> For instance, dd->GetOutput()->GetPointData()->GetGlobalIds() doesn't<br>
>> >>> work as it's a vtkDataObject<br>
>> >>><br>
>> >>> Stupid question I'm sure but the doxy notes say this type returns an<br>
>> >>> unstructured mesh, but I can't seem to get it out?<br>
>> >>><br>
>> >>> Also, why exactly do I need the vtkPieceScalars and<br>
>> >>> vtkDataSetSurfaceFilter again? If the above can be made to work and<br>
>> >>> return<br>
>> >>> the mapping, what are they adding in terms of information?<br>
>> >>><br>
>> >>> Thanks again,<br>
>> >>> Andy<br>
>> >>><br>
>> >>> On 6 November 2012 16:00, Moreland, Kenneth <<a href="mailto:kmorel@sandia.gov" target="_blank">kmorel@sandia.gov</a>> wrote:<br>
>> >>>><br>
>> >>>> I believe vtkDistributedDataFilter will always return with global<br>
>> >>>> point<br>
>> >>>> ids (a mapping from local point ids to global point ids), although it<br>
>> >>>> might<br>
>> >>>> pass them if they already exist. So<br>
>> >>>> dd->GetOutput()->GetPointData()->GetGlobalIds() should return the<br>
>> >>>> array that<br>
>> >>>> gives this mapping.<br>
>> >>>><br>
>> >>>> Ghost cells are only created on demand, and this is usually done by<br>
>> >>>> pipeline convention. If you have a filter that needs a layer of<br>
>> >>>> ghost<br>
>> >>>> cells, it should override the RequestUpdateExtent method to increment<br>
>> >>>> the<br>
>> >>>> value of UPDATE_NUMBER_OF_GHOST_LEVELS from the output information to<br>
>> >>>> the<br>
>> >>>> input information. This method would look something like this.<br>
>> >>>><br>
>> >>>> int vtkDistributedDataFilter::RequestUpdateExtent(<br>
>> >>>> vtkInformation *vtkNotUsed(request),<br>
>> >>>> vtkInformationVector **inputVector,<br>
>> >>>> vtkInformationVector *outputVector)<br>
>> >>>> {<br>
>> >>>> // get the info objects<br>
>> >>>> vtkInformation *inInfo = inputVector[0]->GetInformationObject(0);<br>
>> >>>> vtkInformation *outInfo = outputVector->GetInformationObject(0);<br>
>> >>>><br>
>> >>>> int piece, numPieces, ghostLevels;<br>
>> >>>><br>
>> >>>> // We require an extra layer of ghost cells from upstream.<br>
>> >>>><br>
>> >>>> piece =<br>
>> >>>><br>
>> >>>> outInfo->Get(vtkStreamingDemandDrivenPipeline::UPDATE_PIECE_NUMBER());<br>
>> >>>> numPieces =<br>
>> >>>><br>
>> >>>><br>
>> >>>> outInfo->Get(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_PIECES());<br>
>> >>>> ghostLevels =<br>
>> >>>><br>
>> >>>><br>
>> >>>> outInfo->Get(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_GHOST_LEVELS());<br>
>> >>>><br>
>> >>>><br>
>> >>>> inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_PIECE_NUMBER(),<br>
>> >>>> piece);<br>
>> >>>><br>
>> >>>><br>
>> >>>> inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_PIECES(),<br>
>> >>>> numPieces);<br>
>> >>>><br>
>> >>>><br>
>> >>>> inInfo->Set(vtkStreamingDemandDrivenPipeline::UPDATE_NUMBER_OF_GHOST_LEVELS(),<br>
>> >>>> ghostLevels);<br>
>> >>>><br>
>> >>>> return 1;<br>
>> >>>> }<br>
>> >>>><br>
>> >>>><br>
>> >>>> The operation of the RequestData method should also strip off this<br>
>> >>>> layer<br>
>> >>>> of ghost cells. It might be possible to request a layer of ghost<br>
>> >>>> cells by<br>
>> >>>> setting UPDATE_NUMBER_OF_GHOST_LEVELS at the bottom of the pipeline,<br>
>> >>>> but I'm<br>
>> >>>> not totally sure how to make that work. It's probably easier (or at<br>
>> >>>> least<br>
>> >>>> cleaner) to do it from within a filter.<br>
>> >>>><br>
>> >>>> -Ken<br>
>> >>>><br>
>> >>>> From: Andrew Parker <<a href="mailto:andy.john.parker@googlemail.com" target="_blank">andy.john.parker@googlemail.com</a>><br>
>> >>>> Date: Tuesday, November 6, 2012 8:25 AM<br>
>> >>>> To: "<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>" <<a href="mailto:vtkusers@vtk.org" target="_blank">vtkusers@vtk.org</a>>, "<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>"<br>
>> >>>> <<a href="mailto:paraview@paraview.org" target="_blank">paraview@paraview.org</a>><br>
>> >>>> Subject: [EXTERNAL] [Paraview] vtkDistributedDataFilter and ghost<br>
>> >>>> cells<br>
>> >>>><br>
>> >>>> Hi,<br>
>> >>>><br>
>> >>>> Hope you can help. I have some code running in parallel, that by<br>
>> >>>> other<br>
>> >>>> means I have constructed nprocs worth of vtkRectilinearGrids, one per<br>
>> >>>> process. Each of which is a valid nprocs-worth of the whole serial<br>
>> >>>> mesh,<br>
>> >>>> I've check this and I am happy with that i.e. it's partitioned<br>
>> >>>> properly and<br>
>> >>>> nothing is missing. I need the following information to process my<br>
>> >>>> data in<br>
>> >>>> parallel:<br>
>> >>>><br>
>> >>>> 1) I would like the local -> global cell mapping between the local<br>
>> >>>> rgrid<br>
>> >>>> and the corresponding global single mesh.<br>
>> >>>> 2) I would like to know which cells are on processor boundaries for<br>
>> >>>> parallel exchange purposes.<br>
>> >>>> 3) I would like all the double arrays per processor to be "expanded"<br>
>> >>>> by<br>
>> >>>> the amount of (1 level of) ghost cells such that I can properly do<br>
>> >>>> the<br>
>> >>>> computations I want with the ability to exchange only those<br>
>> >>>> additional cells<br>
>> >>>> given the local to global mapping.<br>
>> >>>><br>
>> >>>> I have tried from the examples to use the following code, which I<br>
>> >>>> call<br>
>> >>>> on every process, each of which has it's own local rgrid as I said.<br>
>> >>>> I do<br>
>> >>>> the following:<br>
>> >>>><br>
>> >>>> vtkSmartPointer<vtkDistributedDataFilter> dd =<br>
>> >>>> vtkSmartPointer<vtkDistributedDataFilter>::New();<br>
>> >>>> dd->SetInput(rgrid);<br>
>> >>>><br>
>> >>>> dd->SetController(getVtkController());<br>
>> >>>> dd->SetBoundaryModeToSplitBoundaryCells();<br>
>> >>>> //dd->SetBoundaryModeToAssignToOneRegion();<br>
>> >>>> //dd->SetBoundaryModeToAssignToAllIntersectingRegions();<br>
>> >>>> dd->UseMinimalMemoryOff();<br>
>> >>>> dd->Update();<br>
>> >>>> vtkPieceScalars *ps = vtkPieceScalars::New();<br>
>> >>>> ps->SetInputConnection(dd->GetOutputPort());<br>
>> >>>> ps->SetScalarModeToCellData();<br>
>> >>>> vtkDataSetSurfaceFilter *dss = vtkDataSetSurfaceFilter::New();<br>
>> >>>> dss->SetInputConnection(ps->GetOutputPort());<br>
>> >>>><br>
>> >>>> The dd object works fine and writing its contents out on each<br>
>> >>>> processor<br>
>> >>>> gives nprocs worth of meshes, each of which look slightly different<br>
>> >>>> to the<br>
>> >>>> way I've partitioned them up, but sum to the same serial mesh so I am<br>
>> >>>> happy<br>
>> >>>> with that working correctly. But I can't for the life of me figure<br>
>> >>>> out how<br>
>> >>>> to obtain local to global cell mappings, allocate ghost cells, or<br>
>> >>>> work out<br>
>> >>>> how to exchange data given the above partition info and comms....<br>
>> >>>><br>
>> >>>> Note I have not provided any additional information to "dd" regarding<br>
>> >>>> global cells as per the doxy notes so I assume it went away and<br>
>> >>>> computed it.<br>
>> >>>> I can't figure out how to extract it however. I also have no idea<br>
>> >>>> how to<br>
>> >>>> modify each local processor rgrid to include the ghost cells for that<br>
>> >>>> processor. Finally given that info, I could exchange between<br>
>> >>>> processors to<br>
>> >>>> write to each local processors ghost cells the corresponding "real"<br>
>> >>>> cell<br>
>> >>>> data from the neighbouring meshes and continue the code.<br>
>> >>>><br>
>> >>>> Any help really appreciated!<br>
>> >>>><br>
>> >>>> Cheers,<br>
>> >>>> Andy<br>
>> >>>><br>
>> >>><br>
>> >>><br>
>> >>><br>
>> >>> --<br>
>> >>><br>
>> >>> __________________________________<br>
>> >>><br>
>> >>> Dr Andrew Parker<br>
>> >>><br>
>> >>> Em@il: <a href="mailto:andrew.parker@cantab.net" target="_blank">andrew.parker@cantab.net</a><br>
>> >>><br>
>> >>><br>
>> >><br>
>> >><br>
>> >><br>
>> >> --<br>
>> >><br>
>> >> __________________________________<br>
>> >><br>
>> >> Dr Andrew Parker<br>
>> >><br>
>> >> Em@il: <a href="mailto:andrew.parker@cantab.net" target="_blank">andrew.parker@cantab.net</a><br>
>> >><br>
>> >><br>
>> ><br>
>> ><br>
>> ><br>
>> > --<br>
>> ><br>
>> > __________________________________<br>
>> ><br>
>> > Dr Andrew Parker<br>
>> ><br>
>> > Em@il: <a href="mailto:andrew.parker@cantab.net" target="_blank">andrew.parker@cantab.net</a><br>
>> ><br>
>> ><br>
>> ><br>
>> > _______________________________________________<br>
>> > Powered by <a href="http://www.kitware.com" target="_blank">www.kitware.com</a><br>
>> ><br>
>> > Visit other Kitware open-source projects at<br>
>> > <a href="http://www.kitware.com/opensource/opensource.html" target="_blank">
http://www.kitware.com/opensource/opensource.html</a><br>
>> ><br>
>> > Please keep messages on-topic and check the VTK FAQ at:<br>
>> > <a href="http://www.vtk.org/Wiki/VTK_FAQ" target="_blank">http://www.vtk.org/Wiki/VTK_FAQ</a><br>
>> ><br>
>> > Follow this link to subscribe/unsubscribe:<br>
>> > <a href="http://www.vtk.org/mailman/listinfo/vtkusers" target="_blank">http://www.vtk.org/mailman/listinfo/vtkusers</a><br>
>> ><br>
><br>
><br>
><br>
><br>
> --<br>
><br>
> __________________________________<br>
><br>
> Dr Andrew Parker<br>
><br>
> Em@il: <a href="mailto:andrew.parker@cantab.net" target="_blank">andrew.parker@cantab.net</a><br>
><br>
><br>
</div>
</div>
</blockquote>
</div>
<br>
<br clear="all">
<div><br>
</div>
-- <br>
<p>__________________________________</p>
<p> Dr Andrew Parker</p>
<p> <a href="mailto:Em@il" target="_blank">Em@il</a>: <a href="mailto:andrew.parker@cantab.net" target="_blank">
andrew.parker@cantab.net</a></p>
<br>
</div>
</div>
</div>
</blockquote>
</div>
</div>
</span></div>
</blockquote>
</div>
<br>
<br clear="all">
<div><br>
</div>
-- <br>
<p>__________________________________</p>
<p> Dr Andrew Parker</p>
<p> <a href="mailto:Em@il" target="_blank">Em@il</a>: <a href="mailto:andrew.parker@cantab.net" target="_blank">
andrew.parker@cantab.net</a></p>
<br>
</div>
</div>
</div>
</blockquote>
</div>
</div>
</span></div>
</blockquote>
</div>
<br>
<br clear="all">
<div><br>
</div>
-- <br>
<p>__________________________________</p>
<p> Dr Andrew Parker</p>
<p> <a href="mailto:Em@il" target="_blank">Em@il</a>: <a href="mailto:andrew.parker@cantab.net" target="_blank">
andrew.parker@cantab.net</a></p>
<br>
</div>
</div>
</blockquote>
</body>
</html>