[vtkusers] is whole extent required for multiblock dataset?
Ufuk Turuncoglu
ufuk.turuncoglu at itu.edu.tr
Fri Jun 2 15:14:38 EDT 2017
Hi,
Actually, i am trying to create simplified case to test some ideas to
improve the performance of my Catalyst application. In this case, i
would like to create C++ code one for point data in vtkMultiBlockDataSet
and other is cell data again in vtkMultiBlockDataSet. If i manage to
create these two cases and ghost regions to have correct representation
of the entire field, that would be great. As i know, for point data, i
need to create ghost region (or overlapped regions) for point data but
not for cell ones. Is it correct? Is there any other way to escape
generation of ghost regions (both for point and cell data)? It is not
easy to find any example for the cases like these. So, i am looking for
an answer in the user lists.
By the way, i could not see any "off-by-1" bug in the point coordinates.
Is it possible to give more information about it. Thanks.
Regards,
--ufuk
On 02/06/2017 20:16, Andy Bauer wrote:
> Hi Ufuk,
>
> If there is one vtkStructuredGrid per multi-block you don't need the
> whole extent. Using a vtkStructuredGrid in a multiblock though will
> result in each grid block to be considered a separate grid so if you
> do something like extracting the surface it probably won't look
> correct unless you generate your own ghost cells (the ghost cell
> generator filter in PV won't work).
>
> I only briefly looked through your code snippet but I suspect there's
> an "off-by-1" bug in the way you're setting your point coordinates for
> the structured grids but once that's fixed it will work the way you
> want it to.
>
> I probably don't know your full use case but I would suggest for
> non-Catalyst stuff to just avoid the multiblock here and just generate
> structured grid pieces.
>
> Best,
> Andy
>
> On Fri, Jun 2, 2017 at 8:13 AM, Ufuk Turuncoglu
> <ufuk.turuncoglu at itu.edu.tr <mailto:ufuk.turuncoglu at itu.edu.tr>> wrote:
>
> Hi Again,
>
> I create following test code to write dummy data (checkerboard
> with 0 and 1) using vtkStructuredGrid in multiblock format.
> Without extent in left and top of the tiles the data looks correct
> but there is a gap between pieces (see attached screenshot). To
> fix the issue, i add a extra piece of code to overlap one row and
> columns (using hasLeft and hasTop control). The problem is that
> when i run the code with this modification, i am getting totally
> wrong grid representation (screen2). I just wonder that do i need
> to set whole extent in somewhere. I tried to add following to set
> whole extent to multiblock dataset,
>
> int extent[6] = {0, N-1, 0, N-1, 0, 0};
> mb->GetInformationObject(0)->Set(vtkStreamingDemandDrivenPipeline::WHOLE_EXTENT(),
> extent, 6);
>
> but it won't work as i expected. Any suggestion? Thanks.
>
> Regards,
>
> --ufuk
>
>
> #include <iostream>
> #include "mpi.h"
> #include "vtkStructuredGrid.h"
> #include "vtkSmartPointer.h"
> #include "vtkMultiBlockDataSet.h"
> #include "vtkMultiPieceDataSet.h"
> #include "vtkNew.h"
> #include "vtkXMLPMultiBlockDataWriter.h"
> #include "vtkMPIController.h"
> #include "vtkDoubleArray.h"
> #include "vtkDataSet.h"
> #include "vtkPointData.h"
> #include "vtkInformation.h"
> #include "vtkStreamingDemandDrivenPipeline.h"
>
> using namespace std;
>
> // global variables
> int const N = 10;
> int const tileX = 2;
> int const tileY = 2;
>
> int main(int argc, char** argv) {
> // initialize MPI
> MPI_Init(NULL, NULL);
>
> // setup global controller, true means initialized externally
> vtkNew<vtkMPIController> controller;
> controller->Initialize(&argc, &argv, true);
> vtkMultiProcessController::SetGlobalController(controller.Get());
>
> // get number of processor and rank
> int nproc, myid;
> MPI_Comm_size(MPI_COMM_WORLD, &nproc);
> MPI_Comm_rank(MPI_COMM_WORLD, &myid);
>
> // create 2d decomposition parameters
> bool hasLeft = false, hasTop = false;
> if (myid%tileX == 0) hasLeft = true;
> if (myid/tileY == 0) hasTop = true;
>
> int istr = myid%tileX*N/tileX;
> int iend = istr+N/tileX-1;
> if (hasLeft) iend = iend+1;
> int jstr = myid/tileX*N/tileY;
> int jend = jstr+N/tileY-1;
> if (hasTop) jend = jend+1;
> int size = (iend-istr+1)*(jend-jstr+1);
> //cout << myid << " of " << nproc << " " << boolalpha << hasLeft
> << " " << hasTop << endl;
> //cout << myid << " " << istr << " " << iend << " " << jstr << "
> " << jend << endl;
>
> // create structured grid
> vtkSmartPointer<vtkStructuredGrid> grid =
> vtkSmartPointer<vtkStructuredGrid>::New();
>
> // define grid coordinates
> vtkSmartPointer<vtkPoints> points =
> vtkSmartPointer<vtkPoints>::New();
> points->SetNumberOfPoints(size);
>
> int id = 0;
> for (int i = istr; i <= iend; i++) {
> for (int j = jstr; j <= jend; j++) {
> points->SetPoint(id, i, j, 0);
> id = id+1;
> }
> }
>
> grid->SetPoints(points);
> grid->SetExtent(istr, iend, jstr, jend, 0, 0);
> points->Delete();
>
> // add field, checkerboard type data with full of 0/1
> vtkSmartPointer<vtkDoubleArray> field =
> vtkSmartPointer<vtkDoubleArray>::New();
> field->SetName("dummy");
> field->SetNumberOfComponents(1);
> field->SetNumberOfValues(size);
>
> id = 0;
> for (int i = istr; i <= iend; i++) {
> for (int j = jstr; j <= jend; j++) {
> int val = (i+j)%2 ? 0 : 1;
> field->SetValue(id, val);
> id = id+1;
> }
> }
>
> // create multi block grid
> vtkMultiBlockDataSet* mb = vtkMultiBlockDataSet::New();
> vtkNew<vtkMultiPieceDataSet> mpds;
> mpds->SetNumberOfPieces(nproc);
> mpds->SetPiece(myid, grid.GetPointer());
> mb->SetNumberOfBlocks(1);
> mb->SetBlock(0, mpds.GetPointer());
>
> // add field
> vtkDataSet *ds = vtkDataSet::SafeDownCast(mpds->GetPiece(myid));
> ds->GetPointData()->AddArray(field);
>
> // create a writer
> vtkSmartPointer<vtkXMLPMultiBlockDataWriter> mbw =
> vtkSmartPointer<vtkXMLPMultiBlockDataWriter>::New();
> mbw->SetFileName("test.vtm");
> mbw->SetDataModeToAscii();
> mbw->SetInputData(mb);
> mbw->Write();
>
> // finalize the MPI environment.
> MPI_Finalize();
>
> return 0;
> }
>
> _______________________________________________
> Powered by www.kitware.com <http://www.kitware.com>
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
> <http://www.kitware.com/opensource/opensource.html>
>
> Please keep messages on-topic and check the VTK FAQ at:
> http://www.vtk.org/Wiki/VTK_FAQ <http://www.vtk.org/Wiki/VTK_FAQ>
>
> Search the list archives at:
> http://markmail.org/search/?q=vtkusers
> <http://markmail.org/search/?q=vtkusers>
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/vtkusers
> <http://public.kitware.com/mailman/listinfo/vtkusers>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtkusers/attachments/20170602/6fbb2286/attachment.html>
More information about the vtkusers
mailing list