vtkDIYKdTreeUtilities now supports creating an assigner that preserves
the kd-tree across ranks even if the number of ranks is not exactly a
power of two.

vtkDIYKdTreeUtilities was not using correct data bounds when
cell-centers were being requested. Fixed that.

Also removed the code to pad the bounding box. Algorithms should pass in
padded bounds if that's what is expected.

The vtkBandedPolyDataContourFilter::ClipEdge method interpolates between
two edge points, creating extra points at the intermediate clip values,
if any. The method retrieves the clip values for the edge points. The
clip value for the highest scalar value may be somewhat larger than the
end point scalar value due to the application of the internal clip
tolerance. When the difference between the end point values is on the
order of this clip tolerance this may result in an interpolation factor
significantly larger than 1. The effect of this overshoot is that spikes
appear extending outside the original cell.

This issue is fixed by excluding the high end point clip value if the
resulting interpolation factor is significantly larger than 1.

Also separate out the parallel functionality and move it into a subclass,
vtkPMergeArrays.

ThreadedData
Modernizing TestLagrangianParticle
Better point data implementation
Better caching
Fix an error in the way sendStream were managed
Fix empty output bug

This ensures field data is passed through
vtkPUnstructuredGridGhostCellsGenerator.

Modify test to check that field data is passed to output.

vtkExtractCells now shallow copies points, if possible thus avoiding
an expensive copy if input dataset is indeed a point-set.

vtkExtractCells changed the cell ids filled by the user in its
RequestData. This was incorrect. It had the potential of dropping cell
ids provided is the input dataset had fewer elements for an earlier
iteration.

Adding API to vtkExtractCells to skip sorting and uniquifying operations
if its already known that the ids are sorted and unique.

vtkExtractCells now has a mode to extract all cells. When set, the
filter avoid building and updating internal cell id list which can be
time consuming.

vtkExtractSelection now uses shortcuts when extracting all or no cells.

vtkExtractCells is tuned for cell extraction.

vtkGenerateGlobalIds needed a check to handle case where there was only 1
block of non-empty data. Added that. Also updated the test.

Addresses paraview/paraview#19638.

This should quiet output when vtk3DLinearGridPlaneCutter executes.

It can be used only when boundary cells alone are extracted.

Yohann Bearzi (Kitware) authored 5 years ago and Cory Quammen committed 5 years ago

(cherry picked from commit 04995e0f)

Yohann Bearzi (Kitware) authored 5 years ago and Cory Quammen committed 5 years ago

(cherry picked from commit 3444bb8d)

Added new threaded methods in vtkPolyData and vtkBoundingBox
for computing GetBounds(). Added a comparison test for the
methods vtkPolyData::GetBounds(), vtkPoints::GetBounds(),
and vtkBoundingBox::ComputeBounds(). The speed improvements on
large meshes is approximately 10x.

Also, the new vtkCellArray API, and the vtkArrayDispatch
framework is used. Note, as before, this new vtkPolyData::GetBounds()
only considers points used by cells to determine the bounding box.

Performance considerations required tuning the vtkCellArray and
vtkCellArrayIterator to share memory when possible (i.e., avoid
copying). Also cleaned up documentation and API for these classes.
A test was added to ensure that iterator is functioning properly
(e.g., is faster, sharing storage buffer as appropriate).

Finally discovered an error in vtkParametricKuen which was
generating NaN on some systems, which messed up the GetBounds()
computation.

Robert Maynard authored 5 years ago and T.J. Corona committed 5 years ago

In C++11 the return type of a lambda can be only inferred if the
lambda is a single statement with an explicit `return`. So for
functions that have no return, we have to explicit state that
they don't have a return.

(cherry picked from commit 14e996fe)

Robert Maynard authored 5 years ago and T.J. Corona committed 5 years ago

By moving to a manual for loop we can work around this compiler
limitation

(cherry picked from commit a89637aa)

In C++11 the return type of a lambda can be only inferred if the
lambda is a single statement with an explicit `return`. So for
functions that have no return, we have to explicit state that
they don't have a return.

By moving to a manual for loop we can work around this compiler
limitation

Simple fix to avoid divide by zero

bunch of missing override qualifiers on destructors

This reverts commit a7800961, reversing
changes made to f89fc4ea.

Updating `vtkPipelineSize` API to avoid dependency on mapper and hence
on `VTK::RenderingCore`.

vtkLineSource now support ability to manually specify the locations
along the line segment(s) where to add intermediate sample points.

This is needed to addresses paraview/paraview#19236.