- Jan 06, 2017
-
-
Sean McBride authored
-
c63e304e add generic report capability code into OpenGLRenderWindow Acked-by:
Kitware Robot <kwrobot@kitware.com> Reviewed-by:
David Lonie <david.lonie@kitware.com> Merge-request: !2322
-
81a5b257 try another fix for tsan condition variable Acked-by:
-
c51f6a84 Add and use ReleaseGraphicsResources to OGLRenderTimer. Acked-by:
-
600e728d Implement vtkPolygonalMeshPointNormals Acked-by:
Ken Martin <ken.martin@kitware.com> Merge-request: !2271
-
143bb211 add two more valid images for gl2ps failures Acked-by:
-
Ken Martin authored
I think the old code was fine but tsan can have false positives with some atomic types and race conditions so try another fix
-
f0729750 Fix NULL socket crash Acked-by:
Joachim Pouderoux <joachim.pouderoux@kitware.com> Merge-request: !2318
-
- Jan 05, 2017
-
-
0f2b56e4 Prefer GetIncrements(incs) over getting a pointer Acked-by:
-
Alexis Girault authored
vtkTriangleMeshPointNormals is a filter that computes point normals for a triangle mesh to enable high-performance rendering. It is a fast-path version of the vtkPolyDataNormals filter in order to be able to compute normals for triangle meshes deforming rapidly. The computed normals (a vtkFloatArray) are set to be the active normals (using SetNormals()) of the PointData. The array name is "Normals", so they can be retrieved either with `output->GetPointData()->GetNormals()` or with `output->GetPointData()->GetArray("Normals")`. The algorithm works by determining normals for each triangle and adding these vectors to the triangle points. The resulting vectors at each point are then normalized. @warning Normals are computed only for triangular polygons: the filter can not handle meshes with other types of cells (Verts, Lines, Strips) or Polys with the wrong number of components (not equal to 3). @warning Unlike the vtkPolyDataNormals filter, this filter does not apply any splitting nor checks for cell orientation consistency in order to speed up the computation. Moreover, normals are not calculated the exact same way as the vtkPolyDataNormals filter since the triangle normals are not normalized before being added to the point normals: those cell normals are therefore weighted by the triangle area. This is not more nor less correct than normalizing them before adding them, but it is much faster. @sa If you do not need to do high-performance rendering, you should use vtkPolyDataNormals if your mesh is not only triangular, if you need to split vertices at sharp edges, or if you need to check that the cell orientations are consistent to flip inverted normals. @sa If you still need high-performance rendering but your input polydata is not a triangular mesh and/or does not have consistent cell orientations (causing inverted normals), you can still use this filter by using vtkTriangleFilter and/or vtkCleanPolyData respectively beforehand. If your mesh is deforming rapidly, you should be deforming the output mesh of those two filters instead in order to only run them once. This filter was implemented for a real-time simulation application where a mesh vertices are deforming at every frame. With 110k points, the framerates were the following: - 100 fps with no normals computation - 5 fps with vtkPolyDataNormals with default options - 15 fps with vtkPolyDataNormals with SplittingOff & ConsistencyOff - 80 fps with vtkPolygonalMeshPointNormals This is an improvement of 5 to 16 times faster for that use case. Note: for high-performance rendering of deformable meshes, two other improvements can be made to boost the performance: - skip frustrum culling ``` renderer->RemoveCuller(renderer->GetCullers()->GetLastItem()); ``` - skip auto shift & scale of VBO: ``` openglmapper->SetVBOShiftScaleMethod(vtkOpenGLVertexBufferObject::DISABLE_SHIFT_SCALE); ``` Those two changes allow to skip calling `GetBounds()`, which requires to recompute the bounds at each frame. This brings the framerate from 80 fps to 200 fps. -
ba4813f2 fix unsafe memory write in vtkImageGradientMagnitude etc Acked-by:
-
David Gobbi authored
The GetIncrements() method calls ComputeIncrements(), and if you do not supply your own array, then the vtkImageData::Increments ivar is modified and returned. This means that the signature of ComputeIncrements() that returns a pointer is not thread safe.
-
David C. Lonie authored
-
17d4747e Restore FBO bindings after SurfaceLIC rendering. Acked-by:
-
70bd204f Fix off-by-one error in TestGLS2PSTextOpacity 2c51d6fe Prefer prefix increment/decrement on iterators Acked-by:
-
David Gobbi authored
This error was found by cppcheck.
-
David Gobbi authored
This fixes performance warnings for cppcheck.
-
Ken Martin authored
Some subclasses did not have it implemented and it makes sense to provide a generic implementation by default.
-
Ken Martin authored
Two more images that appear to be driven by pixel level size changes
-
Ken Martin authored
The class was writing to a memory location from multiple threads bro. Also fix thread issues in the TestAtomic and TestConditionVariable tests.
-
d461040c fix type conversion warnings on new dashboard machine Acked-by:
-
07d84f48 fix CMP0020 issue for Examples/Widgets Acked-by:
-
Ken Martin authored
fixing dashboard issues
-
Mathieu Westphal (Kitware) authored
In some cases, when server closes unexpectendly the client may try to send data to the server while it's socket has already been destroyed, provoking a segfault. This fixes it.
-
- Jan 04, 2017
-
-
4611c773 added some more gl2ps valid images Acked-by:
-
Ken Martin authored
new versions of cmake warn on no setting
-
Ken Martin authored
seems to be a fair bit of variablility on systems
-
4ed3f155 fix failing header tests for pLagrangianParticalTrackker Acked-by:
-
6b9b8c8e Use properly sized MAX values to force newer mtime. ddd40c22 Add min/max defines for vtkMTimeType. Acked-by:
-
David C. Lonie authored
Refs paraview/paraview#17088.
-
9a46ad06 Dan's revert missed this chunk of code Acked-by:
-
David C. Lonie authored
MTimes are more complicated now than they were when this code was written, and may be different types.
-
David C. Lonie authored
These follow the naming convention used for vtkIdType. Also added the missing min/max defines for the vtkType[type][size] typedefs, as they were needed for some of the MTime implementations.
-
363f2155 Fix TubeFilter when generating texture coordinates Acked-by:
Mathieu Westphal <mathieu.westphal@kitware.com> Reviewed-by:
Cory Quammen <cory.quammen@kitware.com>
Merge-request: !2219 -
Ken Martin authored
was missing comment and VTK_DELETE_FUNCTION
-
4e5c7f67 Fixed null dereference warned by cppcheck Acked-by:
-
a8e1ef08 Fixed clang -Wstrict-prototypes warnings Acked-by:
-
43ce2ed0 Correcting vtkCutter progress Acked-by:
-
32b5c8bc fix a compiler warning Acked-by:
-
- Jan 03, 2017
-
-
Sean McBride authored
Strictly speaking, in C, unlike C++, one must use void for an empty parameter list.
-
