vtk-m/docs/changelog/split-contour-filters.md

# Split flying edges and marching cells into separate filters

The contour filter contains 2 separate implementations, Marching Cells and Flying Edges, the latter only available if the input has a `CellSetStructured<3>` and `ArrayHandleUniformPointCoordinates` for point coordinates. The compilation of this filter was lenghty and resource-heavy, because both algorithms were part of the same translation unit.

Now, this filter is separated into two new filters, `ContourFlyingEdges` and `ContourMarchingCells`, compiling more efficiently into two translation units. The `Contour` API is left unchanged. All 3 filters `Contour`, `ContourFlyingEdges` and `ContourMarchingCells` rely on a new abstract class `AbstractContour` to provide configuration and common utility functions.

Although `Contour` is still the preferred option for most cases because it selects the best implementation according to the input, `ContourMarchingCells` is usable on any kind of 3D Dataset. For now, `ContourFlyingEdges` operates only on structured uniform datasets.

Deprecate functions `GetComputeFastNormalsForStructured`, `SetComputeFastNormalsForStructured`, `GetComputeFastNormalsForUnstructured` and `GetComputeFastNormalsForUnstructured`, to use the more general `GetComputeFastNormals` and `SetComputeFastNormals` instead.

By default, for the `Contour` filter, `GenerateNormals` is now `true`, and `ComputeFastNormals` is `false`.

The marching cubes version of contour still has several possible compile paths, so it can still take a bit to compile. To help manage the compile time further, the contour filter compilation is broken up even further using the instantiation build capabilities.
Split flying edges and marching cells into separate filters In order to compile the contour filter more efficiently, we split the contour filter into two separate translation units, corresponding to the new filters ContourFlyingEdges and ContourMarchingCells. The API for Contour filter is left totally unchanged, and tries to use flying edges if the dataset is structured and uniform. All three contour filters inherit from the `AbstractContour` class, providing utility methods used in the implementations. 2023-04-25 08:54:35 +00:00			`# Split flying edges and marching cells into separate filters`

			The contour filter contains 2 separate implementations, Marching Cells and Flying Edges, the latter only available if the input has a `CellSetStructured<3>` and `ArrayHandleUniformPointCoordinates` for point coordinates. The compilation of this filter was lenghty and resource-heavy, because both algorithms were part of the same translation unit.

			Now, this filter is separated into two new filters, `ContourFlyingEdges` and `ContourMarchingCells`, compiling more efficiently into two translation units. The `Contour` API is left unchanged. All 3 filters `Contour`, `ContourFlyingEdges` and `ContourMarchingCells` rely on a new abstract class `AbstractContour` to provide configuration and common utility functions.

			Although `Contour` is still the preferred option for most cases because it selects the best implementation according to the input, `ContourMarchingCells` is usable on any kind of 3D Dataset. For now, `ContourFlyingEdges` operates only on structured uniform datasets.

			Deprecate functions `GetComputeFastNormalsForStructured`, `SetComputeFastNormalsForStructured`, `GetComputeFastNormalsForUnstructured` and `GetComputeFastNormalsForUnstructured`, to use the more general `GetComputeFastNormals` and `SetComputeFastNormals` instead.

Split the compilation of ContourMarchingCells into instantiations Although the contour filter was recently divided into 2 filters, flying edges and marching cubes, the marching cubes version still had many conditions and was the file that took the longest to compile on Frontier. To help speed up parallel compiles and prevent a single run of a compiler from being overwhelmed, the compilation of all the marching cubes conditions has been split up using instantiations. 2023-06-26 18:07:47 +00:00			By default, for the `Contour` filter, `GenerateNormals` is now `true`, and `ComputeFastNormals` is `false`.

			`The marching cubes version of contour still has several possible compile paths, so it can still take a bit to compile. To help manage the compile time further, the contour filter compilation is broken up even further using the instantiation build capabilities.`