Contents • • • • • • • • • • • Motivation This document describes modifications to the Visualization ToolKit that allow data structures with arbitrary memory layouts to be used with a subset of the VTK pipeline. Supporting such data objects enables in-situ analysis of large simulations without the need to copy and repack the data into standard VTK arrays and data sets. For large simulations, this saves time and memory. The greatest obstacle to using a non-standard memory layout in a VTK array or dataset is that the implementations of VTK data structures are only weakly abstracted. For example,, the root of all data arrays in VTK, defines a GetVoidPointer method that returns the memory address of the first element in the array. Many algorithms in VTK rely on obtaining this pointer and assuming that the 2-dimensional data is contiguous in memory and follows a specific ordering.
Oiq, 0884, 03806,. XML Schema Documentation x509types.xsd.htm[11.5.2016 9:41:45] XML Schema Documentation. Table of Contents. Schema Document Properties Global Declarations.
Similarly,, the most flexible dataset type, exposes many details of its implementation in its API, making it impossible to store arbitrary representations of data topologies. Such issues prevent simulation data with alternate memory layouts from being properly handled by the VTK pipeline. The following describes techniques that allow arbitrary data layouts to be used with the and APIs with the goal of enabling in-situ visualization of active simulations.
Mapped vtkUnstructuredGrids A object represents a dataset consisting of points and cells. Cells hold the topology of the dataset, and unlike the other VTK datasets, vtkUnstructuredGrid makes no assumptions about the types of cells in the data. This flexibility makes it an ideal choice for a generic in-situ dataset, as it can losslessly store all other types of datasets.
Like vtkDataArray, the unstructured grid interface exposes much of its implementation. The point information is not a problem, as it is ultimately stored in a vtkDataArray and thus the in-situ dataset’s points can be easily handled by creating a subclass of vtkMappedDataArray. The cell information, however, is distributed across four arrays: • Connectivity (): Stores the connections between points cells as an array: [ n0, ptId1, ptId2., ptId(n0), n1, ptId1, ptId2., ptId(n1). ] where nX is the number of points in cell X, and the point ids that follow are the indices of the points in that cell. • Links (): Stores the cells to which each point belongs. • Types (): Type of cell (quad, triangle, line, etc).
• Locations (): Maps cell ID into Connectivity offset for random-access cell lookup. None of these arrays are fully encapsulated by the vtkUnstructuredGrid API, meaning that filters and algorithms can (and do) retrieve and operate on them directly. Thus, many filters depend on the implementation details of the dataset. A mapped version of a vtkUnstructuredGrid dataset could be created by using vtkMappedDataArrays to emulate the connectivity, types, and locations arrays, but this would not be a straightforward process. In particular, the connectivity array is difficult to map data onto, since the location of each cell is dependent on the size of every cell before it. A higher-level solution is needed.
Approach Rather than reuse the vtkDataArray solution of creating a mapped subclass that emulates the API of the VTK object, we introduce a superclass to vtkUnstructuredGrid that removes the implementation dependant methods. The class defines a set of pure virtual functions, taken from the vtkUnstructuredGrid API, that allow interaction with the cell information in a generic way.
Combined with the cell-related methods in, this provides a suitable level of abstraction for implementing a mapped dataset. Creating a Mapped vtkUnstructuredGridBase Subclass In-situ datasets should make use of the wrapper. Shejrhoi behtarin dar borai modar.
This is a class template that fulfills the vtkUnstructuredGridBase API by delegating to an implementation class. This wrapper is used by writing an implementation class that meets the requirements listed in the vtkMappedUnstructuredGrid class documentation, and then calling a macro from the family to create a new concrete class. See the examples section for a sample implementation. The template approach allows cheap shallow copying of the dataset. Unlike the mapped arrays, calling NewInstance() on a mapped dataset will return a new mapped dataset of the same type. Calling ShallowCopy(.) will only copy a reference-counted pointer to the implementation class, quickly making a low-overhead copy of the dataset. Tablica umnozheniya a4. Making Existing Code Compatible With vtkUnstructuredGridBase Updating existing code for compatibility with mapped datasets is more intrusive than the changes required for mapped data arrays.