VTK/Executives

Streaming of pieces, extents and time steps and multi-block support in VTK:

Executives:

vtkExecutive: General purpose superclass. Stores 1 algorithm, input and output information. Does not implement any pipeline passes.

vtkDemandDrivenPipeline: Implements REQUEST_DATA_OBJECT, REQUEST_INFORMATION and REQUEST_DATA passes.

vtkStreamingDemandDrivenPipeline: Implements REQUEST_UPDATE_EXTENT and REQUEST_UPDATE_EXTENT_INFORMATION (used for dynamic streaming) passes. Introduces concept of extents for structured datasets and piece for unstructured datasets and time steps. Support for extent translator.

Talk about streaming Talk about parallel processing

Streaming in filters (vtkTemporalStatistics - CONTINUE_EXECUTING()), streaming outside filters (vtkPolyDataMapper, looping at the sink)

vtkCompositeDataPipeline: Adds support for iterating over time steps and multiple blocks:

• If a filter can only handle single block datasets (e.g. vtkDataSet, vtkGraph, ...), this executive can iterate over the blocks of a multi-block dataset and invoke the filter for each block
• If a filter can only handle one time step and the input has multiple time step, this executive can iterate over the time steps and invoke the filter for each
• If a reader can only provide one time step at a time and downstream requested multiple time steps, this executive can ask the reader to read each time step and collect the result in a temporal dataset