Caching of AMR blocks (geometry+data) on AMR reader
This issue was created automatically from an original Mantis Issue. Further discussion may take place here.
The demand-driven AMR infrastructure, implemented in vtkAMRBaseReader, supports upstream requests from the pipeline that specify which blocks are to be loaded. However, when the user performs actions within the paraview user-interface that can trigger multiple such requests, it can lead to an overwhelming amount of I/O for multiple requests of the same data. One solution to minimize the I/O and avoid reading data that has already been read is to cache AMR blocks in a separate data-structure, e.g., a vtkAMRDataSetCache, that will be an ivar of the vtkAMRBaseReader.