Hierarchical indexing for out-of-core access to multi-resolution data

Increases in the number and size of volumetric meshes have popularized the use of hierarchical multi-resolution representations for visualization. A key component of these schemes has become the adaptive traversal of hierarchical data-structures to build, in real time, approximate representations of the input geometry for rendering. For very large datasets this process must be performed out-of-core. This paper introduces a new global indexing scheme that accelerates adaptive traversal of geometric data represented as binary trees by improving the locality of hierarchical/spatial data access. Such improvements play a critical role in the enabling of effective out-of-core processing. Three features make the scheme particularly attractive: (i) the data layout is independent of the external memory device blocking factor, (ii) the computation of the index for rectilinear grids is implemented with simple bit address manipulations and (iii) the data is not replicated, avoiding performance penalties for dynamically modified data. The effectiveness of the approach was tested with the fundamental visualization technique of rendering arbitrary planar slices. Performance comparisons with alternative indexing approaches confirm the advantages predicted by the analysis of the scheme.