Employing Complex GPU Data Structures for the Interactive Visualization of Adaptive Mesh Refinement Data

Abstract

We present a framework for interactively visualizing volumetric Adaptive Mesh Refinement (AMR) data. For this purpose we employ complex data structures to map the entire AMR dataset to graphics memory. This allows to apply hardware accelerated visualization algorithms previously only operating on uniform cartesian grids. For mapping the data to graphics memory we consider two approaches, a space-efficient, texture-based octree and a fast method based on page table address translation. We demonstrate the utility of our approach by extending an existing GPU raycasting implementation to render AMR data. As a first step to vector field visualization techniques we successfully integrated our framework into a commercial CFD postprocessing tool and visualized scalar properties of the vector field.