Project Parallel in-Situ Visualization

Parallel in-Situ Visualization of Adaptive Mesh Data

Conforming Interpolation for Raytracing

In collaboration with Chris R. Johnson’s group at the SCI Institute, USA, we have developed an interpolation method to compute a continuous interpolation of element-centered data on recursively refined hexahedral meshes. It is termed “General Trilinear Interpolation,” GTI, and outperforms existing methods in terms of the visual quality of isosurfaces.

We have rendered large data sets out of the simulation of the airflow around an aircraft’s landing gear with this new method; see our 2020 Eurographics paper for details.

The Radiosity Method

As one part of our parallel visualization reasearch we develop algorithms that implement the radiosity method on parallel adaptive octree meshes. For the adaptive octrees we use the implementation provided by the software libary p4est.

For an efficient implementation of the radiosity method on parallel adaptive octrees, we exploit a specific tree traversal algorithm, which is covered in the paper Parallel Tree Algorithms for AMR and Non-Standard Data Access. The radiosity method involves computing the pairwise light transfer between surface patches in a scene, which we determine by enforcing strict local radiation balance.

The top-down search algorithm makes it possible to speed up these computationally demanding calculations by recursively excluding non-visible surface patches.

The load balance of the surface patches of a given scene is done by a weighted Morton space-filling curve partition and the parallelization in general makes use of the properties of the Morton space-filling curve.

As a next step we develop a multigridding-like approach to accelerate the solver and in general design the algorithm in a more hierarchical fashion.

(Project image courtesy of T. Griesbach, INS.)