Parallel Peeling of Curvilinear Grids

In this paper we present a novel hybrid CPU-GPU approach for rendering curvilinear grids. Visibility sorting is accomplished by parallel peeling cells off the grid, utilizing an active cell peeling front. In each step, we compute the ray-cell intersection coordinates on the GPU, perform accurate volume integration (CPU), and determine the set of active cells for the next iteration (GPU). The approach requires only standard graphics capabilities and can therefore be used on any commodity PC, including laptops. Furthermore, the main memory requirements are negligible since the required data structures are minimal. The main advantage of our algorithm is that we exploit hardware acceleration for the expensive visibility sorting which is beneficial over time due to the faster performance increase of GPUs over CPUs. Due to the simplicity of the algorithm and its low requirements on preprocessing and main memory, it is well suited for thin clients. Last but not least, the approach could easily be extended to irregular grids using tetrahedra. CR Categories: I.3.7 [Computer Graphics]: Three-Dimensional Graphics and Realism—Raytracing;