high performance computing on graphics processing units: hgpu.org

In this paper, we present an adaptive multiresolution mesh representation exploring the computational differences of the CPU and the GPU. We build our representation considering a dense-polygon mesh simplified to a base mesh which stores the original geometry by means of an atlas structure. For both simplification and refinement processes, we present a hierarchical method based on stellar operators. During simplification, we compute local parametrizations to generate charts and an atlas structure to be used later in refinement. Unlike previous approaches, we employ the simplified mesh as our base domain in a novel atlas descriptor using a specialized halfedge data structure combined with our charts. Finally, we show that our mesh representation can be used to adaptively control the mesh resolution in CPU-GPU coupled applications, including mesh editing and visualization.