GPU Accelerated Isosurface Extraction on Complex Polyhedral Grids.

in: Proc. 26th Gocad Meeting, Nancy

Abstract

Progress in numerical modeling of physical processes has opened the way to new types of simulation grids made of complex polyhedral cells. Visualizing scalar fields on such grids without increasing the number of cells calls for a generalization of existing isosurface extraction algorithms. We present a method to efficiently tessellate on a GPU tetrahedral, hexahedral and hybrid grids. This method also works for strongly heterogeneous grids, with arbitrary polyhedral cells. It avoids data redundancy and processes efficiently large grids made of millions of cells. To achieve this, textures are used to store most of the needed data and are accessed through vertex texture lookup in the vertex shading unit of modern graphics cards. Results show that this method is faster than the same CPU-based extraction. In the case of tetrahedral grids, our method is complementary with previous approaches based on GPU registers: it is less efficient for small grids, but allows the storage of millions-tetrahedra grids in graphics memory, which was impossible with previous works.

Download / Links

    BibTeX Reference

    @INPROCEEDINGS{P23_Buatois,
        author = { Buatois, Luc and Caumon, Guillaume and Levy, Bruno },
         title = { GPU Accelerated Isosurface Extraction on Complex Polyhedral Grids. },
     booktitle = { Proc. 26th Gocad Meeting, Nancy },
          year = { 2006 },
      abstract = { Progress in numerical modeling of physical processes has opened the way to new types of simulation
    grids made of complex polyhedral cells. Visualizing scalar fields on such grids without
    increasing the number of cells calls for a generalization of existing isosurface extraction algorithms.
    We present a method to efficiently tessellate on a GPU tetrahedral, hexahedral and hybrid grids.
    This method also works for strongly heterogeneous grids, with arbitrary polyhedral cells. It avoids
    data redundancy and processes efficiently large grids made of millions of cells. To achieve this,
    textures are used to store most of the needed data and are accessed through vertex texture lookup
    in the vertex shading unit of modern graphics cards. Results show that this method is faster than
    the same CPU-based extraction. In the case of tetrahedral grids, our method is complementary
    with previous approaches based on GPU registers: it is less efficient for small grids, but allows the
    storage of millions-tetrahedra grids in graphics memory, which was impossible with previous works. }
    }