Implementing D.S.I. On Arbitrary Graphs using Dart Technology

Richard Cognot and Bruno Levy. ( 1997 )
in: 16th gOcad Meeting (Dallas), ASGA

Abstract

This paper summarises the recent advancements in the new topology engine based on Darts, stemmed by a close collaboration with the University of Strasbourg (France). One of our main concerns with this new topology is that a lot of the information is inferred, that is, deduced from a general scheme. Hence, as opposed to the classical topology engine used in the current version of GOCAD, no direct information about, for example, the satellites of a node is known a priori. This has advantages and drawbacks. Among the advantages, nodes, segments, triangles, among others, are known as a single type of entity: cells. Nodes are O-cells, segments 1-cells, and so on. Hence, defining a neighbourhood on cells allows the same code to perform on either nodes or segments or polygons or even polyhedra. The internal topological data base is also much smaller, as less redundancy is used. The main drawback, especially in the case of a crucial process like D.S.I, is performance. Some algorithms will of course perform better on the new topology, in particular algorithms requiring only local knowledge of the topology in order to modify it, as they will have less information to update. But D.S.I. uses all the topological information, without ever modifying it. To make things faster, we have implemented a CellCache mechanism. This CellCache contains all the neighbourhood information in a way that makes it transparent to the fact that we are dealing with cells of whatever dimension, and is filled on demand.

Download / Links

    BibTeX Reference

    @inproceedings{CognotRM1997c,
     abstract = { This paper summarises the recent advancements in the new topology engine based on Darts, stemmed by a close collaboration with the University of Strasbourg (France). One of our main concerns with this new topology is that a lot of the information is inferred, that is, deduced from a general scheme. Hence, as opposed to the classical topology engine used in the current version of GOCAD, no direct information about, for example, the satellites of a node is known a priori. This has advantages and drawbacks. Among the advantages, nodes, segments, triangles, among others, are known as a single type of entity: cells. Nodes are O-cells, segments 1-cells, and so on. Hence, defining a neighbourhood on cells allows the same code to perform on either nodes or segments or polygons or even polyhedra. The internal topological data base is also much smaller, as less redundancy is used. The main drawback, especially in the case of a crucial process like D.S.I, is performance. Some algorithms will of course perform better on the new topology, in particular algorithms requiring only local knowledge of the topology in order to modify it, as they will have less information to update. But D.S.I. uses all the topological information, without ever modifying it. To make things faster, we have implemented a CellCache mechanism. This CellCache contains all the neighbourhood information in a way that makes it transparent to the fact that we are dealing with cells of whatever dimension, and is filled on demand. },
     author = { Cognot, Richard AND Levy, Bruno },
     booktitle = { 16th gOcad Meeting (Dallas) },
     month = { "november" },
     publisher = { ASGA },
     title = { Implementing D.S.I. On Arbitrary Graphs using Dart Technology },
     year = { 1997 }
    }