Keywords: parallel processing, CAD modeler, parametric surface meshing, multithread strategy, curve discretization, anisotropic meshing, surface curvatures.

polynomial or rational parametric patches (NURBS) as is the case for most CAD modelers. There are, essentially, two approaches to meshing parametric surfaces: direct and indirect. Popular direct meshing methods include the octree-based method, the advancing-front-based method and the paving-based method. All these methods work directly in the tridimensional space. On the other hand, the indirect meshing approach consists in meshing the bidimensional parametric domain and mapping the resulting mesh onto the surface. It is conceptually straightforward as a bidimensional mesh is generated in the parametric domain and thus it is expected to be faster than the direct approach. Notice that the bidimensional mesh of the parametric domain is generally anisotropic. Indeed, the specified size map in three dimensions (which may be isotropic or anisotropic) is induced in the parametric domains using the first fundamental metric of the surface which is generally anisotropic. Using the latter approach, we have proposed a general scheme for meshing conformal composite parametric surfaces. The scheme is constituted by sequential procedures which consist of discretizing each interface curve, projecting the discretizations on boundaries of parametric domains, meshing each parametric domain according to the above boundary discretizations and finally mapping these two-dimensional meshes onto the surface. Complex surfaces such as a car engine or a complete aircraft are composed of thousands of patches, and meshing these surfaces using the above sequential scheme can be inefficient. However, it can be noticed that each curve discretization can be created independently, as well as each domain mesh. After a brief presentation of the two principal parallel computing methodologies, namely distributed memory and shared memory architectures, a parallel version of the general meshing scheme using a multithread strategy is proposed. This approach naturally balances the load to each processor and is particularly efficient on present multicore computers. This parallel methodology has been implemented in the BLSURF surface meshing software. Numerical examples are given showing the cost reduction compared with the sequential version. To conclude, some remarks are provided about the integration of the parallel meshing methodology in a CAD system. In particular, the efficiency of the method depends on the parallel evaluation of the mapping function with respect to parametric domains.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £85 +P&P)