Abstract
Solid modeling is a very useful industrial tool in the manufacture and design of industrial parts and assemblies. As a tool in the industrial workplace it has to be able to respond quickly to changes in design. To do this, the intersection algorithms between the solids have to be speeded up. Optimizations such as vector and parallel processing traditionally supported by supercomputers have the potential to solve this problem.
A solid modeler was developed based on the boundary representation approach using a half-edge data structure. Those parts of the solid modeler code that could be vectorized were identified. A method was tested which allows loops involving linked lists to vectorize. It was also shown that this solid modeler has an inherent parallelism that can be exploited. Results are presented for vectorization and parallelization. The practical limits to both vectorization and parallelization are highlighted. Improvements to the geometric intersections algorithms are suggested to take advantage of vector and parallel processing. Results of the speedups possible using these algorithms are presented.
