Mesh intersections: the good, the bad and the ugly.

@inproceedings{Lévy2026RM,
 abstract = {In this article, we propose several methods for gaining significant performance in mesh intersection algorithms while making no compromise concerning robustness. Our new accelerated mesh intersection framework is based on the following components: a new optimized arithmetic kernel that represents floating-point numbers of arbitrary length, that leverages the instruction set of modern CPUs; a parallel implementation of Axis-Aligned Bounding Box Tree construction; a new radial sorting algorithm that uses the input geometry as much as possible (preferred to the intersection points that involve heavier computations); a new symbolically perturbed orient3d predicate used to determine the membership of floating components. In addition, we propose a system to systematically test our implementation as well as other ones, based on a test-suite of Constructing Solid Geometry files in OpenSCAD format and a customizable 3D modeling kernel.},
 author = {Lévy, Bruno and Wan-Chiu, Li},
 booktitle = {2026 {RING} meeting},
 language = {en},
 pages = {317--332},
 publisher = {ASGA},
 title = {Mesh intersections: the good, the bad and the ugly.},
 year = {2026}
}