Partri

partri is a term used in computational geometry to denote a class of parallel triangulation algorithms intended to speed up processing of large triangular meshes. The name blends parallel computing with triangle meshes and is used in some literature to designate methods that partition a mesh for distributed computation while preserving geometric fidelity.

Partri algorithms typically operate by partitioning the vertex-edge graph of a mesh into subdomains, performing local triangulation on each subdomain, and then merging the results across subdomain boundaries. Common strategies include graph partitioning with minimal edge cut, overlapping ghost regions to maintain continuity, and post-processing to ensure consistency of shared vertices.

Applications include finite element analysis, where large meshes must be processed efficiently, computer graphics for high-resolution terrain and surface models, and scientific visualization. partri-based methods aim to balance computational load across processors and minimize communication overhead.

Limitations include challenges at partition boundaries, such as maintaining mesh congruence and avoiding duplicated or conflicting vertex data, as well as potential numerical perturbations introduced during reassembly. Performance gains depend on mesh structure and hardware architecture.

Variants of partri may emphasize different partitioning schemes, reassembly protocols, or error-control criteria. Related concepts include mesh partitioning, domain decomposition, and parallel mesh generation.