BENCH 3D

3D Planar B-Rep & Mesh CSG Benchmark Suite

Comparing polystrata-3d vs. manifold-3d (C++/WASM), three-bvh-csg, and @jscad/modeling

Initializing WASM engines...
CLI Runner: npm run bench

3D Planar B-Rep & Mesh CSG Benchmark Suite

polystrata-3d vs. manifold-3d (v3.5.4 WASM) • three-bvh-csg • @jscad/modeling

3D Adversarial Glancing & Degeneracy Stress Suite (24 Scenarios × 3 CSG Ops = 72 3D Boolean Ops)

Tests kissing UV-spheres, shallow glancing cylinders (1.75°), cone apices grazing planar box faces, interlocking toroids, multi-hole extrusions, and non-manifold dangling fins. Audits every output mesh for watertight 2-manifoldness (∂1∂2 = 0, zero unmatched half-edges) and exact 3D Inclusion-Exclusion Volume Identities: Vol(A ∪ B) + Vol(A ∩ B) = Vol(A) + Vol(B) and Vol(A \ B) + Vol(A ∩ B) = Vol(A).

Library Representation Throughput Total Time Watertight 2-Manifold Volume Identity Fully Valid Crashes Topological Verdict

3D Workload-by-Workload Performance, Face Unification & Manifold Audit (10 Workloads)

Each sample executes union3D, intersect3D, and difference3D. Compare Out Faces / Tris (exact coplanar polygon unification vs. triangulated mesh splitting), Volume Identity Error (f64 vs. f32 vertex quantization), and Unmatched Half-Edges (0 = closed 2-manifold; >0 = T-junction cracks).

Library Ops / sec Relative Speed Mean (ms) p50 (ms) Out Faces / Tris Vol Identity Err Unmatched Edges Status