Dataset and analysis artifact for "Branching-Factor Selection for Wide Bounding Volume Hierarchies in Real-Time Collision Detection: A Pre-Registered CPU-Serial and GPU-SIMT Benchmark"

Pre-registered benchmark artifact for a controlled study of wide bounding volume hierarchy (BVH) branching factor (k in {2, 4, 8, 16}) in the real-time collision-detection narrowphase, measured on a CPU-serial engine and a GPU-SIMT (CUDA) engine over three high-resolution Stanford meshes with a QEM LOD ladder. Contents: the frozen pre-registration document; the PowerShell sweep driver; the full master sweep CSV (450 conditions, 30 samples each, 13,500 rows, fixed seed 1337); the structure profile (node counts, memory footprint, build times per mesh, LOD rung, and k); all derived statistical tables (mechanism, bridge check, structure analysis, secondary Mann-Whitney tests, log-log scaling fits, fastest-arm and ordering tables); validity-gate reports (k=2 AABB regression anchor and control comparison); numpy-only analysis scripts; and publication figures F1 to F6 (PNG, 300 dpi). A README documents the CSV schema, build path, and the pre-declared exclusion of two degenerate cells from per-cell timing figures. Companion code repository: https://github.com/RazordousJR/BVH_CollisionDetection