Hakan Seven
|
45f9f96c93
|
perf: implement L–S performance improvements
- L: pre-project wire endpoints to screen before ApparentIntersection pair loops
- O+S: tight per-segment AABB cull + move a0/a1 Vec3 outside inner snap loop
- P: epoch-cached entity→block reverse map replaces O(E×B) scan
- Q: lazy per-point projection in hit tests eliminates per-wire Vec<Point> alloc
- R: HatchModel.boundary Arc<Vec> so clone is pointer bump not heap alloc
|
2026-05-06 22:42:22 +03:00 |
|