45278467d3
Finish refactor to use BVH_Triangulation instead of BRepExtrema_TriangleSet for more control over how BVHs are constructed sans OCCT shapes, and also remove need to track valid_tris.
2024-02-19 17:35:06 +11:00
2af1e2196f
WIP attempt to use BVH_Triangulation
2024-02-15 21:34:19 +11:00
a5cf4cd421
Remove 1:N clash functions
2024-02-12 11:57:23 +11:00
911b7153d7
Implement naive multithreading for N:N clashes. The clash task queue is divided by num_threads equally.
2024-02-12 11:47:20 +11:00
6cec57816e
Implement BVH based many-many clash with return struct.
2024-02-11 22:18:08 +11:00
c90f9c374f
Forgot to commit some critical bugfixes.
2024-02-08 16:28:06 +11:00
6caacd6024
Attempt a manifold check. If it isn't closed, then we use a collision test instead of intersection which is much faster.
2024-02-08 15:54:04 +11:00
2cb280932c
Indirection using vertex indices saves 20% memory but speed is now worse than when we used unordered maps.
2024-02-08 14:36:20 +11:00
e089d4c78a
Use vectors instead of unordered_map for 15% speed boost
2024-02-08 11:44:08 +11:00
155f0abbfe
Refactor out BVH-BVH clash into its own function.
2024-02-08 10:35:35 +11:00
77d9688296
Replace threeyd tri-tri intersect library with physx implementation.
2024-02-08 10:21:56 +11:00
09efd68e6e
Move vendor functions into clash_utils
2024-02-07 23:01:21 +11:00
0b3e9e7c5a
Reuse geom iterator triangulation data, don't use optimal OBBs for speed, verts now use gp_Pnt instead of BVH_Vec3d, and optimise add_triangulated.
2024-02-07 17:49:12 +11:00
b658b74b90
Fix missing clearance AABB check and add check_all clearance bool. Minor cleanup to use vert/normal maps.
2024-02-07 09:28:19 +11:00
47dd8ec4dd
Use map for verts and normals, basic implementation of piercing checks.
2024-02-06 22:21:33 +11:00
9ba1596e30
Implement early return bool for intersection tests.
2024-02-05 18:47:08 +11:00
930020bf05
Use NVIDIA's ray-tri intersection implementation under BSD-3 license.
2024-02-04 22:51:51 +11:00
1fe69e5f19
Simple implementation of clearance clash.
2024-02-04 16:02:45 +11:00
988c8f16f3
Implement collision clash type (pure tri intersection regardless of protrusion)
...
This should be equivalent to select(element) but faster and with an allow_touching toggle.
2024-02-04 09:01:35 +11:00
080b628df5
Optimise by removing unnecessary (very slow) triangle shrinking, more optimal OBBs, check OBB first for points_in_b prior to doing raycast (much faster now), and return the surface point of the protrusion too for convenience.
2024-02-03 16:14:26 +11:00
d0333ac7cd
Penetration distances now are influenced by the normal for better results.
...
Add cache for points that have already been checked.
Separate out new add_triangulated function with new kwarg add_element(should_triangulate=True) so that you can still have the old geom tree using aabb only if you want, also because add() is used in boolean_utils. Tolerance is now a kwarg.
I misunderstood the OBB dimension and should *2 not /2 which now makes it much slower since there's more triangles to check.
2024-02-02 23:35:06 +11:00
e95e16c102
First attempt at implementation of intersection clash checks.
...
This considers only protruding clashes. It does not consider touching nor encroaching clashes and does not consider protrusion direction in the protrusion distance.
2024-02-01 18:30:20 +11:00
c18458b3db
Refactoring
2022-11-13 11:21:58 +01:00
3736a8b4b4
Refactoring
2022-11-13 11:04:16 +01:00
bc2d76e685
Include-what-you-use
2022-06-15 13:41:13 +02:00
Dion Moult