I heard somewhere that large collision mesh faces perform worse than smaller ones? Was/is that still true? For an example, imagine a plane that is around half the size of the GridMap heightmap. Which (collision) mesh would be better to use for performance? Thanks, NaxNir
Yeah, me too. It's just that I heard that it's actually worse performance that way. I don't know if what I heard was true, though.
Physics wise, the second image (in the OP - first image in this post) would be better performance (IIRC we had to subdivide the bridges in ECA a bit more to avoid performance issues) This is my own theory, so I might be wrong: edit - Actually confirmed from estama. It's somewhat like this. As far you keep the sizes (of the tris) uniform, performance won't suffer from it. Static collisions are calculate in a certain radius around the car. Now if you have a big plane which is subdivided like in the first case, the 'radius' is limited somehow as shown in the Static Collision debugger (not sure about the actual radius, I think the debugger has a different layout than the real one, but it should be something similar) If you have a big, simple, plane, the radius is no more limited, but bound the that actual face size. So it ends up to be bigger than case 1, and somehow affecting performance.
Wow, thanks for your constructive post. /s :| "Does your post contribute to the thread? Please be aware that this is not instant messaging: Every post should contribute to the thread and be written out in a semi-formal manner. This is not a text message to someone. We have Private Messaging for that." Thanks! The only thing I don't understand is why if they're different sizes, performance will be affected.
The Static Collision Manager tries to divide a surface and split in into groups. If faces are uniform, the manager can group them more efficiently: But if you start having very large faces along with small faces, things go wrong as the manager won't be able to uniformly split the groups: For example, the ECA bridges. We used to have performance issues on them as the faces were narrow and very long. It means that as soon you pass on the bridge, the static collision would had to put that big faces in a group, ruining it. Now the bridge colmesh looks like this: It's splitted in uniform parts, letting the static collision manager do it's job more efficiently. If you have 2 simple planes (4 vertices, 2 triangles), one sizes 500x500m and the other 1000x1000m meters, the performance would be the same on both. The thing that counts is that the size of the colmesh tris is uniform along the level.
Wow.. beam is more advanced than i thought. I never knew static collision was activated when you got close to it, that makes way more sense than just having everything on at once. Its like a little bubble of physics around the vehicle.