Mesh linking and land impact

[xigaro]

Hi! A friend of mine made me a mesh frame with rounded corners. It has a land impact of 1.

When I link it to a bunch of 20 normal prims, the linked set shoots up to 40 land impact! The root (mesh) prim had a script in too, I don't know if that makes a difference. I deleted the script but its still 40.

I understand this kind of behaviour is expected to some degree. Is there a way around it? Thanks in advance!

[Drongle McMahon]

Has to be a guess without more information. However....

When you link a mesh to a linkset of traditional prims, those prims become subject to the new LI accounting rules. There are two possible ways the increase you see could happen, depending on whether it is download weight or physics weight that is the higher and therefore proucing the LI. (It can't be server weight because that is never more than 1.0 per prim, even if all are scripted.)

The download weight for prims brought into the new accounting this way are now capped, at 1.0 for ordinary prims and at 2.0 for sculpties. So the increase you see could be a download weight, but only of at least 19 of the 20 prims are sculpties. If not, then we must conclude that it is physics weight.

Physics weights are still not capped. By default for ordinary prims, they will be physics-shape-type* "Prim" which can be very high for certain prims, especially if they have cuts and other distortions. It might be possible to get the weight, and therefore the LI, down by setting the physics-shape-type of (some of) those prims to "Convex Hull".

However, even convex hull weights can be quite high for prims with roundd surfaces, twists, cuts etc. Ultimately, if you don't need their physics (of some of them) for correct collision behaviour, you can set the physics-sjape-type of (some of) the prims, possibly including the mesh, to "None". Then those prims will not contribute physics weight at all. Just note that the root prim of the linkset (the last selected before linking) cannot be set to "None". So use the simplest possible prim as the root.

*visible and settable for individual prims in the linkset, on the "Features" tab of the Edit dialog, by checking the Edit-Selected checkbox and clicking on the target prim.

[xigaro]

Aah thank you both! With a simple root prim and physics disabled on child prims, it's down to 15 LI - much better! The child prims were tortured for text-on-prim which is what was causing the naughtiness.

I'm glad the new calculation system highlighted the needless impact I was having on the server with physics. I had no idea. Great explanation thanks Drongle!

[Kwakkelde Kwak]

---

xigaro wrote:

 

I'm glad the new calculation system highlighted the needless impact I was having on the server with physics.

---

I'm not 100% sure, but I do really think the physics weight is more a "potential" physics cost. Unless the server does pre-calculations on collisions, buried prims, or prims never collided with can have a real high physics number but still not add any actual load. Anyway, if objects aren't ment to be collided with, I'd always set the shape to "none".

[Drongle McMahon]

That's a very good point. I suppose we could ask Falcon to do some sort of phyical occlusion calculation, if we want to make him really mad at us. I guess it's a lot more sensible to leave it to the creator or assembler to note and eliminate the irrelevant physics (and safer :matte-motes-smile:).

[xigaro]

I don't really know how SL functions. With my limited knowledge I assumed when an avatar/object moves, the region needs to run through all physical objects searching for a possible collision, then do calculations based on actual collisions. It probably has a clever algorithm which discounts far-away objects within a microsecond and only does proper searches within a radius/box/whatever.

Anyway setting physics to "off" on my wall-mounted sign satisfied my OCD no end :matte-motes-big-grin:

Edit: And my friend read this thread and now understands physics calculations too. We were both very happy last night to read your explanation, many thanks again!