LI Calc gone nuts when mesh linked to prims sculpts

[Toysoldier Thor]

Just wanted to throw you all a weird story of what I encountered last week regarding linking one of my MESH models with an already liked group of prims and a sculpty (an animated texture art statue).  Thankfully I had 3 expert mesh builders come to my gallery to witness what I did as they didnt believe it either but they all saw what I saw.

Background:

I have a 5 prim art statue on the top floor of my art gallery.  It is made up of 1 hollowed cylinder prim, 1 sculpty with an animated water texture (script removed that made the texture move), and 3 "tortured" torus prims about 10m tall that also have animated textures on them (script removed as well).  So, this 5 prim art statue has a prim count of 5 and stands about 10-12meters tall.

I also have a single mesh woman model about 6m tall that has an LI of 3.

 

What I wanted to do was link the mesh woman into the middle of this "animated lights pillar" statue.  It looked cool that they were together.  No biggy.... I simply linked the already linked-set of prims and added the mesh to the linkset.

When I did this my LI count went to....   drumroll please ..........   237 !!!

Yup - yo uread that correctly - 237 !!

I tried unlinking the mesh from the initial linkset (which brought the counts back to 5 & 3) and then linking the linkset to the mesh instead.  NOPE... 237.  Three other mesh experts on chat didnt believe me and TPed in to see.  They suggested I make it convex hull - but already was.  Turning off convex hull moved the count from 237 to 238.

They all blamed the 3 tortured TORUS prims for the weird linked anomaly and figured there wasnt much I could do about it.  But as a last thought I decided to unlink the previous linkset and simply link all of them back up together with he mesh.  Magically the new resulting linkset was only 8.

We tried to reverse what I did to see if we could recreate the anomally but relinking the old linkset back to what it was.... but nothing I could do would recreate the problem. 

No one had an answer and the problem is permanently gone.

[Theresa Tennyson]

Next time something like that happens try changing the hollow/tortured prims to physics type "none." It's the physics calculations that make LI on shapes like that explode. With physics off they should have a LI no more than 2.

[Innula Zenovka]

I've noticed something similar when doing some experiments changing linksets made out of legacy prims to mesh accounting -- I can't be certain because it's difficult reliably to replicate but I think that if you first change something to none or convex hull leaving the rest of the prims as prims, and the LI goes through the roof because of the physics weight, it doesn't recalculate immediately when you change the rest to convext hull or none.    A couple of times, I've changed things and not immediately seen the reduction in LI that I expected but, when I've looked again a few minutes later, it's gone down to what it should be.

I'm very unsure if it's the sim taking time to do the recalculation or the viewer taking time to reflect it.      

Thanks -- that trick of unlinking everthing and relinking it is worth knowing.

[Toysoldier Thor]

@Theresa

myself and 3 other mesh experts tried a few things including setting the prims to none.  But it seemed the only way to fix it was to unlink the old Legacy linkset and relink it all.

A theory I have is that the old linkset way back when had texture animation scripts in each of the animated prim of the set.  I removed the scripts from the linkset as they had no value (once a texture is animated - the script is no longer needed).  But I am wondering of the linkset held some parameter telling the new LI calculation algorithm that there are scripts in this linkset.

When I unlinked everything and relinked it together - this parameter was reset? 

I dont know but I could not recreate the problem at all.  It seems something to do with the linkset holding some legacy parameter.

 @Innula... your welcome. I thought it was pretty freaky so I thought I would point it out here.

[Drongle McMahon]

Before the pathfinding servers came out, the LI system would count the tortured toruses as mesh for download weight and physics weight. Download weights of several hundred for what you describe were quite normal in that system. Now, the download weight per legacy prim is capped at 1.0 (2.0 for a sculpty), although tortured prims can still make very high physics weights. You seem to be confident that is was not a physics weight effect. So it sounds as if your linkset was still being calculated under the pre-pathfinding system even after when you added the mesh to it, and that it only switched to the new capped scheme after unlinking and relinking. It would be useful to know if that is a general problem. I don't know whether the weights are cached on the server for rezzed objects. If that's the case, taking into invenyory and re-rezzing might have a similar effect to unlinking and relinking. I'll have to see if I can find any suitable old rezzed linksets to test this with.

[Toysoldier Thor]

@Drongle,

I am pretty confident it has nothing to do with physics since I wasn't unable to re-create the anomaly.  If it had something to do with physics, then I should have been able to reverse the linkset to the original configuration and the problem should have re-appeared.   I am not 100% though since I didnt spend another 15 minutes trying each reverse combination.

I have a feeling that the problem is rooted closer to your theory.

This linkset has been rezzed at my art gallery since mid-July (prior to pathfinder upgrades).  As soon as the other mesh experts saw what I was linking together with mesh they right away pointed at "wow its the 3 tortured torus's that you are using".

BUT, maybe by unlinking the entire set and re-linking the set set, the new formula with the capping engaged.  This does make sense.  If so, taking the legacy linkset into inventory and re-rezzing it would have also worked.

[Drongle McMahon]

I would call it a bug if the recalculation isn't being done until a linkset is unlinked. If we can find a reproduction that doesn't destroy itself, we could make a jira and see if they will fix it. I don't suppose you have another copy of the linkset rezzed somewhere from before the change? Even then it might be difficult because copying makes a new linkset and would probably trigger the recalculation in at least the new copy. Hmm.

[Toysoldier Thor]

---

Drongle McMahon wrote:

I would call it a bug if the recalculation isn't being done until a linkset is unlinked. If we can find a reproduction that doesn't destroy itself, we could make a jira and see if they will fix it. I don't suppose you have another copy of the linkset rezzed somewhere from before the change? Even then it might be difficult because copying makes a new linkset and would probably trigger the recalculation in at least the new copy. Hmm.

---

Yes I suspect it is a bug but if so it will be one that diminishes in impact as time goes on.  I will have to look around my store/gallery to see if there are some examples.  Problem is that very few of my builds would use tortured torus.  Ohh hold on there might be!  I will check when I get back inworld.

[Toysoldier Thor]

OK Drongle... Here are my tests on an older linkset.  I dont think the old prims were a factor.    I believe Theresa was correct.

I did have a few more older animated art pillars that used my tortured torus prims that were still rezzed inworld prior to the pathfinder upgrade you mentioned.

One of them had 6 prims (5 prim 2 of them tortured torus & 1 sculpty).  When I linked it to a 3 LI mesh statue (I added the mesh to the older prim linkset) the total LI went from 9 to 212.  Even when I added the prim linkset to the mesh - LI was still 212.

THEN... I simple dragged a copy of the old linkset to make an identical new copy of the linkset and did the exact same linking.  I did not unlink the linkset.  The results were identical.  LI = 212.

THEN....  I took the old linkset and completely unlinked it and then simply relinked the linkset.  Then I linked this refreshed linkset to the mesh and the count was still  LI = 212

THEN...  I unlinked EVERYTHING and relinked ALL of the prims and mesh tegether as one.  And the count was STILL  LI=212.

 

I was completely lost so I took the entire linkset with the mesh at its count of 212 while in Convex Hull and switched it to Physics shape of NONE.  And surprise surprise..... the count dropped down to LI = 8

I thought... ok let me switch the entire linkset back to CONVEX HULL and it should go back up to 212.....  It didnt.  It stayed at 8.  If I switch the shape type to PRIM then the count flew back to 212.  But swithing it back to convex hull or none brought the linkset back to 8.

It seems when I changed the entire linkset to NONE from CONVEX HULL - that fixed things - even if I switched it back to convex hull.  PRIM broke it again.

I guess it was the physics type that was the root cause.

 

[Kwakkelde Kwak]

---

Toysoldier Thor wrote:

Thankfully I had 3 expert mesh builders come to my gallery to witness what I did as they didnt believe it either but they all saw what I saw.

---

I can't help wondering how they were experts if the issue was the physics weight/setting of your items. That's a well known issue to say the least and has been written about on the forums many many times. It shows how horribly demanding the physics can be if you don't watch out. At least now it's shown in landimpact, unlike before.

Anyway...glad you got it solved.

[Drongle McMahon]

Oh good. That's what's supposed to happen. I guess the physics shape type of the old one got switched at some point without you or your visitors realising. So it looks like there is no bug then. Whenever you have this sort of problem, it's always worth clicking the "More Info" link which should tell you whether it's the download or physics weight that's high.

Here are the rules for physics weights of a linlkset ..

If everything is legacy prim (including sculpty) AND all are set to type "Prim" then it doesn't matter because it's old style accounting. If one member is mesh OR any is not "Prim", then it's new style accounting. Tortured prims can be VERY expensive in physics weight when they are type "Prim" and the new accounting is used.

Thus, for a tortured torus (root) linked to a cube: Both "Prim", LI=2. Set cube to "None", LI=668. Set cube to "Convex Hull", LI=669. Set cube back to "Prim", LI=2. Set torus to "Convex Hull", LI=2. Now set cube to "None", LI=1. If you do all that whuile switching on physics shape display (Dveelop->Render Metadata->Physics Shapes), you can see exactly what's going on. You can see the torus go red when it gets switched to the new accounting system and it's "Prim".

[Kwakkelde Kwak]

---

Toysoldier Thor wrote:

I dont think the old prims were a factor.

---

From what you describe they most certainly were.

As soon as you link a "normal prim"  or sculpted prim to a mesh, SL treats them as if they were meshes aswell.

So the grandfathered 1 prim sculpt and torus, show their true colours so to speak. The 212 you noticed is representing the actual load on the physics engine, or at least potential load. This load is no higher than it was before you linked the mesh to them, or them to a mesh, it just never showed in landimpact.

You can see this load when editing the objects, whether they are calculated as mesh or not for landimpact.

This is a standard torus and will cost you only 1 in landimpact. Notice the physics weight of 66.9. As soon as you link a mesh to this and leave the physics at "prim" that 66.9 will really count.

[Toysoldier Thor]

---

Kwakkelde Kwak wrote:

---

Toysoldier Thor wrote:

Thankfully I had 3 expert mesh builders come to my gallery to witness what I did as they didnt believe it either but they all saw what I saw.

---

I can't help wondering how they were experts if the issue was the physics weight/setting of your items. That's a well known issue to say the least and has been written about on the forums many many times. It shows how horribly demanding the physics can be if you don't watch out. At least now it's shown in landimpact, unlike before.

Anyway...glad you got it solved.

---

They were good experts as all three of them immediately said it was the physics of the "tortured torus" - a term I had never even heard of prior to their visit.  They even saw the info that showed it was high (something I didnt know how to check).  What stumped them in the 10 minutes of their visit without full rights to my legacy linkset was that they said "change it from prim to convex hull" and I told them (because that is exactly what it showed) that the linkset already was set to CONVEX HULL.

And what seemed to have fixed the problem was to ignore what the Edit window said about the linkset being CONVEX HULL and flip-flop it to anything other than convex hull and then back to it (or as Theresa said - to NONE as well).

 

Final point that might shock some of you .... the vast majority of mesh creators likely DO NOT follow this forum nor all of its created threads on a daily or weekly basis... if they do at all.  So even though you are correct that maybe this issue was extremely obvious to you and the forum loyalists and it has been spoken about "many many times" in this forum, it really doesnt help when most mesh creator dont follow this forum unless they need help.

There is a reason that this topic has come up "many many times" in this forum and I can safely predict that it will come up many many many many more times in this forum in the months and years to come (as it would also be coming up in the other SL mesh forums that you likely are not a part of)......  because the vast majority of Mesh creators have not been on this forum nor do not watch this forum and only come to this forum if they encounter this problem and need an answer.

[Kwakkelde Kwak]

The convex hull you saw for the "linkset" was not for the linkset. It was for the parent object. SL doesn't show "multiple"  in that window. After you flipflopped the settings, I am sure it said convex hull for the "linkset" again. You have to set it per child, manually, or set it for the entire linkset, manually. SL prims are set to "prim" by default.

I never ment builders should have known because they should have visited the forums, nor did I say so. I ment the "problem"  has been here from the day LL introduced mesh to the betagrid, so I was assuming people dealing with mesh on a regular basis shouldn't be surprised by an increase in prims from 3 to 212. Just from personal experience. 

[Toysoldier Thor]

Well I have been working with and learning mesh for 8 months and I even occassionally peek in and contribute to this forum at times (when I need help or want to share what I learned) and I even did know the general inworld mesh algorithm no-no's that linking Mesh with anything else like non-mesh prims & sculpts & scripts often causes horrible things to happen with LI (in addition to countless other factors that penalize the use of mesh inworld)....

But until this week, I didnt know what a "Tortured Torus" was other than I knew that I created one in the past as art statues.  I didnt know that even though a linkset of prims says one type of object "convex hull" that it doesnt mean all the prims in the linkset are not convex hull.  Nor did I know how to fix it until I kept fiddling with the various combinations of linking until the massive LI dropped.

But then again I would guess I am like a large portion of the mesh creating population (i.e ex casual sculpty makers) that are not heads down MESH ONLY creators.  Mesh is just one of countless things I do in SL and my RL around SL.  I do sculpties, textures, write some scripts, assemble some prims, do SL/RL/Digital art, participate in SL's art community, sing karaoke, go to live SL concerts, participate in the Merchant community of SL, etc. etc. etc.  I maybe create a mesh creation every couple months.

Maybe if my prime focus in SL was to create mesh then I would focus all my time on following and participating in MESH forums and reading all the threads and going to the LL User Group meetings on mesh...etc.  But sadly i am not.... so I come to the forums like this when i need to and ask the dumb questions some of you mesh loyalists have known long time ago (likely even before mesh was officially released) and have been asked countless times before. 

[Kwakkelde Kwak]

Well that's why I asked about the "expert"  level of your experts. Nothing more nothing less.

I don't expect everyone building or texturing even just playing SL to know all these things at all and am always willing to answer "dumb questions"  as you call them over and over. Occasionally I have some myself. That's why the forums are here afterall, to get some answers to those.

A "tortured" torus isn't that much worse than a normal torus btw, as the picture showed that's quite bad by itself. Hollowing one has a huge impact though. Well over 200 for a normal hollow one against "only" 66.9 for a solid one.