Why Doesn't Convex Hull Work Magic As It Used To?

[Prokofy Neva]

I wonder if anyone else has had this experience, or it is just idiosyncratic or incomplete on my part.

In the early glory days of the first appearance of the Convex Hull miracle, you could toggle to this option on "feature" on the item menu, from "prims," and it would significantly reduce prims in a prim object.

This didn't work so well on houses or buildings with doors and certain other things, where it would then block your entry.

But for something like an old-school piece of heavy-prim furniture, or let's say a utility belt that you put out inworld instead of wearing, it would cut the prims in half or more.

It seemed many things would reduce. I and others gloried in going around sims and reducing everything -- this was in the days before more prims were added, but it's still relevant. You can never be too rich, too thin, or have too many prims.

But in the last few months I've noticed that the Convex Hull magic just never works.

It either "does nothing" or worse, it ADDS to prims so you back out of that fast.

It rarely ever works.

Although I did discover with constant efforts that sometimes, while a prim count will appear to remain the same, the convex hull bit will reduce the LAND IMPACT, which is usually the same as prims, but can be a bit different, not by much.

For this, you have to know what your whole sim or parcel's prims were before doing this, and watch to see if it jumps UP, i.e. making MORE land impact available. So that happens now, but still, not as much as the original magic.

I don't know why convex hull does this but it may be that it has changed.

[Chic Aeon]

There are working the way they have for ages -- for me anyway. You never actually reduced the PRIMS, only the land impact which is what counts now and has for many years. Prims designate the number of OBJECTS in a linkset be it the legacy prims or mesh objects or a combination of both.  

Six prims - 3 land impact when turned to convex hull. The only thing that changed from the VERY beginning (really the VERY BEGINNING) is that sculpt objects can no longer skyrocket the land impact. They CAN increase though but I think it is only up to 2 from one. You would have to research that. I haven't used scupts in at least six years except for some old plants and I KNOW not to link sculpts with other items so it isn't an issue. 

And before someone catches that "mesh prim mapped" detail, here is a very OBVIOUS prim photo with the same info.

 

Edited April 25, 2018 by Chic Aeon
spelling

[Klytyna]

3 hours ago, Prokofy Neva said:

This didn't work so well on houses or buildings with doors and certain other things, where it would then block your entry.

Convex hull was never "magic", it simply told SL to use the physics shape of the original untortured prim, so setting convex on a hollow box as a way to make a low prim room, never worked, as the physics ignored the hollows and slices.

3 hours ago, Prokofy Neva said:

it would cut the prims in half or more

Best you could realistically hopew for was a 50% decrease in Land Impact, it never "reduced prims" the count of prims in the linkset remained constant, only the land impact fell, and how much it fell depended on factors such as what shape prims you were using, had you tortured them, did they have scripts in etc.

Boxes worked best,  a torus with a script in, didn't work nearly as well.

Convex hulling parts of a linkset was always a "pro prim pusher" trick used by the more tech savy builders, but magic it wasn't, it had and still has certain limitations, it's not a sure fire way to halve land impact on everything.

I once built an object with 15 stretched cubes (a wooden animal stall), convexed it came in at 8 LI (7.5 LI rounded up), but adding a torus with a leash holder script in, pushed LI up to 11 for a 16 prim object.
 

[Qie Niangao]

8 hours ago, Chic Aeon said:

The only thing that changed from the VERY beginning (really the VERY BEGINNING) is that sculpt objects can no longer skyrocket the land impact. They CAN increase though but I think it is only up to 2 from one. You would have to research that. I haven't used scupts in at least six years except for some old plants and I KNOW not to link sculpts with other items so it isn't an issue. 

Yes, at "the very beginning" sculpts could be disastrous, so it was a welcome change when they were "tamed" to a maximum of 2 Land Impact per sculpt in the linkset. (I too am hesitant to claim that's a hard and fast rule, but it seems to be.) I also think I remember different Land Impact calculations for embedded scripts at "the very beginning" but I could be misremembering.

I find I'm at a point now where squeezing Land Impact savings is more difficult just because most stuff I have is already using "mesh" LI accounting: even if I haven't set Convex Hull I've usually applied Materials maps or set Alpha Masking on some face. So now it's a more challenging hunt, seeking opportunities to change linksets. This can result in unrelated stuff linked together, and you have to be sure any scripts will work with different parts linked / unlinked, but it's certainly possible to scrounge a few precious LI points this way.

I'm not sure of a hard-and-fast algorithm for this, but the opportunity to benefit from breaking apart a linkset may arise when either

• there are lots of scripts in an object, or
• the object contains one or more links with individual Download weights having fractional components between .0 and .5. 

Splitting apart the latter can result in both parts rounding down; for example, a 1.2 linked with a 1.4 will total 2.6, rounding to 3, whereas unlinked they round down to 1 each. In the former case, splitting apart scripted objects can take advantage of the ceiling on Server Weight (affected by scripts) of no more than 1 per linked element so for example a two-person, four-link furniture item may contain enough scripts to score 4 Land Impact despite low Download complexity; split apart, the scripted pair will score 2 LI and the other pair only 1.

Combining items into a larger linkset may win when they're totalling more than 0.5 LI per element (notably, singletons cost 1, so are likely candidates) and they have "complementary" weights (e.g., an item with LI driven by Server weight can save when linked with an item driven by Download weight -- as long as that Server weight didn't come from scripts, see above).

[ETA: The arithemetic for rounding down multiple fractional impact scores can theoretically apply to other weights than Download, too, but in practice that seems to be where I find actual savings. YMMV.]

Edited April 25, 2018 by Qie Niangao

[Bitsy Buccaneer]

Man, I'm not able to connect the dots well enough to figure out if this is what Qie's talking about but at some point adding a script to an object of linked bits (be they prim or mesh), over-rode any .5 + .5 LI reductions and made each bit count as 1 LI. AFAIK a single script was sufficient for this to happen.

example of the chair I learned this with (which Prok later bought):

chair = 1 LI, two cushions = .5 LI each, prim to make it show me as creator = .5 LI

unscripted = 2.5 LI, averaged up to 3 LI

scripted = 4 LI

If this is still how it works, it could look like it broke the trick of changing linked items to convex hull to reduce LI.

So, question for Prokofy: are the items you aren't able to reduce now scripted or unscripted? Have you tried recently with unscripted objects or freshly rezzed cube prims?

[Qie Niangao]

30 minutes ago, Bitsy Buccaneer said:

chair = 1 LI, two cushions = .5 LI each, prim to make it show me as creator = .5 LI

unscripted = 2.5 LI, averaged up to 3 LI

scripted = 4 LI

I've read a little more carefully that article I linked to above, and indeed there was a change in how scripts affect Land Impact. That change very well may have been for the better in this case.

The same thing -- a 4 element linkset getting a 4 Land Impact when scripted -- will still happen with enough scripts in the object (common enough with modular animation engines such as AVsitter), but with the original calculation, adding even a single script to an object made the whole thing "non-static" so it automatically scored a Server weight of 1 LI per element in the linkset. Now that 1-per-element score is the maximum Server weight when counting scripts at 0.25 each plus 0.5 per linked element.

Edited April 25, 2018 by Qie Niangao

[Chic Aeon]

7 hours ago, Qie Niangao said:

I also think I remember different Land Impact calculations for embedded scripts at "the very beginning" but I could be misremembering.

Yes, I remember that issue and conversations on it. At the time I wasn't making furniture with animations so not cemented in my brain.

The OTHER thing that folks can do to cut their land impact down is LINK some "low prim" (below .7 land impact mesh in most cases but it works with prims and mesh) together. So say you have a stack of books (.5 li) on a table (.8 li). If you LINK them (and they aren't sculpts of course) you get 1.3 li which rounds down to 1 land impact count.  Typically I don't use linksets as I prefer to join items before upload to save on both land impact and upload costs :D, but for personal decorating this is a great way to get the biggest bang for you land impact buck. Just check what the land impact is of the objects (More Info pane) before linking. 

 

[Bitsy Buccaneer]

4 hours ago, Qie Niangao said:

I've read a little more carefully that article I linked to above, and indeed there was a change in how scripts affect Land Impact. That change very well may have been for the better in this case.

Thanks Qie. I like this as a change Will try to remember to check the chair's current LI when I'm inworld later.

[Chic Aeon]

It took me some searching (5000 ish posts) but I found my blog post on the change and apparently I wrote it on Christmas!  LOL   

https://chicatphilsplace.blogspot.com/2012/12/when-is-prim-not-prim.html