A mental experiement on optimisation and LI...

[Da5id Weatherwax]

I will say up front that the comparable project I'm working on at the moment is not yet complete so I have nothing to compare against. However, before I see what my project ends up as, I'd like to see what the real experts on here would come up with as a guess for the final weight of a similar building.

The brief is a 4-bedroom stone-built scottish mansion over three floors with drawing room, large dining room, grand hall (double height), kitchen, conservatory, library (double height with gallery), study, ballroom (double height), master suite to include an additional bathroom and a private balcony. There will be an additional stairwell downwards to a wine cellar but the wine cellar is not included in this buildings budget as it will be a separate construction accessed via an experience-TP half way down the stairs.

The question which I invite you to answer is this:

If you were contemplating making this, before you even put a single vertex into Blender, what would you guess would be the sort of LI budget you'd be shooting for? You can assume land constraints are not a factor, so you have no hard target to aim for, you just want to be as efficient as you can while still making it look good and it being a seamless experience walking through the main building.

[ChinRey]

1 hour ago, Da5id Weatherwax said:

If you were contemplating making this, before you even put a single vertex into Blender, what would you guess would be the sort of LI budget you'd be shooting for?

There are quite a few unknown details so I can't be very exact but I'm inclined to say I'd try to keep it below 200 LI but wouldn't be too unhappy if it went up to 300-ish. I might even go as high as 350-400 if the house was to be heavily ornamented.

Edited October 16, 2021 by ChinRey

[Da5id Weatherwax]

yeah, @ChinRey - there are details missing, like the size of the plot its on and the footprint for the building, to name only two, but this is exactly what I was looking for, a mesh expert like yourself throwing out a number they'd find acceptable for that kind of build. The idea is NOT to measure myself against your estimates, I'm building what I'm building and it will fit the land. It's purely a mental exercise to see if my own estimate is reasonable and how close I can get to it in the final build

You were one of the folks I hoped would respond, BTW. You're better at optimization than I'll ever be and your "gut instinct" is a valuable source of info.

[Charlotte Bartlett]

22 hours ago, Da5id Weatherwax said:

I will say up front that the comparable project I'm working on at the moment is not yet complete so I have nothing to compare against. However, before I see what my project ends up as, I'd like to see what the real experts on here would come up with as a guess for the final weight of a similar building.

The brief is a 4-bedroom stone-built scottish mansion over three floors with drawing room, large dining room, grand hall (double height), kitchen, conservatory, library (double height with gallery), study, ballroom (double height), master suite to include an additional bathroom and a private balcony. There will be an additional stairwell downwards to a wine cellar but the wine cellar is not included in this buildings budget as it will be a separate construction accessed via an experience-TP half way down the stairs.

The question which I invite you to answer is this:

If you were contemplating making this, before you even put a single vertex into Blender, what would you guess would be the sort of LI budget you'd be shooting for? You can assume land constraints are not a factor, so you have no hard target to aim for, you just want to be as efficient as you can while still making it look good and it being a seamless experience walking through the main building.

Unfurnished if you want details eg cornices I would say up to 275 LI.

if you can do the window LODs and imposters that’s a big win on saving the LI…..

 

[Chic Aeon]

And before I read the other replies I said to myself "200 - 250.  So we seem to be more or less in agreement.  It isn't so much SIZE or even ROOMS -- it is about maximizing the details simply so that they don't start racking up land impact. Since I am a minimalist at heart it isn't difficult for me to do that.

 

And to @Charlotte Bartlett  I just moved from one of you houses (modern, two rooms with one mostly open to the elements and a very big pool) to a little beach cabin sitting on a big rock.   This over at the community beach sim.  I like them both but the rock one is now on the corner with almost no neighbors (my original one WAS the corner for awhile LOL) and it seems more cozy  for winter.  There are palm trees all around  my back entrance but I mostly blocked out the windows and can pretend I am on the Scottish coast or whatever . The other views are all water.   

 

 

[Charlotte Bartlett]

1 hour ago, Chic Aeon said:

And before I read the other replies I said to myself "200 - 250.  So we seem to be more or less in agreement.  It isn't so much SIZE or even ROOMS -- it is about maximizing the details simply so that they don't start racking up land impact. Since I am a minimalist at heart it isn't difficult for me to do that.

 

And to @Charlotte Bartlett  I just moved from one of you houses (modern, two rooms with one mostly open to the elements and a very big pool) to a little beach cabin sitting on a big rock.   This over at the community beach sim.  I like them both but the rock one is now on the corner with almost no neighbors (my original one WAS the corner for awhile LOL) and it seems more cozy  for winter.  There are palm trees all around  my back entrance but I mostly blocked out the windows and can pretend I am on the Scottish coast or whatever . The other views are all water.   

 

 

Oh you have the rock one yay!! ❤️❤️❤️

[ChinRey]

On 10/17/2021 at 4:48 AM, Da5id Weatherwax said:

 there are details missing, like the size of the plot its on and the footprint for the building, to name only two,

Also, the number of windows, doors and other smaller details.

For a well optimized house you can make a fairly good estimate of the LI by counting the meshes you need:

• If the LI is exactly half the number of meshes, it's as good as it can possibly be. (It's actually only possible for single door houses, see note below.)
• If it's lower than the number of meshes and the house is big, you're probably doing fine.
• If it's slightly higher, you shouldn't be doing too bad.
• If it's twice the number of meshes, the house is probably not well optimized.
• If it's three times the number of meshes, there's definitely something wrong.
• Unless the house is so big it extends beyond the link distance limit anyway, you want to keep it all as a single linkset so you don't want it to have more than 255 meshes. For that reason it may be a good idea to accept a less than ideal mesh count/LI ratio but 1/2 is still not good and 1/3 is still bad.

As for the number of meshes you shold aim for:

• For a big house the exterior walls should be split up with at least one separate mesh for each side. This is mainly because of draw distance (you want the wall nearest you to be the first mesh to appear as you approach the house) but also because you really need full LoD for these parts and that means you don't want too many tris for each mesh.
• The roof is at least one mesh. For a big and complex house it may well be more. In your case it's definitely at least two since the conservatory will need a roof of its own.
• The foundation may be a separate mesh or included in the exterior wall meshes. I don't think there's a hard rule there.
• The interior structure needs to be split up because of the number of faces a mesh can have. The standard is two rooms in one mesh (six faces: two walls, two floors and two ceilings) but sometimes it's better to have a mesh with a single room, either because it needs more faces or to reduce LoD swap distances and/or balance the download and server weights.
• Each window, doorframe and door needs to be a separate mesh for LoD reasons, the doors also of course because they're supposed to be movable.
• Each staircase is at least two meshes, one for the steps and one for the railing. Complex staircases need more.
• Then there are all those other details: cornices, quoins, pillars, balconies, dormers, galleries, railings, fireplaces, chimneys, chimney caps, molding, maybe some wainscoting etc., etc., etc.

---

Edit

A note about the "LI half the number of meshes" point: Each part of a linkset, whether it's a prim, a sculpt or a mesh has a server weight of 0.5 and each active script a server weight of 0.25. Since the server weight is one of the three values that can count as LI, it's not even theoretically possible to get the LI lower than half the number of parts in the linkset and with more than on script, that limit increases slightly.

Specially for math nerds, if you want to squeeze the land impact system for all it is worth, go for this:

• DW=LI+0.499
• SW=LI+0.3
• PW<=DW

Edited October 18, 2021 by ChinRey

[Da5id Weatherwax]

On 10/18/2021 at 12:11 PM, ChinRey said:

Also, the number of windows, doors and other smaller details.

For a well optimized house you can make a fairly good estimate of the LI by counting the meshes you need:

• If the LI is exactly half the number of meshes, it's as good as it can possibly be. (It's actually only possible for single door houses, see note below.)
• If it's lower than the number of meshes and the house is big, you're probably doing fine.
• If it's slightly higher, you shouldn't be doing too bad.
• If it's twice the number of meshes, the house is probably not well optimized.
• If it's three times the number of meshes, there's definitely something wrong.
• Unless the house is so big it extends beyond the link distance limit anyway, you want to keep it all as a single linkset so you don't want it to have more than 255 meshes. For that reason it may be a good idea to accept a less than ideal mesh count/LI ratio but 1/2 is still not good and 1/3 is still bad.

As for the number of meshes you shold aim for:

• For a big house the exterior walls should be split up with at least one separate mesh for each side. This is mainly because of draw distance (you want the wall nearest you to be the first mesh to appear as you approach the house) but also because you really need full LoD for these parts and that means you don't want too many tris for each mesh.
• The roof is at least one mesh. For a big and complex house it may well be more. In your case it's definitely at least two since the conservatory will need a roof of its own.
• The foundation may be a separate mesh or included in the exterior wall meshes. I don't think there's a hard rule there.
• The interior structure needs to be split up because of the number of faces a mesh can have. The standard is two rooms in one mesh (six faces: two walls, two floors and two ceilings) but sometimes it's better to have a mesh with a single room, either because it needs more faces or to reduce LoD swap distances and/or balance the download and server weights.
• Each window, doorframe and door needs to be a separate mesh for LoD reasons, the doors also of course because they're supposed to be movable.
• Each staircase is at least two meshes, one for the steps and one for the railing. Complex staircases need more.
• Then there are all those other details: cornices, quoins, pillars, balconies, dormers, galleries, railings, fireplaces, chimneys, chimney caps, molding, maybe some wainscoting etc., etc., etc.

---

Edit

A note about the "LI half the number of meshes" point: Each part of a linkset, whether it's a prim, a sculpt or a mesh has a server weight of 0.5 and each active script a server weight of 0.25. Since the server weight is one of the three values that can count as LI, it's not even theoretically possible to get the LI lower than half the number of parts in the linkset and with more than on script, that limit increases slightly.

Specially for math nerds, if you want to squeeze the land impact system for all it is worth, go for this:

• DW=LI+0.499
• SW=LI+0.3
• PW<=DW

I'm close to those design rules, depart from them a little. I'm going for one mesh per interior room, because of exactly what you describe, balancing the download and server weighting. Cornicing, wainscotting, other decoration will be involved and while ideally I would love to get the whole thing, including all doors, decorations, light fixtures and hopefully curtains too, into a single linkset I am aware that I may not be able to. Not because of distance, but because of mesh count. I deliberately took a LI hit on the exterior shell of the house because of its shape - I'd need around 10 meshes to separate out exterior walls and roof segments into their own meshes so instead I made the exterior shell a single mesh that lods gracefully into a couple of boxes. This was one place where I could save mesh count without taking too great a LI hit and visually it shouldn't be too bad because if the windows and exterior doorways lod into planar impostors as they should they will mask the fact that the exterior has lost so much at minimum lod. With the interior rooms being so much "smaller" meshes and interior lines of sight not long enough to make an interior room display at minimum lod, Ive deliberately made all interior structures lod into a couple of placeholder tris at minimum, to guarantee that none of them show any higher level until the exterior has displayed a sufficiently high lod that they wont give any pokethrough.

I am pleased, though, that off my own bat I came up with a set of design rules for my project that so closely reflect these practices.

One of the reasons I really really want to fit it in a single linkset with all decorations is that I can then use just a single script to control all its moving parts, whether that's opening or closing a door, or drawing curtains etc. I dont want the inefficiency of having to set up multiple scripts with listener channels to communicate between different linksets. I have an aversion to scripts keeping listeners open when there isn't anything happening. If I do have to make it multiple linksets though, the "main" one will be scripted to handle rezzing and derezzing the others, so the building can be moved if necessary without too much hassle. This is for me, not for sale, so if that system is a little clunky I'll probable be more tolerant of that than somebody who paid me money for it would be