Modern Sim Lag Times

[Klytyna]

50 minutes ago, Theresa Tennyson said:

The additional physics time that pathfinding characters use comes out of script time instead of general frame time and someone has piffled very authoritatively that scripts don't cause server lag because the time per frame scripts can use is capped.

i've seen a sim that was, what was the term you used in that thread "a hot mess", with lot's of linksets that hadn't been included in the pathfinding navmesh, where a single pathfinding scripted thing added 7-9 ms to the physics time.

Physics time can't "come out of general frame time" because there is no such thing...

A regions "Frame Time" is the TOTAL of the list of component times, physics, agent, simulator, script, spare etc.

In a bid to try and maintain Frame time at about 1/45th of a second (22ms or there abouts) the servers prioritise, and reduce script time inline with the increase in physics time, but there is a limit...

If the sim runs out of script time for physics time to steal, then the excess physics time leads to an increase in Frame time, and the Sim-FPS/Physics-FPS reported on "SL-Fossil" lag meters drops from 45/45. Server side lag.

There is some sort of limit to how much script time gets stolen... The Servers seem to be set up to prevent it bottoming out completely and thus stopping all script execution.

It's also worth mentioning that the pathfinding 9 ms lagbomb is in ADDITION to the regular physics time, for physical set linksets, vehicles, elevators, wandering prim-beasts, collisions etc.

Let's quote the last comment on the thread you linked to...

On 25 January 2018 at 2:38 PM, ChinRey said:

That would imply that the server is supposed to allocate 9 ms of the frame time to pathfinding. And no, I don't believe that. I can't believe that. No matter how much I have criticized the pre-2014 LL developer team before, that would be taking it too far.

I can believe it, I've seen it.

Bear in mind that this pathfinding lag-tech crap was dreamed up by people who's idea of testing was to run something for 15 mins on a flat empty green beta grid square for 15 minutes, while watching some test npc walk around a few 5 m cubes...

The 2012 pathfinding pre-release "infomercial" thread you linked to in that January thread, stated that we shouldn't see performance degradation because on a full region pathfinding would only use a maximum of 4 ms of frame time.

But your own test, and Chinrey's, showed that claim by the 2012 pathfinding project manager to be... FALSE.

9 ms...

Remember that 2012 was back in the days when Official Policy was "there is no such thing as lag". Be wary of quoting 6 year old claims made by vanished non-technical Linden Clan members during the Era of Lag Denial

 

[Theresa Tennyson]

3 minutes ago, Klytyna said:

i've seen a sim that was, what was the term you used in that thread "a hot mess", with lot's of linksets that hadn't been included in the pathfinding navmesh, where a single pathfinding scripted thing added 7-9 ms to the physics time.

And a dozen scripted pathfinding things won't increase that physics time because the hit is for the single routine that handles all pathfinding characters.

Physics time can't "come out of general frame time" because there is no such thing...

A regions "Frame Time" is the TOTAL of the list of component times, physics, agent, simulator, script, spare etc.

But it's never going to be less than 22 ms, because that's the general length of a frame. If a frame is less than 22 ms, spare time is added in.

In a bid to try and maintain Frame time at about 1/45th of a second (22ms or there abouts) the servers prioritise, and reduce script time inline with the increase in physics time, but there is a limit...

If the sim runs out of script time for physics time to steal, then the excess physics time leads to an increase in Frame time, and the Sim-FPS/Physics-FPS reported on "SL-Fossil" lag meters drops from 45/45. Server side lag.

There is some sort of limit to how much script time gets stolen... The Servers seem to be set up to prevent it bottoming out completely and thus stopping all script execution.

And since I've never seen pathfinding go over 9 ms and the limit for scripts is in the range of 15-20 ms, how will that happen?

It's also worth mentioning that the pathfinding 9 ms lagbomb is in ADDITION to the regular physics time, for physical set linksets, vehicles, elevators, wandering prim-beasts, collisions etc.

That would require both pathfinding and non-pathfinding wandering prim-beasts simultaneously, wouldn't it?

 

 

[Klytyna]

6 minutes ago, Theresa Tennyson said:

...

Do something useful today, learn how to quote sections of somebody elses post rather then inserting your stuff into a single quote of theirs... It's not hard.

8 minutes ago, Theresa Tennyson said:

But it's never going to be less than 22 ms, because that's the general length of a frame. If a frame is less than 22 ms, spare time is added in.

If you had bothered to READ my post, you would have seen that i said...

28 minutes ago, Klytyna said:

A regions "Frame Time" is the TOTAL of the list of component times, physics, agent, simulator, script, spare etc

So, your comment that spare time is added to the other times to keep Frame time from dropping below 22 ms is... "Completely Redundant".

12 minutes ago, Theresa Tennyson said:

And since I've never seen pathfinding go over 9 ms and the limit for scripts is in the range of 15-20 ms, how will that happen?

Physics time for a sim is whatever the NORMAL physics time is, from physical set prims, vehicles, collisions etc., PLUS the Pathfinding 9 ms Lagbomb. In an attempt to maintain total Frame Time at 22 ms, the server cuts script time, and it can drop below 15 ms, that you havn't seen this, isn't MY problem.

When the script time bottoms out at the lowest value the server will permit, and spare time is zero, if the TOTL Frame time is still over 22 -23 ms, then... Sim FPS drops...

Let me clarify that in terms EVEN a tech-illiterate pseudo-intellectual philosofailure graduate could understand, should one happen to read this...

If Spare Time + Script Time + Net Time +Agent Time + Simulator Time + Physics Time etc., etc., etc., is GREATER than 1/45th of a second, then LAAAAAAAAAG...

You're on a sim with an already laggy build, physics times at 5 ms, simulator times at 2 ms, agent & net time's about the same, scripts are running at 15%, and THEN some CRETIN rezzes a pathfinding self propelled lagbomb and jacks the sims physics time to 14 ms...

22 minutes ago, Theresa Tennyson said:

That would require both pathfinding and non-pathfinding wandering prim-beasts simultaneously, wouldn't it?

Yes, and you forgot to add physical vehicles, elevators, grabable dice, tumbling rockfalls (yes those are a thing), physical projectiles (all those bullets on ePeener combat sims) etc., etc., etc.

Oh... and in case you hadn't noticed, despite pathfinding being 6 years old, we still have non-pathfinding wandering prim beasts.

Just an FYI there, I know you like to dwell in the past, quoting 6 year old infomercial rubbish etc... So you might not have noticed.
 

[Theresa Tennyson]

19 minutes ago, Klytyna said:

Physics time for a sim is whatever the NORMAL physics time is, from physical set prims, vehicles, collisions etc., PLUS the Pathfinding 9 ms Lagbomb. In an attempt to maintain total Frame Time at 22 ms, the server cuts script time, and it can drop below 15 ms, that you havn't seen this, isn't MY problem.

When the script time bottoms out at the lowest value the server will permit, and spare time is zero, if the TOTL Frame time is still over 22 -23 ms, then... Sim FPS drops...

Indeed.

But it won't be the pathfinding that does it, will it?

(Or alpha-blended textures, of course, but somehow you forgot you said that.)

Edited September 23, 2018 by Theresa Tennyson

[Theresa Tennyson]

25 minutes ago, Klytyna said:

You're on a sim with an already laggy build, physics times at 5 ms, simulator times at 2 ms, agent & net time's about the same, scripts are running at 15%, and THEN some CRETIN rezzes a pathfinding self propelled lagbomb and jacks the sims physics time to 14 ms...

 

And then the scripts are going to run at 2% in their same subframe. The region you're describing is probably going to be lagging already.

[Klytyna]

21 minutes ago, Klytyna said:

When the script time bottoms out at the lowest value the server will permit, and spare time is zero, if the TOTL Frame time is still over 22 -23 ms, then... Sim FPS drops...

2 minutes ago, Theresa Tennyson said:

But it won't be the pathfinding that does it, will it?

21 minutes ago, Klytyna said:

Let me clarify that in terms EVEN a tech-illiterate pseudo-intellectual philosofailure graduate could understand, should one happen to read this...

If Spare Time + Script Time + Net Time +Agent Time + Simulator Time + Physics Time etc., etc., etc., is GREATER than 1/45th of a second, then LAAAAAAAAAG...

You're on a sim with an already laggy build, physics times at 5 ms, simulator times at 2 ms, agent & net time's about the same, scripts are running at 15%, and THEN some CRETIN rezzes a pathfinding self propelled lagbomb and jacks the sims physics time to 14 ms...

Adding 9 ms to the physics time on some sims WILL push them over the 1/45th of a second Frame time, and cause sim fps to drop, claiming otherwise is either tech-illiterate BS or fraudulent hypocritical BS, or both.

Which kind of BS are you intending here?

6 minutes ago, Theresa Tennyson said:

Or alpha-blended textures, of course, but somehow you forgot you said that

I didn't forget, but I felt that your hypocritical BS on that point didn't deserve a reply, but since you insist...

Unlike you, when I advise somebody on de-lagging a sim, I don't limit that advise to ONE type of lag.

I don't accept that it's ok to shag peoples viewers into the ground with crappy 2006 build-fail techniques as long as the 2006 lag meter reports "sim fps 45 - everything is fine!"
 

[Theresa Tennyson]

1 minute ago, Klytyna said:

Adding 9 ms to the physics time on some sims WILL push them over the 1/45th of a second Frame time, and cause sim fps to drop, claiming otherwise is either tech-illiterate BS or fraudulent hypocritical BS, or both.

That would explain why people rezz pathfinding objects when they want to grief a region.

Only they don't, do they?

It's trivially easy to create a pathfinding character, and griefers are certainly the creative type yet I haven't heard the howling that would result if this was a significant problem. Never mind the fact that an estate owner can simply turn pathfinding off anyway.

[Klytyna]

1 minute ago, Theresa Tennyson said:

And then the scripts are going to run at 2% in their same subframe. The region you're describing is probably going to be lagging already.

And yet MORE hypocritical BS...

Now, having had your earlier BS blown out of the water, you're claiming its OK, to add a 9 ms lagbomb to a sim if it's already laggy...

Outstanding...

Terrible Theresa's guide to lag reduction:

"If your sim is laggy, what the hell, add more lag till the sim dies!"

 

[Klytyna]

1 minute ago, Theresa Tennyson said:

and griefers are certainly the creative type yet I haven't heard the howling that would result if this was a significant problem

Actually, no, most are NOT that creative, most use griefing tools made years ago, by ancient SL-Fossils, deformers for example are not the problem they once were, now that you can reset your avatars skeleton without having to sit on or click an undeformer, or relog...

But griefers still use that ancient outdated crap.

Oh and good try at derailing from your failed attempts to defend pathfinding into some nonsense about griefers...

Moving on...

3 minutes ago, Theresa Tennyson said:

Never mind the fact that an estate owner can simply turn pathfinding off anyway.

I know this, and have recommended it to some in the past, and apparently you know this, but I've seen people a damn site smarter than you, on these forums who didn't know this...

And just because it CAN be disabled, doesn't mean pathfinding isn't a laggy piece of crap, it just means that it's a laggy piece of crap that can be disabled, out in the Islands...

But Madlanders are stuck with it...
 

[Theresa Tennyson]

1 hour ago, Klytyna said:

Actually, no, most are NOT that creative, most use griefing tools made years ago, by ancient SL-Fossils, deformers for example are not the problem they once were, now that you can reset your avatars skeleton without having to sit on or click an undeformer, or relog...

But griefers still use that ancient outdated crap.

Oh and good try at derailing from your failed attempts to defend pathfinding into some nonsense about griefers...

Moving on...

I know this, and have recommended it to some in the past, and apparently you know this, but I've seen people a damn site smarter than you, on these forums who didn't know this...

And just because it CAN be disabled, doesn't mean pathfinding isn't a laggy piece of crap, it just means that it's a laggy piece of crap that can be disabled, out in the Islands...

But Madlanders are stuck with it...
 

I just went to a mainland region with a total agent count, including "child" agents, of over 40. (It's a Japanese language welcome area.) The server framerate was already fluctuating somewhat and scripts run were well under 50%. Rezzing a pathfinding character did cause a small drop in simulator framerate and reduced the percentage of scripts run but the effect was barely noticeable overall, even with more than one characters. Using pathfinding characters in a region like that is a bad idea but it hardly brings the region to its knees. In fact, it was running better than a region whose owner has posted about its performance issues which has no pathfinding characters but lots of RLV/BDSM items (that region has fluctuating moments of script time over 28 microseconds.)

Edited September 23, 2018 by Theresa Tennyson

[Macrocosm Draegonne]

15 hours ago, Klytyna said:

No...

Adding "additional scripting languages" would simply add complication to the system, while doing absolutely nothing to reduce server load or "optimise performance".

An utterly useless idea.

Thats just silly, there are quite a few languages already in the mix, and each for specific reasons, that is and will always be the paradigm, I was simply reasserting a standing fact, and extending it to what people can do in-world, IF, and ONLY if, it was better to do so for optimizing and performance, eg. LESS complicated and better running.

15 hours ago, Klytyna said:

RCI (Render Complexity Index) is an ESTIMATE of how difficult it is to render an avatar on YOUR computer... It has almost NOTHING to do with simulation server performance, as the sim servers do NOT render any of your avatar at all, nothing.

Its the best we got, several numbers can be used, are you suggesting someone with a score in the millions is somehow not effecting performance to themselves and others around them, compared to a median number around 150k or less?

15 hours ago, Klytyna said:

It's just an arbitary excuse for insta-booting randomly selected visitors based on YOUR assumptions of what constitutes "excessive complexity".

Why instaboot that seems harsh?  Why not just give people a warning and 5 min to let them decomplex themselves to get better performance and enjoyment out of an experience that specifically requires such things?  This is all related to only those things which would require the utmost performance, not someones house parcel, or a simple shopping sim.  Having a bunch of multi-million scoring avi's running around seems like an easy thing to mitigate for a nice benefit, that everyone would feel.

15 hours ago, Klytyna said:

Goodies back from Project Stupid? Name one? 

Lowered tier and higher transaction fee's an excellent idea that came from Sansar.  ...and most everyone loves it too.  Also we've gotten some developers back from over there I hear.  Im sure as time goes on there will be many mutually beneficial exchanges between the two.

15 hours ago, Klytyna said:

HINT:

If you intend to "fight server side lag", MAYBE you should take the trouble to find out what CAUSES server side lag first...

HINT:

Learing how to delag your builds for example, by eliminating as much of the alpha blended textures from it, avoiding physical set prims or linksets, avoiding pathfinding scripted wandering npc's ( Animesh will no doubt see an explosion of such horrors ) etc.

Eliminating damn awful llSensor scripted "detectors" from the "Olden Times" before the existence of people called "[something] Resident" , such as ancient ARC kickers, script counters, greeter spammers, security orbs etc., also helps.

Great points on that bit, although Server Side Lag is not the only lag to mitigate, and I wasnt only speaking to that.

Also why are you so angry? ...chill out and express yourself, people will hear more, not everyone is immune to such rhetoric like I am, and would probably just ignore everything you're saying, even though there is a lot of value in what you're explaining in various posts ive seen of yours.   ?

[Klytyna]

1 hour ago, Macrocosm Draegonne said:

Lowered tier and higher transaction fee's an excellent idea that came from Sansar.  ...and most everyone loves it too.  Also we've gotten some developers back from over there I hear.  Im sure as time goes on there will be many mutually beneficial exchanges between the two.

Madlanders, a minority, were happy to hear about lower tier, it didn't impress Islanders much...

I haven't seen ANYONE happy about higher transaction fees...

Yeah they transfered a few Project Stupid people to SL, which seems curious, since Project Stupid is a) 100 % incompatible with SL, the whole "reinvent all the wheels from scratch" thing and b) if project Stupid is so wow fab groovy futureness, why are they cutting it's dev team back?

Realistically, we've had NOTHING of any real use or benefit from that white elephant money pit.

But that's a discussion for another thread...

1 hour ago, Macrocosm Draegonne said:

are you suggesting someone with a score in the millions is somehow not effecting performance to themselves and others around them

1. The number of avatars with RCI scores of over 1 million are tiny...  They are a rare and noteworthy sight. The range most people complain about are the 200-400k range.

2. The 1 million club is composed almost entirely of... Rancid old SL-Fossils on system avis covered in flexi-prims, bling jewelery m and scripts-in-every-prim 2006 lagtech...

3. Their performance issues are CLIENT SIDE, for the most part, only their massive 500+ script counts (all those 12 year old colour and resize scripts in their flexi-crap), have any significant effect on Server side performance, so booting on RCI ius...

Completely bloody useless...

 

[ChinRey]

16 hours ago, Theresa Tennyson said:

someone has piffled very authoritatively that scripts don't cause server lag because the time per frame scripts can use is capped.

That was Kyle Linden and he ought to know. But I'm beginning to suspect there may be more to it than that:

14 hours ago, Klytyna said:

Bear in mind that this pathfinding lag-tech crap was dreamed up by people who's idea of testing was to run something for 15 mins on a flat empty green beta grid square for 15 minutes, while watching some test npc walk around a few 5 m cubes...

For all her British understatement, Klytyna may have a point there. Before animesh testing new functions seems to have been a bit low on LL's priority list and may not always have been as thorough as it should have been.

I know of at least two occasions where scripts seem to have had a seriously negative effect on overall server performance. But it's second hand info and I don't have any details so I can't say for sure. It may be worth looking into though.

[Pamela Galli]

5 minutes ago, ChinRey said:

For all her British understatement,

?

[Phil Deakins]

8 hours ago, Macrocosm Draegonne said:

Also why are you so angry? ...chill out and express yourself, people will hear more, not everyone is immune to such rhetoric like I am, and would probably just ignore everything you're saying, even though there is a lot of value in what you're explaining in various posts ive seen of yours. 

Klytyna always posts in bold. It's not anger. She always objects to most stuff she writes about. I suspect she imagines that she's funny, but I don't know. What I do know is that the 'Ignore' option works a treat for me as far as Klytyna is concerned 

[Macrocosm Draegonne]

1 hour ago, Phil Deakins said:

Klytyna always posts in bold. It's not anger. She always objects to most stuff she writes about. I suspect she imagines that she's funny, but I don't know. What I do know is that the 'Ignore' option works a treat for me as far as Klytyna is concerned 

lol, it doesnt bother me, I grew up around verbose people, I'm tempered. ?

[Phil Deakins]

It doesn't bother me either, but she writes long posts and, knowing that there's rarely anything worth reading in them, it's easier for me not to even see them, so that I'm not tempted.

[Macrocosm Draegonne]

9 hours ago, Klytyna said:

1. The number of avatars with RCI scores of over 1 million are tiny...  They are a rare and noteworthy sight. The range most people complain about are the 200-400k range.

2. The 1 million club is composed almost entirely of... Rancid old SL-Fossils on system avis covered in flexi-prims, bling jewelery m and scripts-in-every-prim 2006 lagtech...

3. Their performance issues are CLIENT SIDE

Well that depends, if you're using Black Dragon, the scores are higher a touch, its not as lenient on the onion skined meshes, but yes, it seems there are less people these days overdoing it, and yes some of that classic stuff can be heavy, a prim for every part lol.

We need an SL museum of "Rancid old SL-Fossils on system avis covered in flexi-prims, bling jewelery m and scripts-in-every-prim 2006 lagtech"  but we should make them holograms so not to disturb the living lol. ?

[NiranV Dean]

On 9/24/2018 at 12:01 PM, Macrocosm Draegonne said:

..if you're using Black Dragon, the scores are higher a touch, its not as lenient on the onion skined meshes...

*laughs in multi-quadrillion high complexity values*

Puny human, think you can run from my complexity numbers? *shoots lazers* 5 million complexity for you. *shoots more lazers* 10 million for you *laser-light shows all over the room into a discoball reflecting off at everyone in the room* 50 million complexity FOR ALL OF YOU. *makes pew-pew noises*

No seriously. "a touch", "not as lenient" haha! That's a very "lenient" way of putting it. My complexity calculation will absolutely hate you when you wear anything resembling even remotely an onion-skinned mesh body... not that it could do much about you... except reporting it to scripts and getting you kicked off of places that have ARC checkers there (yes those exist, i had a good laugh when someone reported getting kicked with a million complexity).

But seriously with my Viewer's complexity calculation, 1 million complexity is really easy to get, just wear maitreya or belleza and put on some clothes and the usual stuff like bento hands, head, hair and so on. You'll quickly end up with 1-2 million polygons and since all of is rigged you'll quickly end up with the same (if not more) complexity. I really wouldn't call this a light approach at all. It's harsh and it's extremely punishing just how LL should have made it from the very beginning.

Edited October 1, 2018 by NiranV Dean

[Macrocosm Draegonne]

On 9/30/2018 at 10:51 PM, NiranV Dean said:

*laughs in multi-quadrillion high complexity values*

Puny human, think you can run from my complexity numbers? *shoots lazers* 5 million complexity for you. *shoots more lazers* 10 million for you *laser-light shows all over the room into a discoball reflecting off at everyone in the room* 50 million complexity FOR ALL OF YOU. *makes pew-pew noises*

No seriously. "a touch", "not as lenient" haha! That's a very "lenient" way of putting it. My complexity calculation will absolutely hate you when you wear anything resembling even remotely an onion-skinned mesh body... not that it could do much about you... except reporting it to scripts and getting you kicked off of places that have ARC checkers there (yes those exist, i had a good laugh when someone reported getting kicked with a million complexity).

But seriously with my Viewer's complexity calculation, 1 million complexity is really easy to get, just wear maitreya or belleza and put on some clothes and the usual stuff like bento hands, head, hair and so on. You'll quickly end up with 1-2 million polygons and since all of is rigged you'll quickly end up with the same (if not more) complexity. I really wouldn't call this a light approach at all. It's harsh and it's extremely punishing just how LL should have made it from the very beginning.

lmao.... well it was tweeked a bit from different versions, its not as bad as the first round, I was scoring over 500k in this outfit, and now 315k, I think you gave leeway for hidden invisible tri's.  The same Avi in LL viewer 70k lol.  I am wearing all bento rigged mesh and bento head and all the things, so im not in the millions score, you may be correct about other bodies though, I think Signature really put in some effort to optimize over their previous versions, the score has kept going down with each update.  Catwa said she will update her heads for bakes on mesh when the time comes which will reduce complexity.  I think BOM will reduce onion skinning somewhat, but there will always probably be one onion skin, because really cool things can be done with that. Soapy, dirty, or wet skin, or food on the face etc.

Nothing drives up the complexity score like some old flexy prim hair lol, can easily double the score over all other items combined.

Edited October 2, 2018 by Macrocosm Draegonne

[NiranV Dean]

On 10/2/2018 at 7:28 AM, Macrocosm Draegonne said:

lmao.... well it was tweeked a bit from different versions, its not as bad as the first round, I was scoring over 500k in this outfit, and now 315k, I think you gave leeway for hidden invisible tri's.  The same Avi in LL viewer 70k lol.  I am wearing all bento rigged mesh and bento head and all the things, so im not in the millions score, you may be correct about other bodies though, I think Signature really put in some effort to optimize over their previous versions, the score has kept going down with each update.  Catwa said she will update her heads for bakes on mesh when the time comes which will reduce complexity.  I think BOM will reduce onion skinning somewhat, but there will always probably be one onion skin, because really cool things can be done with that. Soapy, dirty, or wet skin, or food on the face etc.

Nothing drives up the complexity score like some old flexy prim hair lol, can easily double the score over all other items combined.

Pretty sure i've lowered flexi so hard that it's almost akin to a normal prim. (see what i did there?)

Flexi's are by no means bad, sure they take some calculation cycles but not even close as much as LL wants to tell us they do. You can have hundreds, thousands of flexis all at the same time doing their flexi thing their flexi status will pretty much almost have no impact beyond the increased polycount (which is needed to allow them to bend smoother or at all)

[Macrocosm Draegonne]

1 hour ago, NiranV Dean said:

Pretty sure i've lowered flexi so hard that it's almost akin to a normal prim. (see what i did there?)

Flexi's are by no means bad, sure they take some calculation cycles but not even close as much as LL wants to tell us they do. You can have hundreds, thousands of flexis all at the same time doing their flexi thing their flexi status will pretty much almost have no impact beyond the increased polycount (which is needed to allow them to bend smoother or at all)

Ok so its not the flexy its the tri's?  I suppose that makes sense with a flexy hair they would use all prims, one for each small section that moves and that increases the tri's a ***** ton, that must be why its so heavy, its true in all viewers too, some of the classic stuff is really super un-optomized compared to mesh.  I believe sometime soonish we will have fleximesh too right?

[NiranV Dean]

We won't get fleximesh. Although we could technically have it at any time. Different implementations would allow something like physically correct calculated clothing physics, it would be a matter of designating a few bones or collision volumes for that and rigging to them to allow the viewer to apply flexi-similar physical calculations. Though with our unoptimized junk that goes around in SL you'd make short work of your framerate with that.

Also, MOST of the impact comes from the triangle count in combination with them being transformed by the flexi calculation, obviously more triangles will be slower to transform since more calculations have to be done unless it is done in chunks (lets say the entire collision volume is one part only and one uniform calculation is applied, rather than inside this collision volume all vertices can be individually calculated and transformed differently). I've often seen lots of people claim that "triangles is not everything", that's sadly only half-true. Triangles is pretty much everything, more triangles always means more calculations, sometimes linearly, sometimes exponentially, depending on what other things are applying more calculations on top. A static mesh with lots of triangles is not as bad as an animated one because the animated one needs extra calculations to transform all of these triangles but again if it wouldn't use so many triangles to begin with, the calculation for them would be much faster and thus much more trivial. In the end, triangles are your best, most concrete pointer as to whether something is bad or not. Something with many polygons is (outside of it's size) ALWAYS bad, the only exception being that you are using more triangles because the mesh you are creating is covering more things (such as creating an entire level, you can't create more and more parts of the level while staying the same in triangle count unless you take away detail from everywhere else). So whenever you see an avatar with 1 million triangles, you can safely say that it's a completely unoptimized avatar, no exceptions ever. There is no reason a single avatar should ever have that much triangles, not even half of it, not even a quarter. 20K triangles is enough for an entire body with clothes and everything, that is however for games where clothes and body are one whole object, rather than being separated pieces like in SL but taking modularity into account you could estimate roughly 50K triangles for body + clothes. That is 20K for your base body, another 20K for all clothes covering your entire body (facial features included, basically duplicating a skin-tight layer, like a catsuit) and you'd still have 10K triangles free for something else, like items you wear, accessories and that stuff.

Edited October 5, 2018 by NiranV Dean

[Macrocosm Draegonne]

3 hours ago, NiranV Dean said:

We won't get fleximesh. Although we could technically have it at any time. Different implementations would allow something like physically correct calculated clothing physics, it would be a matter of designating a few bones or collision volumes for that and rigging to them to allow the viewer to apply flexi-similar physical calculations. Though with our unoptimized junk that goes around in SL you'd make short work of your framerate with that.

Also, MOST of the impact comes from the triangle count in combination with them being transformed by the flexi calculation, obviously more triangles will be slower to transform since more calculations have to be done unless it is done in chunks (lets say the entire collision volume is one part only and one uniform calculation is applied, rather than inside this collision volume all vertices can be individually calculated and transformed differently). I've often seen lots of people claim that "triangles is not everything", that's sadly only half-true. Triangles is pretty much everything, more triangles always means more calculations, sometimes linearly, sometimes exponentially, depending on what other things are applying more calculations on top. A static mesh with lots of triangles is not as bad as an animated one because the animated one needs extra calculations to transform all of these triangles but again if it wouldn't use so many triangles to begin with, the calculation for them would be much faster and thus much more trivial. In the end, triangles are your best, most concrete pointer as to whether something is bad or not. Something with many polygons is (outside of it's size) ALWAYS bad, the only exception being that you are using more triangles because the mesh you are creating is covering more things (such as creating an entire level, you can't create more and more parts of the level while staying the same in triangle count unless you take away detail from everywhere else). So whenever you see an avatar with 1 million triangles, you can safely say that it's a completely unoptimized avatar, no exceptions ever. There is no reason a single avatar should ever have that much triangles, not even half of it, not even a quarter. 20K triangles is enough for an entire body with clothes and everything, that is however for games where clothes and body are one whole object, rather than being separated pieces like in SL but taking modularity into account you could estimate roughly 50K triangles for body + clothes. That is 20K for your base body, another 20K for all clothes covering your entire body (facial features included, basically duplicating a skin-tight layer, like a catsuit) and you'd still have 10K triangles free for something else, like items you wear, accessories and that stuff.

If im not mistaken there is a fleximesh thing somewhere I saw, perhaps in the beta grid it was a sims name, or one of the beta viewer builds, or both.

Thanks for sharing that detailed information about the calculations, it makes it seem quite possible, yet still needing strict rules if ever implemented.   It could be wicked cool too!

Yeah, too many triangles can be such a waste, often they're providing literally zero visual enhancement too lol, or what scant little they provide can be in the normal map and textures.  That said, it is subjective somewhat, (depending on the avatar) how much tri's would be needed.  Really large or multi limbed things could expand quite a lot.  Seems like most of the added trouble currently is because we dont have Bakes on Mesh yet, that should eliminate all but one of the onion skins.  I think thats a considerable potential  improvement, also if Linden gets more strict with scores it will compel creators to optimize more when possible.

[Qie Niangao]

41 minutes ago, Macrocosm Draegonne said:

fleximesh

Maybe you mean "Animesh"? Seems one could make and script animations that approximate a few strands of basic repetitious "flexi" motion. It would be a lot more restrictive than just hanging more flexiprims onto a linkset, though, and adjusting their flexi parameters.

I could imagine tentacle fetishists wearing a Gorgon head of animesh snakes, for example. (free product idea, hey! ?)

I'm not sure if anybody is still actively discussing other, more general ways for mesh to distort.

7 hours ago, NiranV Dean said:

Flexi's are by no means bad, sure they take some calculation cycles but not even close as much as LL wants to tell us they do. You can have hundreds, thousands of flexis all at the same time doing their flexi thing their flexi status will pretty much almost have no impact beyond the increased polycount (which is needed to allow them to bend smoother or at all)

This is very interesting. It matches my totally unscientific, subjective observations (and nothing better than getting my naive biases confirmed!). I still sometimes add a few flexi strands to mesh hair to achieve a bit of motion. Given the Lab's current accounting, it's easy to use a lot of Land Impact that way if the flexis are linked to the mesh or if, god forbid, one applies to the flexis any materials properties such as masked alpha.