Chat log from the meeting on 2022-08-16

From OpenSimulator

Jump to: navigation, search
[10:58] Andrew Hellershanks: Hello, everyone.
[10:59] Vincent.Sylvester Let's see this week we had more of Ubit beating things into submission hehe
[11:00] Vincent.Sylvester Is much left on that?
[11:00] Ubit Umarov: ofc
[11:00] Ubit Umarov: that was a minor thing
[11:01] Ubit Umarov: changes of the type "useless changes" uff
[11:01] Ubit Umarov: looking to code, clear  time saved
[11:01] Ubit Umarov: in use..  bahh
[11:02] Vincent.Sylvester Not entirely useless if they make things easier to read
[11:02] Ubit Umarov: there was there a O(n^2) now a O(nlogn)

The meeting was moved to Hurliman Plaza due to a problem with the region or its host machine.

[11:09] Vincent.Sylvester That was the OpenSim experience brought to you by the year 2010
[11:09] Vincent.Sylvester lol
[11:10] Andrew Hellershanks: :)
[11:10] Vincent.Sylvester Brings back memories
[11:11] Ubit Umarov: well even yesterday dan told he is considering not renew the rental of that box
[11:11] Ubit Umarov: but today, dev meeting, lbsa etc where on it
[11:11] Ubit Umarov: they say they have a bad router
[11:11] Vincent.Sylvester Used to happen a lot back in the days that I could still reach my toes
[11:11] Ubit Umarov: and keep doing that for like 2 months
[11:12] Vincent.Sylvester Who is providing that box you know?
[11:12] Ubit Umarov: forgot :)
[11:13] Ubit Umarov: let me go start lbsa on another
[11:13] Vincent.Sylvester I have not found a single provider in north america that was decent, most of them seem to be run by some dudes out of some basement or so it feels at least
[11:13] Andrew Hellershanks: I've used servers in NAm run by OVH and didn't have a problem with them.
[11:13] Vincent.Sylvester Yeah they are like the only one worth anything in that price bracket
[11:14] Ubit Umarov: this is a cheaper isp
[11:14] Vincent.Sylvester For a place also home to silicon valley you'd think there would be more decent options
[11:14] Vincent.Sylvester Most are way more expensive and you get barely any hardware in configurations usable for OpenSim type load
[11:15] Jagga.Meredith can't log in as myself. I'm here as my Aviworlds avatar
[11:15] Andrew Hellershanks nods
[11:15] Andrew Hellershanks: np, Jagga.
[11:15] Ubit Umarov: nothing like a tp to change gender :p
[11:15] Ubit Umarov: yeack :p
[11:15] Vincent.Sylvester Rofl
[11:16] Ubit Umarov: remember in 2007 at sl we seen as male, and everyone else seen as female with beard?
[11:16] Ubit Umarov: so irritating :)
[11:16] Vincent.Sylvester Those were the days
[11:16] Vincent.Sylvester Feels like yesterday
[11:16] Ubit Umarov: yeah
[11:17] Ubit Umarov: wel possible it is back again
[11:17] Ubit Umarov: yeah it is back again
[11:17] Vincent.Sylvester I had a region eat my hair last week cause it was running on some ancient release of 7.x, that was fun to restore
[11:18] Andrew Hellershanks: hehe
[11:19] Ubit Umarov: well i have to move outreach to other box
[11:19] Ubit Umarov: uff
[11:19] Andrew Hellershanks: We might as well stay here to save time.
[11:19] Ubit Umarov: yeha lets go on
[11:19] Vincent.Sylvester I seriously considered implementing something to block hypergrid on major version difference to prevent the ticket flood from broken inventory
[11:19] Ubit Umarov: the usuals are here :)
[11:19] Andrew Hellershanks: I've sent IMs to Selby and Kayaker.
[11:20] Selby.Evans I am here
[11:20] Ubit Umarov: well box is unreachable again
[11:20] Andrew Hellershanks: Welcome, Selby.
[11:20] Ubit Umarov: and there is kayaker
[11:20] Andrew Hellershanks: Hello, Kayaker.
[11:21] Kayaker Magic: Hello all!
[11:21] Andrew Hellershanks: Before the forced changed in venue we had started discussing the recent changes to the code. Mostly changes affecting ubODE.
[11:21] Ubit Umarov: ofc the the machine was fine.. had to lose comms when i was telling very important things
[11:21] Ubit Umarov: that  now i forgot
[11:21] Ubit Umarov: :)
[11:22] Vincent.Sylvester Well I am wondering if you have a lot more still to be committed or if I should pull the changes and start testing
[11:22] Andrew Hellershanks: Ubit, you were talking about changes to make clearer and to save time. You mentioned a change from O(n^2) to O(nlogn).
[11:22] Ubit Umarov: :)
[11:23] Ubit Umarov: last week code changes again of the class "useless changes"
[11:23] Kayaker Magic: That sounds like somethint llCastRay needs!
[11:23] Ubit Umarov: ie try to save a few ns here and there
[11:23] Ubit Umarov: this time on avatar/avatar collisions
[11:24] Ubit Umarov: no i do the broadband phase in managed code
[11:24] Ubit Umarov: ie  the AABB boxes checks
[11:25] Ubit Umarov: as i say the algo that was there before was even a O(n * n)
[11:25] Ubit Umarov: now it is a O(n * log n)
[11:25] Ubit Umarov: also has less unmanaged/managed code crossings that are a bit heavy
[11:26] Ubit Umarov: ofc code that here now we are not using at all, bc sitting avatars do not get it LOL
[11:26] Ubit Umarov: if oyu look, physics time per frame is 2us here :)
[11:27] Vincent.Sylvester Said as if collisions and physics are slow these days. I was testing object crossing and have hurled thousands of prims across region borders without much trouble. I guess at those scales it might even become measurable heh
[11:27] Ubit Umarov: sitting avatars have no physics, so physics here does nothing :)
[11:28] Ubit Umarov: what is the relation btw collisions and crossings??
[11:28] Ubit Umarov: duhh :p
[11:28] Vincent.Sylvester Well they do collide with each other before they and while the cross apparently, prim spheres flying everywhere hehe
[11:28] Ubit Umarov: so as i said, this changes i made are only on avatar to avatar collisions,
[11:29] Ubit Umarov: if on physics, so, standing avatars
[11:29] Ubit Umarov: prims are like before..
[11:29] Ubit Umarov: did got a few changes in recent weeks
[11:29] Ubit Umarov: on the dll code changes
[11:30] Ubit Umarov: this last are diferent, because instead of changing code on the dll, i moved it to managed code
[11:30] Ubit Umarov: since with some care JIT does good code
[11:31] Ubit Umarov: but one change that was ok on win,  made mono more tha 20x slower
[11:31] Ubit Umarov: duhhh
[11:31] Ubit Umarov: a situation that used 10us on physics went to close to 300
[11:32] Ubit Umarov: insane
[11:32] Andrew Hellershanks: ouch
[11:32] Vincent.Sylvester Not the first time JIT is being stupid
[11:32] Ubit Umarov: the code change to fix it was commit
[11:32] Ubit Umarov: 2ba53faa5c688c9378e21b108da97e8583089890

* ubode: try ot make mono happy again
[11:32] Ubit Umarov: bahh
[11:33] Andrew Hellershanks: I remember seeing that commit message.
[11:33] Vincent.Sylvester Has a tendency to blow up massively whenever it does fail
[11:34] Vincent.Sylvester Ubit put some aot scripts in there a while back even for those that want to bypass jit entirely for whatever reason
[11:34] Ubit Umarov: yeha rays to avatars maybe the next thing i move to manage only
[11:35] Ubit Umarov: not the avatars are simpler
[11:35] Ubit Umarov: because the way viewers work, they are always vertical
[11:35] Ubit Umarov: evne when we fly and see horizontla, physics is still a veritcla capsule
[11:36] Ubit Umarov: so the colliders for avatar do not need to rotate them
[11:36] Ubit Umarov: saving some cpu also
[11:36] Vincent.Sylvester That's good to hear
[11:36] Ubit Umarov: ofc for now they all do rotate..
[11:37] Ubit Umarov: but on the ray code i may add, that rotation will be out
[11:37] Vincent.Sylvester Over the last few years the memory usage has massively reduced, but cpu load still remains a large factor so every bit helps
[11:37] Ubit Umarov: well cpu load is also operation costs of a grid
[11:37] Ubit Umarov: does matter
[11:38] Ubit Umarov: contrary to what coders told in 2010's
[11:38] Ubit Umarov: specially teachers
[11:38] Ubit Umarov: many did write that a 10% performance improvement is useless
[11:38] Vincent.Sylvester OpenSim's resource profile is quite dissimilar to most others so finding machinery geared towards that can be difficult and costly yeah
[11:38] Andrew Hellershanks: High CPU load says you may be doing something the wrong way.
[11:39] Ubit Umarov: fun how we all so unhappy with 10% inflaction
[11:39] Ubit Umarov: bahh
[11:39] Vincent.Sylvester Stands to reason less memory means less for the cpu to shift around too so it has improved through that as well
[11:39] Ubit Umarov: whatever .. other 10& thing :)
[11:40] Ubit Umarov: but fun... then google started to worry abotu performance
[11:40] Ubit Umarov: telling how xml is a total crap..
[11:40] Vincent.Sylvester There is some form of cpu leak, if that can even be a thing, that I was looking at a few months ago, over the course of a day the average cpu load on that region went up by 3% like clockwork
[11:40] Ubit Umarov: and then all start telling how performance matters
[11:41] Ubit Umarov: isn't ppl so ...  well whatever
[11:41] Ubit Umarov: :)
[11:41] Vincent.Sylvester xml in some cases in OpenSim is 90% structur
[11:41] Ubit Umarov: wlel physics engine is a area where performance always matters
[11:41] Ubit Umarov: ll still did not got that menu
[11:41] Vincent.Sylvester Landtool for example, just a single success keyword and tons of useless structure around it
[11:42] Ubit Umarov: well the new materials will use JSON
[11:42] Andrew Hellershanks nods to Vincent.
[11:42] Ubit Umarov: as if that is a great improvement :)
[11:42] Ubit Umarov: ms also now changed to json
[11:43] Ubit Umarov: on the .net things
[11:43] Vincent.Sylvester I'm somewhat inclined to call for changing notecard storage while making the changes for that so empty notecards stop taking up new assets each time, but of course that would be quite difficult to get compatible
[11:44] Jagga.Meredith can you make empty notecard a pointer to one asset?
[11:45] Vincent.Sylvester An empty notecard has a timestamp in the description, which is part of the data, so a new asset is made
[11:45] Jagga.Meredith ah
[11:45] Vincent.Sylvester Ideally we'd save that part in the asset table rather than the physical data on disk in case of fsassets so it could indeed just point to a single asset
[11:45] Jagga.Meredith you should also look at default box
[11:45] Ubit Umarov: well is more silly than that
[11:45] Vincent.Sylvester But doing that change would break compatibility
[11:46] Ubit Umarov: to make a not card we need to create a new one
[11:46] Ubit Umarov: that creates a new asset with empty data
[11:46] Ubit Umarov: when we do edit and add something, ofc a new asset is done
[11:46] Ubit Umarov: we can't change that.. it is how viwers want it
[11:47] Vincent.Sylvester I cleaned over 30 million notecards out of assets from some script that was creating them
[11:47] Ubit Umarov: ( similar happens with landmarks etc )
[11:47] Ubit Umarov: that is viewer bad design
[11:47] Vincent.Sylvester People seem to forget you can save data on prim description as well instead of using notecards
[11:48] Ubit Umarov: well we can' consider that a option LOL
[11:48] Jagga.Meredith and you can read/write prim description programmaticaly
[11:48] Ubit Umarov: ofc new scritps on inventory do the same
[11:48] Ubit Umarov: also on prims
[11:48] Ubit Umarov: but on prims those assets stay region only longer
[11:49] Ubit Umarov: well not scripts
[11:49] Vincent.Sylvester Yeah but most are not as crazy as me and write a osMakeScript function, so most problematic is still just osMakeNotecard thankfully
[11:51] Vincent.Sylvester I been thinking about adding some stream handler or something to ask a region to find assets and if they cannot find it, it doesn't exist on inventory either, then it is virtually orphaned and can be deleted
[11:51] Vincent.Sylvester Cleaning routines like that could work, but are complex to implement and slow, plus all the overhead from doing that all the time on region comms
[11:52] Vincent.Sylvester Most are just very unaware of these things and never clean their databases or assets and then wonder why they balloon to massive proportions
[11:52] Vincent.Sylvester Running grids is fun...
[11:53] Vincent.Sylvester But fixing, if even possible, notecards is something for whenever we break hg protocol or something
[11:58] Jagga.Meredith have there been any major changes to llParticleSystem since SL last updated the Wiki?  I'm building a tutorial.
[11:59] Ubit Umarov: no idea
[11:59] Ubit Umarov: sl does update their wiki
[11:59] Vincent.Sylvester Should be operating to spec when I last checked that about a year ago
[11:59] Jagga.Meredith last week the topic of documentation was brought up.  I LOVE writing documentation.  What needs doing?
[11:59] Ubit Umarov: cna compare with the old one
[11:59] Jagga.Meredith ok
[12:00] Vincent.Sylvester There is a list on the opensim wiki about what functions are implement and not, that needs checking and changes according to what's currently working
[12:00] Jagga.Meredith ok
[12:01] Vincent.Sylvester Might also provide a good way to see if there are changes to that we need to make or things not working right
[12:01] Jagga.Meredith yup
[12:02] Andrew Hellershanks: We are at the top of the hour. As we had a late start due to the region problem I'll keep this meeting going a bit longer.
[12:02] Andrew Hellershanks: Jagga, you have a couple of questions.
[12:02] Jagga.Meredith that was them
[12:03] Andrew Hellershanks: :)
[12:04] Vincent.Sylvester Speaking of particle system that reminds me I was trying to create a way to spawn particles in a radius and have them move towards a target prim without changing their course
[12:04] Andrew Hellershanks: Jagga, I have noticed in the past some issues with the behaviour of particles. Some effects don't show up the same way as in the past. Not sure if that has anything to do with grid code or whether it is more of a viewer issue.
[12:05] Jagga.Meredith I'm noticing some inconsistenciees
[12:05] Jagga.Meredith hence my question
[12:05] Ubit Umarov: regions only store things
[12:05] Ubit Umarov: many times don't even look to what they are
[12:05] Vincent.Sylvester Well particles are mostly viewer side so yeah even making changes to force them along the vector of origin to target they still move, cause that's viewer rendering
[12:06] Ubit Umarov: in particles guess it does so lsl can work
[12:06] Ubit Umarov: but viewer rendering did change a lot
[12:06] Ubit Umarov: specially recently
[12:06] Ubit Umarov: like ALM is mandatory now
[12:07] Vincent.Sylvester We used to be able to make particle TVs to play media on, think that's gone now too iirc
[12:07] Jagga.Meredith i'll check
[12:08] Ubit Umarov: last viewers have also a performance improvment effort
[12:08] Ubit Umarov: like 2x the fps improvemnt
[12:08] Vincent.Sylvester heh yeah five years too late, but still at least it is finally happening
[12:08] Ubit Umarov: that ofc may mean some things will look diferent
[12:08] Jagga.Meredith yup
[12:08] Andrew Hellershanks nods
[12:09] Ubit Umarov: nad a lot more changes with PBR
[12:09] Ubit Umarov: but have no viewer with those still
[12:09] Vincent.Sylvester Probably the best part of it all given how "flat" the world looks, now we only need materials on terrain textures and some higher resolutions
[12:10] Ubit Umarov: fs beta does have some of the performance changes
[12:10] Ubit Umarov: well the materials a a great need mess in the workings
[12:10] Ubit Umarov: each is a new full asset
[12:11] Ubit Umarov: 4 more textures per face..
[12:11] Ubit Umarov: fun fun
[12:11] Vincent.Sylvester That getting a rework might finally mean they stop randomly disappearing though, which I still cannot track down the cause of
[12:11] Ubit Umarov: no idea why but a prim can have like 14 materials
[12:12] Ubit Umarov: guess for avatars
[12:13] Vincent.Sylvester Gonna need that detail when you stare at your avatars behind all day
[12:13] Andrew Hellershanks: hehe
[12:14] Ubit Umarov: wlel all sçl users where crying for those
[12:14] Ubit Umarov: most have now idea what a materials is, but since all games now have PBR option, SL must have also
[12:14] Vincent.Sylvester Buzzwords are a hell of a drug
[12:15] Ubit Umarov: well lets see what code if any does show up viewers for opensim
[12:15] Ubit Umarov: .. up on...
[12:16] Ubit Umarov: oh hi neovo
[12:16] Neovo Geesink: Hello Ubit, and everyone :-)
[12:16] Ubit Umarov: osgrid is having a meeting here?
[12:16] Neovo Geesink: No, I just wandered in beceause I saw quite a few Avatars here. :-)
[12:17] Ubit Umarov: ok
[12:17] Andrew Hellershanks: np. :)
[12:17] Neovo Geesink: :-)
[12:17] Ubit Umarov: this is our meeting backup region
[12:17] Ubit Umarov: plaza14 box is out again
[12:17] Neovo Geesink: Ah, Thats why. Dev Outreach is downb?
[12:17] Jagga.Meredith is this working towards NFT's?
[12:17] Ubit Umarov: so was lbsa :p
[12:17] Ubit Umarov: i moved it
[12:17] Neovo Geesink: Oopsie....Lbsa is allso acting up again.
[12:17] Jagga.Meredith (PBR)
[12:18] Ubit Umarov: well it was at plaza14 machine also..
[12:18] Neovo Geesink: :-)
[12:18] Neovo Geesink: Then Sandbox Plaza III and Event Plaza are allso affected.
[12:18] Andrew Hellershanks: Someone should define PBR as I'm sure it doesn't mean Pabst Blue Ribbon in relation to OpenSim. :)
[12:18] Ubit Umarov: possible on it laso
[12:18] Ubit Umarov: i only moved lbsa
[12:19] Neovo Geesink: :-)
[12:19] Selby.Evans bye all
[12:19] Ubit Umarov: well started on other box..
[12:19] Andrew Hellershanks: Bye, Selby.
[12:19] Ubit Umarov: cya Selby.Evans
[12:19] Neovo Geesink: Bye Selby
[12:19] Andrew Hellershanks: We are at the hour mark from when we started here. Any last minute questions/comments before we wrap up todays gathering?
[12:19] Jagga.Meredith (rewording) Is the movement towards PBR's an attempt at implementing NFT's?
[12:19] Ubit Umarov: nft ?
[12:20] Jagga.Meredith non-fungible topen
[12:20] Andrew Hellershanks: Non-fungible tokens?
[12:20] Jagga.Meredith yep
[12:20] Andrew Hellershanks: ewww(?)
[12:20] Ubit Umarov: guess is just a better way to represent how light and materials interact
[12:20] Jagga.Meredith oh, ok
[12:20] Ubit Umarov: previus models where kinda "empirical"
[12:21] Jagga.Meredith ah
[12:21] Vincent.Sylvester Physically based rendering is what PBR stands for, it's applying a material to a texture to create things like bricks or wooly carpets without having to create polygons
[12:21] Vincent.Sylvester Like bump and normal maps on steroids
[12:21] Ubit Umarov: pbr os physics based rendering, tries to use physics grounds
[12:21] Jagga.Meredith kewl
[12:21] Vincent.Sylvester NFTs are... stupidity
[12:21] Andrew Hellershanks: Vincent, ty. It helps others know what is being discussed who may be reading the transcript of this meeting.
[12:21] Ubit Umarov: well more play with light and shadows to improve 3d feeling
[12:22] Ubit Umarov: same way as a painter in 2d paper uses shadows to make it look 3d
[12:22] Neovo Geesink: Leaving now... I cannot add value since this is way above my knowledge.
[12:22] Jamie.Jordan Have a great week yall
[12:22] Vincent.Sylvester PBR is the thing everyone now uses in order to make it easier to create nice looking environments and objects by just stringing together different modifiers in something like substance painter
[12:22] Ubit Umarov: :)
[12:22] Andrew Hellershanks: ok, np Neovo. Nice to meet you.
[12:23] Vincent.Sylvester It's commonly used in most popular game engines since the overhead is lower than actual geometry and it can be easily scaled back for lower end systems
[12:23] Ubit Umarov: thkn is that is another thing not for amatuers
[12:23] Ubit Umarov: its now a simple thing
[12:24] Ubit Umarov: like 4 of 6 textures per face..
[12:24] Ubit Umarov: hmm 4 ?
[12:24] Vincent.Sylvester Oh it is a pain in the neck to figure them out and most of the programs that can create them are paid software there only exist a handful that are free
[12:24] Ubit Umarov: if you do use that, you just flood a region
[12:24] Ubit Umarov: need to know very well how to use and when
[12:25] Ubit Umarov: for most ppl is just BETTER to use a higher res mesh than a bump map
[12:25] Vincent.Sylvester The upside is we only need to provide them, viewer does the rendering, at least we don't have to maintain that shader code
[12:26] Ubit Umarov: well that means you will need to understand what a LL material is
[12:26] Ubit Umarov: bc sure will not be as a autodesk one
[12:26] Ubit Umarov: or blender..
[12:27] Andrew Hellershanks nods
[12:27] Ubit Umarov: ie render like..
[12:27] Ubit Umarov: sl can't make skies render the same way
[12:27] Ubit Umarov: bahh
[12:28] Kayaker Magic: How are bump and specular maps stored now? Hacked into the Extra blob?
[12:29] Ubit Umarov: well opensim made that a extra material also
[12:29] Ubit Umarov: but they are stored as textures extensions at sl
[12:29] Kayaker Magic: Is that an LL structures? Or did the guy who wrote it for opnensim just not know how to do relational databases?
[12:30] Ubit Umarov: the new ones will be on the Extra blob
[12:30] Ubit Umarov: ir the uuids
[12:31] Andrew Hellershanks: On that note let's call this meeting to a close.
[12:32] Andrew Hellershanks: We will meet at the usual location next week unless we have a problem with the region like we did today. Thank you all for coming. See you again next week.
Personal tools
About This Wiki