Chat log from the meeting on 2012-07-03

From OpenSimulator

Revision as of 12:20, 3 July 2012 by Justincc (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search
[11:43] Dahlia Trimble: hi
[11:43] Justin Clark-Casey: well, knowing if there are high physics times would help
[11:43] Nebadon Izumi: ya its actually ok if we have less than 15 avatars
[11:43] Justin Clark-Casey: if not, then you could proabbly rule out physics
[11:43] Justin Clark-Casey: hi dahlia
[11:44] Nebadon Izumi: once we cross a certain threshold of avatars
[11:44] Nebadon Izumi: it just explodes
[11:44] BlueWall Slade: Hello Dahlia
[11:44] Nebadon Izumi: everyone complains prims are not loading
[11:44] Nebadon Izumi: cant see any avatars
[11:44] BlueWall Slade: those stats should give some clues too.
[11:44] Justin Clark-Casey: It wouldn't enormously surprise me if some other process is soaking up all the sim time
[11:44] BlueWall Slade: and you can use the in-world debugginh things
[11:44] Justin Clark-Casey: like lots of invalid group requests or something
[11:44] Nebadon Izumi: but why doesnt Lbsa do it
[11:44] Nebadon Izumi: we can have 40 avatars on lbsa
[11:44] Nebadon Izumi: same box
[11:44] Justin Clark-Casey: I do not know
[11:44] Nebadon Izumi: never ever once has done it on Lbsa
[11:45] Nebadon Izumi: but everytime we use Bade it happens
[11:45] Bea Witched: flying isn't allowed on lbsa so height is restricted
[11:45] Nebadon Izumi: 4-5 meetings in a row it happened
[11:45] Nebadon Izumi: finally we just moved it here for first time this month
[11:45] Nebadon Izumi: we had 25 avatars
[11:45] Nebadon Izumi: stats were perfect
[11:45] Nebadon Izumi: literally no lag
[11:45] Nebadon Izumi: you would think if any region was going to have issues
[11:45] Nebadon Izumi: it would be this one
[11:46] Nebadon Izumi: lol
[11:46] Nebadon Izumi: bade has no neighbors
[11:46] Nebadon Izumi: its completely isolated
[11:46] Nebadon Izumi: low prim, low scripts
[11:46] Nebadon Izumi: so bizzare
[11:46] Richardus Raymaker: thats same as with my physical train test (i stopped with that) 1 train no timeings. when i rezz aseconds train phyics time jump up to 180,ms. like am used toin the past with 1 train
[11:46] Justin Clark-Casey: well, not seeing objects is often a synptom of bad network comms
[11:46] Justin Clark-Casey: object update packets being completely dropped and the resend system failing
[11:46] Nebadon Izumi: ya this felt more like a hung thread or something
[11:47] Nebadon Izumi: and the only thing i can really pinpoint and say its completely different than the other plazas
[11:47] Nebadon Izumi: is that we meet above 100m above ground
[11:47] Richardus Raymaker: i wish group IM did not be handles by region.
[11:47] Nebadon Izumi: otherwise nothing is different
[11:47] Nebadon Izumi: anyway i wasnt looking for a solution
[11:47] Nebadon Izumi: i just wanted to present the idea
[11:47] Richardus Raymaker: hi dahlia
[11:47] Nebadon Izumi: i'll try to gather more info
[11:47] Justin Clark-Casey: yeah
[11:48] Dahlia Trimble: object updates dropping? is this a prim/texture heave region?
[11:48] Dahlia Trimble: *heavy
[11:48] Nebadon Izumi: no Dahlia, its quite low prims low textures
[11:48] Nebadon Izumi: its maybe 7000 prims
[11:48] Nebadon Izumi: few 100 textures maybe
[11:48] Dahlia Trimble: so much for that idea :/
[11:48] Nebadon Izumi: ya ive tried everyting i could think of
[11:49] Nebadon Izumi: i literally inspected every script on the sim
[11:49] Nebadon Izumi: i went through the database
[11:49] Nebadon Izumi: and located every script
[11:49] Nebadon Izumi: to make sure it wasnt some kind of visitor scanner or something
[11:49] Nebadon Izumi: some kind of radar
[11:49] Nebadon Izumi: i couldnt find any heavy scripts
[11:49] Nebadon Izumi: its all like sit scripts and hover text
[11:49] Richardus Raymaker: also checked for lost AO's and other objects people lose ?
[11:49] Nebadon Izumi: ya, i couldnt find anything unusual
[11:50] Nebadon Izumi: thats not on all the other plazas
[11:50] Justin Clark-Casey: anyway, any other opensim topics anybody wants to bring up today?
[11:50] Mike Kayaker: I have a question about timers...
[11:50] Nebadon Izumi: hrmm I dont think I have anything else myself
[11:51] Nebadon Izumi: shoot Mike
[11:51] Mike Kayaker: I seem to be running into a script resource I don't understand
[11:51] Richardus Raymaker: well justin like i sayed in mailinmglist. new version seems to work betetr . sofar i have doen tests on the sim no crashing anymore
[11:51] Dahlia Trimble: I need advice on adding properties to SOG
[11:51] Mike Kayaker: If I have a bunch of scripts running that all have a short (1 second timer)
[11:51] Dahlia Trimble: like for pathfincing, whether an object is a walkable
[11:51] Mike Kayaker: the FPS starts to drop
[11:52] Dahlia Trimble: how it affects database and oars
[11:52] Mike Kayaker: I want to have two kinds of timers I guess,
[11:52] Mike Kayaker: those that I care if the timer is on time,
[11:52] Richardus Raymaker: strange mike. i have used 0.5s timers (no llSLeep) and never seen frame drops
[11:52] Mike Kayaker: and others that I don't care
[11:52] Justin Clark-Casey: I would find it very surprising if fps drops due to script timers
[11:53] Justin Clark-Casey: unless you're using hundereds and hundreads of them and never removing them
[11:53] Mike Kayaker: Is there some way I can allow the don't care timers to volunteer to be time sliced?
[11:53] Mike Kayaker: like calling llSleep(.0001) or something?
[11:53] Nebadon Izumi: i dont think so
[11:53] Richardus Raymaker: llSLeep.00001) never works
[11:53] Nebadon Izumi: there is a limit how low sleep can go
[11:53] Dahlia Trimble: use not_at_target() for a per-frame event
[11:53] Nebadon Izumi: its 0.01?
[11:53] Nebadon Izumi: or 0.05?
[11:54] Nebadon Izumi: i cant remember
[11:54] Mike Kayaker: Well, of cource I do a vector calculation and a resize or something on every tick
[11:54] Nebadon Izumi: maybe its even less
[11:54] Richardus Raymaker: fastest i have seen is around 0.01 maby but i resits to use llSLeep at all. only in rare cases
[11:54] Justin Clark-Casey: smallest sleep would be 1ms (0.001)
[11:54] Nebadon Izumi: hrmm i ok didnt think it went that far
[11:54] Nebadon Izumi: i thought it was 0.01
[11:54] Mike Kayaker: but not-at-target would slow down as the FPS goes down...
[11:54] Richardus Raymaker: i agree with nebadon. i have not seen any difference betwen 0.01 and lower
[11:55] Dahlia Trimble: why would you want to update anything faster than frame rate?
[11:55] Richardus Raymaker: at_target stuff. thats still so unreliable for me
[11:55] Mike Kayaker: Actully, I don't care how long the slep sleeps, just want timers on the other side of the sim to stay accurate when have a lot of ddon't care timers running
[11:55] Justin Clark-Casey: yes. In fact, anything less than 0.001 will get trunacated to 0, no sleep at all
[11:56] Justin Clark-Casey: mike: I wouldn't be surprised if timers are not very accurate atm
[11:56] Mike Kayaker: thing is, I build one item, timers work good enough,
[11:56] Justin Clark-Casey: I think the way they operate is somewhat inaccurate
[11:56] Richardus Raymaker: Justin, timers, sensors, at_target are never 100% save for me
[11:56] Dahlia Trimble: its not a real-time simulation running on a real-time os
[11:56] BlueWall Slade: working on some stuff a couple of months back - looks like the events priority is burried below other things
[11:56] Mike Kayaker: then I rez a bunch of the item, and things get very strange
[11:57] BlueWall Slade: at_target was delayed for a long time
[11:57] Dahlia Trimble: not at_target(), use not_at_target() *note underscores*
[11:57] BlueWall Slade: when it finally showed up (way past the target point) it spammed the output pretty bad with many lines
[11:58] Justin Clark-Casey: the problem is timers are run on the same loop as a lot of other lsl stuff
[11:58] Justin Clark-Casey: if there's a lot going on then I suspect they could be delayed
[11:58] Mike Kayaker: What resoure am I using up that causes the timers to slow down?
[11:58] Justin Clark-Casey: and minimum timer resolution on this loop is 100ms
[11:58] Richardus Raymaker: lsl is to bad not really multitasking
[11:59] Dahlia Trimble: any event has to be queued anyway
[11:59] Justin Clark-Casey: url listeners, sensors, dataserver requets
[11:59] Justin Clark-Casey: and yes, then there's a delay/inaccuracy with timer event queueing
[11:59] Simulator Version v0.5 shouts: OpenSim 0.7.4 Dev          1f34c82: 2012-06-13 04:05:02 +0100 (Unix/Mono)
[11:59] Richardus Raymaker: i use sensor(s0 to, but the maby get called once every second or longer
[12:00] Justin Clark-Casey: I suspect this area could be improved
[12:01] Mike Kayaker whispers: So is there a way to do a lag-friendly timer? Like get a timer evey second, but not slow down other scripts if the sim is buisy?
[12:01] Dahlia Trimble: probably not if the sim is busy
[12:01] Mike Kayaker: I mean, a timer that does allow itself to get slowed down, so other timers will not get delayes.
[12:02] Justin Clark-Casey: I'm really not sure where the slow down is happening
[12:02] Richardus Raymaker: Mike TimerEvent is possible most save, but not reliable if you ned to have strict timeings
[12:02] Justin Clark-Casey: this would require rearchitecting the way that timers are handled
[12:02] Justin Clark-Casey: possibly so they sit on a separate thread rather than with every other event
[12:02] Justin Clark-Casey: or even setup actual system resource timers
[12:02] Mike Kayaker: Every timer event in my sim starts getting delays.
[12:02] BlueWall Slade: when we can build on C# 5.0 we should do that anyway
[12:02] Richardus Raymaker: Mike , do you have physics on the sim ?
[12:03] Dahlia Trimble: given that all events are queued I'd say its near impossible
[12:03] Justin Clark-Casey: well this doesn't need c# 5.0, bluewall
[12:03] Mike Kayaker: Yes, but nothing physical rezzed
[12:03] Richardus Raymaker: ehh ?
[12:03] Richardus Raymaker: phsiscal but not physical.
[12:03] Mike Kayaker: The sim has ODE, but I don't have any physical objects in the sim ATM
[12:03] Richardus Raymaker: ok, thats what i mean
[12:04] Richardus Raymaker: its a dedicated server ? not VPS ?
[12:04] Mike Kayaker: Im not sure, VPS???
[12:04] Justin Clark-Casey: anyway, I'd like to answer dahlia quickly before she disappearas
[12:05] Richardus Raymaker: go justin
[12:05] Justin Clark-Casey: dahlia: at the moment, extended sog is difficult. In fact, my favourite solution would be to serialize a separate piece of data to the region, a bit like the moap xml
[12:05] Justin Clark-Casey: the alternative is to add more hard coded fields to prims or primshapes
[12:05] Dahlia Trimble: ewwww
[12:06] Dahlia Trimble: its a SOG property
[12:06] Justin Clark-Casey: which has to be persisted?
[12:06] Dahlia Trimble: yes
[12:06] Justin Clark-Casey: I know, unofrutnately there is no separate table for sogs
[12:06] Dahlia Trimble: ok how about to a prim with parent id of 0?
[12:06] Justin Clark-Casey: if it was a serialization one could conceivably add it to the root prim only
[12:07] Dahlia Trimble: so SOP then?
[12:07] Justin Clark-Casey: when you stat serializating properties you get into the expense of that, thuogh I personally don't think it's a huge issue
[12:07] Justin Clark-Casey: the problem is putting this stuff on the region itself
[12:07] Justin Clark-Casey: you have to have a field in the regions db tables
[12:08] Dahlia Trimble: if I add one how do I handle migrations and oars?
[12:08] Justin Clark-Casey: you can add a migration like the existing ones to add an extra field. Serialization covers oars, inventory, attachment teleports, etc.
[12:09] Dahlia Trimble: kk
[12:09] Justin Clark-Casey: that needs an explicit change to the serialization/deser code in OpenSim/Region/Frametwork/Scene/Serialization/SceneObjectSerializer
[12:10] Justin Clark-Casey: I would ask that this isn't put directly into master in case there are any issues - something in a branch could be more easily checkable
[12:10] Dahlia Trimble: I'll ask you in IRC before adding anything
[12:11] Dahlia Trimble: anyway it would be fields that LL is adding but I dont think they use the normal protocols to share them with viewers
[12:11] Justin Clark-Casey: ok, I would certainly like the opportunity to think about the future data format
[12:12] Justin Clark-Casey: dahlia: using caps instead of udp?
[12:12] BlueWall Slade: does OAR get the V3 environment settings ATM ?
[12:12] Justin Clark-Casey: bluewall: I don;'t think so
[12:12] Dahlia Trimble: looks more like some proprietary havok encoding
[12:12] Dahlia Trimble: over caps
[12:12] Justin Clark-Casey: uhhhhhhh :)
[12:12] Dahlia Trimble: ya
[12:13] Richardus Raymaker: brrr, havok :)
[12:13] Dahlia Trimble: but still the properties need to exist
[12:13] Justin Clark-Casey: yes
[12:13] Justin Clark-Casey: well, maybe they could just be new fields - that would just be historical practice after all
[12:13] Justin Clark-Casey: just a bit of a pita
[12:14] Dahlia Trimble: need a process for this kind of thing
[12:14] Justin Clark-Casey: yes
[12:14] Justin Clark-Casey: ok, I need to go and get some stuff done
[12:14] Dahlia Trimble: bye :)
[12:14] BlueWall Slade: bye Justin
[12:14] Bea Witched: be well Justin
[12:14] Dahlia Trimble: me too lol... bye all :)
[12:15] Mike Kayaker: thnks all
[12:15] BlueWall Slade: bye Dahlia
[12:15] Justin Clark-Casey: mike: do you ever compile opensim or do you stick wityh binaries?
[12:15] Mike Kayaker: binaries
[12:15] Nebadon Izumi: see you guys
[12:15] Nebadon Izumi: thanks for coming
[12:15] Richardus Raymaker: bye justin
[12:15] Bea Witched: ty have a good day all
[12:15] Justin Clark-Casey: ok, I was going to suggest a tweakable paramter but I'm not sure how much it would help
[12:15] BlueWall Slade: I need to take off too - take care all
[12:15] Justin Clark-Casey: maybe I should expose it as a config thing
[12:15] Mike Kayaker: next time
[12:16] Richardus Raymaker: bye blue
[12:16] Justin Clark-Casey: l
[12:16] Justin Clark-Casey: k
[12:16] Justin Clark-CaseyJustin Clark-Casey waves
Personal tools
About This Wiki