Chat log from the meeting on 2012-09-04

From OpenSimulator

Jump to: navigation, search
[11:06] Richardus Raymaker: Now i still have only a weird memory usage in my sandbox... :) its getting beter. but its going slowly up again
[11:06] Arielle Popstar: it kept looping synchronous restobjectrequester before
[11:06] BlueWall.Slade Hi Justin
[11:06] Richardus Raymaker: uhmm unassisted flight is still around 160 meters sofar i know
[11:07] VivK Lowlag: hello JCC
[11:07] Justin Clark-Casey: hi folks
[11:07] Arielle Popstar: Hi Justin
[11:07] Richardus Raymaker: Arielle, do you have appdomain = true ?
[11:07] Richardus Raymaker: otake. can you rebake
[11:07] Arielle Popstar: whatever is default
[11:07] Richardus Raymaker: default = false. and that dont work with windows always
[11:07] Richardus Raymaker: then it dont boot.
[11:08] Richardus Raymaker: ok, it dont boot if you have to many scripts running
[11:08] Richardus Raymaker: thats beter otaku
[11:08] Justin Clark-Casey: I thought that appdomain = false worked better with windows these days
[11:09] VivK Lowlag: :)
[11:09] Justin Clark-Casey: hmm, don't see nebadon on skype which is unusual
[11:09] Richardus Raymaker: hmm not sure. it seems to fail for some. but i dont like the true. i wish someone could hint me when my memory still go up in my sandbox. and i hzave now only a few scripts
[11:09] Arielle Popstar: doesnt seem to be Justin
[11:10] Arielle Popstar: tryiong it other way now
[11:10] BlueWall.Slade Teravus was in IRC for a little while this morning
[11:10] Justin Clark-Casey: heh, he probably clicked the wrong button by mistake
[11:10] BlueWall.Slade haa
[11:11] BlueWall.Slade neb isn't in IRC either
[11:11] Justin Clark-Casey: well, any opensim topics today people want to particularyl discuss?
[11:11] Justin Clark-Casey: that is unusual
[11:11] Richardus Raymaker: well, we already agreed warp3d is memory hungry. i have disabled it.
[11:11] BlueWall.Slade it's good to have another release behind us :)
[11:11] Justin Clark-Casey: yes
[11:12] BlueWall.Slade that didn't take too long
[11:12] Andrew Hellershanks: I have one minor optimization for Warp3D that I haven't placed in mantis yet
[11:12] Richardus Raymaker: i still cant find the trouble why memory usage go steady up in sandbox..
[11:12] Andrew Hellershanks: I've seen a *huge* array in warp3d that gets allocated. It may be main reason its memory hungry.
[11:12] BlueWall.Slade Rich - did you try to disable scripts?
[11:12] Richardus Raymaker: only 7 scripts running
[11:13] BlueWall.Slade one script could do it
[11:13] Justin Clark-Casey: andrew: well , provided those aren't permanently held for some reason they shouldn't conitually inflate memory
[11:13] Richardus Raymaker: Well, if i cant run scipts i cant test the object
[11:13] Richardus Raymaker: but i can try it this night
[11:13] Andrew Hellershanks: justin, no. But its not very efficient.
[11:13] BlueWall.Slade sometimes you have to do that - divide and conquer
[11:14] BlueWall.Slade I had a region eat over 7G memory after an upgrade back in the 0.6.9 days
[11:14] Richardus Raymaker: afraid some (lsl) script commando leaks..
[11:14] BlueWall.Slade it was one script
[11:15] Richardus Raymaker: This scripts i wrote myself.
[11:15] BlueWall.Slade didn't do that before the upgrade and a small change fixed it
[11:15] BlueWall.Slade disable them all for a while and see if the problem is still there
[11:15] BlueWall.Slade if so, then disable them one by one
[11:16] Richardus Raymaker: i do that tonight after i restart it fresh
[11:16] Key Gruin: hi folks
[11:16] Richardus Raymaker: hi key
[11:16] Justin Clark-Casey: hi key
[11:16] BlueWall.Slade Hi Key
[11:16] Arielle Popstar: Hi Key
[11:16] Justin Clark-Casey: key: any idea if nebadon is around?
[11:16] Key Gruin: hi Justin Rich, all
[11:16] VivK Lowlag: hi Key
[11:16] Key Gruin: no he's on the road Justin
[11:16] Richardus Raymaker: brb, let me disable it and see what happens
[11:16] Key Gruin: said he will check in late tonight
[11:17] Justin Clark-Casey: ok, thanks
[11:17] Arielle Popstar: ok thx Rich. Setting it to true allowed it to start finally
[11:17] Arielle Popstar: that region has 697 n scripts
[11:17] Justin Clark-Casey: arielle: which version of opensim is this?
[11:17] Arielle Popstar: the other that started up with setting at false has very few
[11:18] Arielle Popstar: osgrid aug 20 release
[11:18] Justin Clark-Casey: ok
[11:18] Justin Clark-Casey: interesting, I don't seem to have any issues on widnows starting up a regio nwith 2200 scripts
[11:18] Justin Clark-Casey: with appdomain = false, though I don't have the test stuff to hand at this point
[11:18] Justin Clark-Casey: but it sounds like there are still issues with appdomain= false on windows
[11:19] Richardus Raymaker: 2200 scipts. hmm then you need to wait for my mega. where already happy with 400 scipts
[11:19] Justin Clark-Casey: arielle: are you 32 or 64 bit?
[11:19] Arielle Popstar: is that region connected to osgrid |Justin?
[11:19] Arielle Popstar: 32
[11:19] Justin Clark-Casey: no, unfortuantely contains content which is not mine to make aailable
[11:19] Justin Clark-Casey: available
[11:19] Justin Clark-Casey: arielle:: so what happens when you try to start up the regio nwith appdomain = false?
[11:20] Arielle Popstar: i think that makes a diff because the looping message is slow request
[11:20] Justin Clark-Casey: this is a regio nconnected to osgrid?
[11:20] Arielle Popstar: yes
[11:20] Justin Clark-Casey: so what happens when you try to startup?
[11:21] Richardus Raymaker: about slow get's i see lowts of slow gets now with saveing IAR file. and that i started 5 hours ago
[11:21] Richardus Raymaker: slow get to assets
[11:22] Simulator Version v0.5 shouts: OpenSim 0.7.5 Dev          15d5f3d: 2012-09-04 00:11:14 +0100 (Unix/Mono)
[11:22] Richardus Raymaker: how close is the 08-20 release to the final 0.7.4 ?
[11:23] Justin Clark-Casey: very
[11:23] Richardus Raymaker: ok. then i dont expect much big updates. thx
[11:23] Andrew Hellershanks: 0.7.4-post-fixes is the branch to use for 074 with any additional fixes that may be applied?
[11:23] Justin Clark-Casey: in fact, 08-20 would be after 0.7.4 because 0.7.4 was already branched at that point
[11:23] Justin Clark-Casey: andrew: yes
[11:24] Andrew Hellershanks: k. Seems you've standardized on post-fixes for after release branch which is good.
[11:24] Justin Clark-Casey: though 0.7.4 would then contain some fxies that went in after 08-20, so, kind of before and behind
[11:25] Justin Clark-Casey: any other opensim topics for today?
[11:26] Andrew HellershanksAndrew Hellershanks doesn't have anything
[11:26] Richardus Raymaker: i disabled scipts in my sandbox. and the memory seems to keep stable now. but need to wait a few hours to be sure
[11:26] Andrew Hellershanks: Is there a particular LSL script that seems to trigger a memory leak?
[11:26] Andrew Hellershanks: or particular function that is
[11:27] Justin Clark-Casey: richardus: are you r gunther?
[11:27] Richardus Raymaker: the 4 train scripts ar almost the same, only the car and train are a bit different.
[11:27] Richardus Raymaker: Yes justin
[11:27] Richardus Raymaker: :)
[11:27] Justin Clark-Casey: so, I would be really surprised if memory leaks are created by variaables in functions
[11:27] Richardus Raymaker: i use sensor in that scipt and some lists
[11:28] Arielle Popstar: @Justin, restarted the prolem region with appdomain=false and now it started up fine
[11:28] Justin Clark-Casey: since I think they translate directly to local variables inside c# methods but I can't be totally sure
[11:28] Arielle Popstar: maybbe it needed the one good load
[11:28] Justin Clark-Casey: did you try echo 3> /prox/sys/vm/drop_caches as melanie suggested?
[11:29] Richardus Raymaker: Well it feels like there's memory reserved somewhere. but not released. and the CG seems to do nothing with it to
[11:29] Richardus Raymaker: Justin, its windows
[11:29] Justin Clark-Casey: ok
[11:29] Richardus Raymaker: thats why i cleaned my sandbox back the only what i really need for testing 1 train and 3 cars.
[11:30] Richardus Raymaker: it still goes up
[11:30] Richardus Raymaker: it only goes lower up i now only have 7 scipts, before it where 300
[11:30] Richardus Raymaker: slower ^
[11:32] Key Gruin: ;
[11:32] Richardus Raymaker: unless, llGetObjectDetails have some leak ?
[11:32] Justin Clark-Casey: unfortunately, it's very difficult to act on subjective information
[11:32] Richardus Raymaker: i know justin. its very hard to figure out.
[11:33] Justin Clark-Casey: one really needs a specific simple-as-possible testcase that demonstrates a specific problem
[11:33] Richardus Raymaker: try to find some straw to grab for a direction
[11:33] Justin Clark-Casey: the only thing I can think of immediately with scripts is if one expands one of its variables forever
[11:33] Justin Clark-Casey: e.g., continually adds to an array
[11:33] Justin Clark-Casey: because xengine cannot enforce any memory limits, that could just keep going forever
[11:33] Hiro Protagonist: a lot of times, distilling the complex problem down into a simpler one reveals a scripting problem as the source of ones troubles
[11:33] Justin Clark-Casey: but then that would affect specific scripts, not scripts in general
[11:33] Hiro Protagonist: though I doubt that would be the case of a memory leak
[11:34] Richardus Raymaker: thats why i moved the list a = above default. and clear every time i need the list everything to []
[11:34] Andrew Hellershanks: Run llGetObjectDetails in a loop with no other scripts in an empty region and see if memory usage goes up.
[11:34] Hiro Protagonist: oth it might, considering lists are involved
[11:34] Richardus Raymaker: Indeed. Good change something else triggers it in combination with what am doing
[11:34] Richardus Raymaker: hi hiro
[11:35] BlueWall.Slade I think the script I had was just setting primitive parameters over and over
[11:35] Richardus Raymaker: its to bad you cant delete a list complete after use. li llListRemove or so
[11:35] Justin Clark-Casey: it would be nice to hav ea script engine that could enforce these limits but unforutnatley it would require something with a very different architecture
[11:35] Richardus Raymaker: well if you move object you need to do that bluewall
[11:36] Richardus Raymaker: i use the "llSetLinkPrimitiveParamsFast(" you told me blue. safed me lots of problems
[11:37] Andrew Hellershanks: no automatic CG in LSL or its just a bit harder to do properly?
[11:37] Justin Clark-Casey: gc should be done for scripts just as it is anywhere else
[11:37] Hiro Protagonist: LSL compiles to c# assemblies
[11:37] Richardus Raymaker: i think i need to moe the train bakc in inventory and make litte scipts with every command sperate in timer loop
[11:37] Arielle Popstar: 1 problem i been seeing is that looking at a neighbours sim increases my ping and fps 10 fold
[11:37] Justin Clark-Casey: yes, xengine compiles to c#
[11:37] Hiro Protagonist: when executing they are binaries
[11:38] Arielle Popstar: could that happen because of a bad script?
[11:38] Richardus Raymaker: well lucky i have not really neighborns around my sandbox. not one i see
[11:38] Andrew Hellershanks: Arielle, lots of stuff can happen if you have a bad script :-)
[11:38] Justin Clark-Casey: arielle: probably not - I would suspect that the neighbour might be on a bad network link
[11:38] Justin Clark-Casey: arielle: do you think that could be the case?
[11:38] Hiro Protagonist: Arielle that's because of all the data that has to be sent to your viewer when they come into your view
[11:38] Arielle Popstar: no, they are on a lot better one then me
[11:38] Hiro Protagonist: from the neighbor region
[11:39] Arielle Popstar: i know my neighbours setup well
[11:39] Hiro Protagonist: right but even if (especially if?) it's all up to par, all of their content has to be sent to your viewer
[11:39] Justin Clark-Casey: arielle: what do you mean by looking - when your avatar turns in that direction?
[11:39] Hiro Protagonist: I reckon via the child agent?
[11:39] Richardus Raymaker: i only know that you get lag when you fly and you approch a not rezzed neighborn.. but thats explainable.
[11:39] Arielle Popstar: yes Justin
[11:40] Justin Clark-Casey: Do they have a lot of content on their sim?
[11:40] Hiro Protagonist: I doubt you cache that data either
[11:40] Arielle Popstar: soon as i turn away my ping drops to 60
[11:40] Hiro Protagonist: since it isnt on your region
[11:40] Arielle Popstar: fair bit
[11:40] Richardus Raymaker: Thats a weird one arielle..
[11:40] Richardus Raymaker: ohh ping drop.. aah
[11:41] Arielle Popstar: or if i reduce draw to less then the distance to their sim
[11:41] Justin Clark-Casey: might even be lag on the viewer side if its having to process more content when you look in certain directions
[11:41] Hiro Protagonist: Arielle that last bit supports my assertion ;)
[11:41] Hiro Protagonist: precisely Justin :)
[11:42] Arielle Popstar: but then when i am on any sim with that much content i should experience the same?
[11:42] Arielle Popstar: though tbh i see 500 ms here
[11:42] Justin Clark-Casey: if you are looking back into a sim with lots of content on as well, then maybe
[11:42] Arielle Popstar: good thing i am sitting down ;)
[11:42] Justin Clark-Casey: 500 ms is very high
[11:42] BlueWall.Slade have to take off - will stay parked here
[11:42] Justin Clark-Casey: I get 200-300ms and I'm on the other side of the atlantic
[11:42] BlueWall.Slade see ya next week
[11:43] Richardus Raymaker: i see that only when a sim is,loading content or its on terrible connection
[11:43] Justin Clark-Casey: bye bluewall
[11:43] Arielle Popstar: Tc Blue
[11:43] Hiro Protagonist: you will find that any sim with no neighbors is dramatically faster than one with a neighbor, and really the oamount of content wont impact it as much as you would expect
[11:43] Hiro Protagonist: tc Blue!
[11:43] Justin Clark-Casey: arielle: what kind of gfx card do you have?
[11:43] Arielle Popstar: on a laptop atm
[11:43] Richardus Raymaker: bye blue
[11:43] Arielle Popstar: dual process 4 gb
[11:44] Arielle Popstar: ati card (covers head )
[11:44] Justin Clark-Casey: what typ eof ati card?
[11:44] Arielle Popstar: checking
[11:46] VivK Lowlag: .
[11:46] Arielle Popstar: radeon express 1200 series
[11:47] Hiro Protagonist: are you on a mac?
[11:47] Richardus Raymaker: now you can cove ryour head. :)
[11:47] Arielle Popstar: no, amd
[11:47] Hiro Protagonist: nods
[11:47] Arielle Popstar: LG machine
[11:47] Richardus Raymaker: LG, cant hear that brand. :O
[11:48] Arielle Popstar: looking behing me i have a ping of 627 ms and fps of 1.6
[11:48] Richardus Raymaker: but it dont explain to my why graphics card have something todo with ping times
[11:48] Richardus Raymaker: 139ms 17fps
[11:48] Arielle Popstar: fps and ping are interralated
[11:48] VivK Lowlag: strange my current connection is 3g and my ping is 128ms
[11:49] Justin Clark-Casey: that's a pretty low performance card really
[11:49] Arielle Popstar: yes not a great one
[11:49] Arielle Popstar: but still should be adequate
[11:49] Justin Clark-Casey:
[11:49] Richardus Raymaker: you on high or ultra ? i use 256 draw distance
[11:50] Arielle Popstar: on medium
[11:50] Justin Clark-Casey: well, sl and opensim even more do not/cannot use gfx cards efficiently
[11:50] OtakuMegane Desu: ^ that
[11:50] Justin Clark-Casey: so to be honest, it doesn't surprise me if you get a drop in performance when looking onto a high content sim
[11:51] Justin Clark-Casey: and gfx/cpu are tangled together, so it would not enormously surprise me if a high content sim also had an impact on ping times
[11:51] VivK Lowlag: how does that even pass the viewers usability list?
[11:51] Arielle Popstar: most regions it is fine on
[11:52] Arielle Popstar: just certyain one it isnt
[11:52] Richardus Raymaker: are there lots of moveing prims in that sim ?
[11:52] Arielle Popstar: wright plaza has never been good for me
[11:52] Justin Clark-Casey: wright plaza has a alot of content, objects and textures
[11:52] Arielle Popstar: yes
[11:52] Justin Clark-Casey: it is highly stressful
[11:53] Justin Clark-Casey: moving prims really make no difference - I think it's all down to number of prims, number of textures and viewer settings
[11:53] Arielle Popstar: i used to think it was cause of the amount of people here at meetings but i came before the rush and still had high ping
[11:54] Arielle Popstar: oh well no biggie. Just wondered if scripts could have a bearing on it
[11:54] Richardus Raymaker: what amd cpu. to ask somethinbg fagerous..
[11:54] VivK Lowlag: probably a turion
[11:55] Arielle Popstar: oh...duh...its a intel core duo
[11:55] Arielle Popstar: the other lappy is a amd
[11:55] Richardus Raymaker: :)) good not so happy with amd and 3d graphics
[11:55] Andrew Hellershanks: fagerous??
[11:55] Justin Clark-Casey: I don't ask nowadays :)
[11:55] VivK Lowlag: mobile graphics are laughable
[11:56] Key GruinKey Gruin takes notes for the dictionary of made-up word :P
[11:56] Andrew Hellershanks: :-)
[11:57] VivK LowlagVivK Lowlag pictures the rush to get it entered into Webster's
[11:57] Richardus Raymaker: eep franzi
[11:57] Arielle Popstar: cant disagree Viv as ihavean od petium 4 with a Geforce Ti4200 that operates faster though not as much detail as this one
[11:57] Franziska Bossi: sh...
[11:58] Franziska Bossi: eep!!
[11:58] Justin Clark-Casey: a p4, sweet jesus
[11:58] Andrew Hellershanks: Does anyone here do execution profiling with C#?
[11:58] Justin Clark-Casey: some of the intel guys might have
[11:58] Justin Clark-Casey: in general difficult to do with opensim since it's such an intensive program
[11:59] Andrew Hellershanks: k. Just wondering what tools can be used (preferably free ones). I would like to profile the warp3d code change to see how big a difference it makes.
[11:59] Justin Clark-Casey: unfortunately you'll find that difficult to do on free tools
[11:59] Justin Clark-Casey: possibly some of the commercial windows ones but I couldn't say
[11:59] Andrew Hellershanks: got rid of a lot of y*256+x calcs in loops and changed them to simple increment on variable.
[11:59] Justin Clark-Casey: but tbh, you might be more interested in memory profiling
[11:59] Justin Clark-Casey: andrew: in general that kind of stuff simply won't make much difference
[12:00] Andrew Hellershanks: Doesn't hurt.
[12:00] Justin Clark-Casey: the problems in opensim are more to do with certain patterns of memory usage and network delays, I think
[12:00] Andrew Hellershanks: Warp3D takes long time to generate maps. Every little improvement helps.
[12:00] Justin Clark-Casey: yes, I'm just sayihng that such improvements will be practically zero
[12:01] Justin Clark-Casey: but hey, prove me wrong :)
[12:01] Richardus Raymaker: hmm the speed seems not to slow. but i heared people talking about 30 minutes. but that it use 200MB more meory is a bigger hit
[12:01] Andrew Hellershanks: justin, I'd like to but can't unless I can find a way to do execution profiling. :-)
[12:01] Justin Clark-Casey: andrew: well, you could do simple time measurements embedde in the code itself
[12:01] Key Gruin: have a small bit of news before the meeting ends
[12:01] Justin Clark-Casey: print out how long it takes to generate with the old code versus the new code, stuff like that
[12:02] Andrew Hellershanks: yeah. THere are some timer stuff but its only down to tick level or so.
[12:02] Justin Clark-Casey: key: ok, well we're at the top of the hour so I'd say that's about now
[12:02] Justin Clark-Casey: andrew: if you don't see an improvmenet therer then it's negibile anyway
[12:02] BlueWall.Slade back
[12:03] Andrew Hellershanks: justin my regions don't have much in them so I'm not going to see a huge difference in my setup.
[12:03] Justin Clark-Casey: andrew: so how are you hoping to tell without real data?
[12:03] Richardus Raymaker: last note for justin and bllue. i have disabled scipts and more keeps stable seems to drop a very little amount
[12:03] Justin Clark-Casey: andrew: you still need that even if you were profiling
[12:03] Richardus Raymaker: i go try to test some commands seperate
[12:03] Richardus Raymaker: Andrew, get a free region from linda
[12:04] Key Gruin: ok dve coyle switched the asset server over to the new sras system yesterday
[12:04] Key Gruin: everything went well
[12:04] Richardus Raymaker: aha, sneaky key. nice
[12:04] Andrew Hellershanks: justin, 65536 variable increments has to be less time than the same number of multiply plus an add.
[12:04] Justin Clark-Casey: key: cool, but I thought osgrid was already using sras?
[12:04] Key Gruin: it's handled over a million assets already so far
[12:04] BlueWall.Slade nice - congrats
[12:04] Justin Clark-Casey: andrew: modern cpus do thos ops so fast it's probablyi not noticeable
[12:05] BlueWall.Slade still in ruby?
[12:05] Justin Clark-Casey: andrew: you might even find the compiler has already optimized it away
[12:05] Key Gruin: not sure Justin, maybe an upgrade?
[12:05] UUID Speaker: London Silversmith: 9a9031c0-e13b-496f-8dc2-41c454b67cb0
[12:05] Key Gruin: in dave's words, " will take some time to convert all the old assets to be used by the new version, but that will all
 happen asynchronously"
[12:05] Andrew Hellershanks: It makes the code look cleaner at least. :-) I never trust the compiler to optimize code in the way I think it could/should do it.
[12:05] Justin Clark-Casey: key: ah, so a new version
[12:06] Key Gruin: its faster than the old version, too, and uses less memory
[12:06] Richardus Raymaker: Simple Ruby Asset Server (SRAS) right ?
[12:06] Key Gruin: yeah
[12:06] Justin Clark-Casey: cool. Sounds like its needed if there are lots of delayed asset requests
[12:06] Justin Clark-Casey: or uploads perhaps
[12:07] Justin Clark-Casey: alright, I need to go and try and figure out why viewers dislike dynamic textures
[12:07] Arielle Popstar: would like to mention that after last weeks meeting a few of us discussed about opening a Metaversity in the hopes of creating a place where we can learn about opensim coding. Anyone have thoughts on that or willing to provide some help in that area would be greatly appreciated. Hoping that maybe it will interest people in leaning how to submit patches etc.
[12:07] Key Gruin: take care Justin
[12:07] BlueWall.Slade laterz jcc
[12:08] Richardus Raymaker: Arielle would also be a good place for townhgall meetings maby
[12:08] Richardus Raymaker: bye JCC
[12:08] Arielle Popstar: Hiro and Bluewall have created a region for it already
[12:08] Justin Clark-CaseyJustin Clark-Casey waves
Personal tools
About This Wiki