User Experience Interest Group/Transcripts/2010-04-01

Topic

User Experience Interest Group Discussion for April 01, 2010.

Topic: Reducing the viewer's footprint.

Summary

We talked about reducing the viewer's "footprint" -- how much memory, bandwidth, CPU power, etc. that it uses.

Transcript

[15:19]	Morgaine Dinova:	You can ignore it to some extent. Depends on the details
[15:19]	Geneko Nemeth:	And then nobody upheld the Tao anymore.
[15:19]	Morgaine Dinova:	Hiya Jacek!
[15:19]	Nimra Irata:	hi Kecaj
[15:19]	Jacek Antonelli:	Hi all. Sorry I'm late. I had to inspect the TOS before logging in
[15:19]	Aklo Modan:	i don't think anything we're doing here has anywhere near that much impact, roight?
[15:20]	Geneko Nemeth:	Not from anywhere I have seen.
[15:20]	Geneko Nemeth:	But, maybe, one day...
[15:20]	Nimra Irata:	:)
[15:20]	Aklo Modan:	ツ
[15:20]	Geneko Nemeth:	You know, I missed the days where Malbers would bring prototypes to the office hour.
[15:21]	Boroondas Gupte:	prototypes of policies?
[15:21]	Jacek Antonelli:	Okay, so. Shall we begin? Today's topic is reducing the viewer's "footprint" -- how much memory, bandwidth, CPU power, etc. that it uses
[15:21]	Geneko Nemeth:	I wish.
[15:21]	Morgaine Dinova:	I clicked-through the ToS, as usual. Anything reasonable I expect to adhere to. Anything unreasonable I reject, period.
[15:21]	Morgaine Dinova:	Cool, Jacek
[15:22]	Morgaine Dinova:	Great topic too
[15:22]	Geneko Nemeth:	I think that's kinda...
[15:22]	Charlette Proto:	woot woot the new TOS?
[15:22]	Geneko Nemeth:	If you make the viewer more effecient, people would just add stuff until it's slow as usual.
[15:22]	Morgaine Dinova:	Hiya Charlette :-)
[15:22]	Charlette Proto:	sorry... hi everyone
[15:23]	Boroondas Gupte:	ToS didn't load for me. Maybe that's because I flipped a bool in llfloatertos.cpp, dunno :-P
[15:23]	Jacek Antonelli:	Geneko: Maybe. But imagine we were creating a viewer meant to run on very old computers. What would be different?
[15:23]	Morgaine Dinova:	Jacek: are we working on the assumption that people WANT to reduce the footprint? (I see no sign that LL wants to, for example)
[15:23]	Geneko Nemeth:	Heavy fog@!
[15:23]	Charlette Proto:	it took ages to load for me too and wouldn't scroll for the first few seconds
[15:23]	Aklo Modan:	Hey, Jacek, isn't the network the biggest pb, really?
[15:24]	Geneko Nemeth:	It would be like Superman 64!
[15:24]	Geneko Nemeth:	Ugh-ugh. I used to have a laptop with a GeForce 8400 card, and the framerate was annoying.
[15:24]	lufpleh Obstreperous:	I asked Esbee & Amada wha the expected footprint for Viewer2 was at the VWBPE conferenace when they were trying to sell it to educators, they had no ide a what the footprint was
[15:24]	Nimra Irata:	yeah, still have a c64 ... er ... somewhere in a box
[15:24]	Jacek Antonelli:	Network use is a bit problem, yes. Especially for people with monthly bandwidth quotas
[15:25]	Charlette Proto:	I still think a communications-only wiewer option is in order - 1FPS or adjustible to use on low-power laptops etc
[15:25]	Jacek Antonelli:	Think also about "netbooks" -- lightweight, low-power laptops
[15:25]	Aklo Modan:	iPad?
[15:25]	Morgaine Dinova:	Now about taking each aspect of footprint in turn?
[15:25]	Jacek Antonelli:	What would you do to make the viewer run acceptably well on such hardware?
[15:25]	Charlette Proto:	if you throttled the rendering task the rest (comms) would pull ok
[15:25]	Geneko Nemeth:	Maybe it would be isometric...
[15:25]	Boroondas Gupte:	I hear Cool Viewer, lagacy renderer edition works quite good for low-end systems (old computer. netbooks not so much)
[15:26]	Geneko Nemeth:	Maybe it would look like Furcadia!
[15:26]	Charlette Proto:	yeah cool is very old and there in no way to throttle rendering there either
[15:26]	Charlette Proto:	think of user controlled rendering load
[15:26]	Geneko Nemeth:	(Furcadia is a 2d furry online world with user-customizable sprites and orthographic projection)
[15:26]	Morgaine Dinova:	I heard of a brilliant viewer than works great on both high and low powered systems. It was called something like Imprudence, odd name :-)
[15:27]	Geneko Nemeth:	Really? It seems to be the same client with all others, only freezes less....
[15:27]	Charlette Proto:	still morgaine - imagine being able to adjust how much CPU load the rendering task gets
[15:27]	Morgaine Dinova:	Yeah
[15:27]	Boroondas Gupte:	Haven't tested Imprudence recently. Does it run well on > 5 year old laptops?
[15:28]	Aklo Modan:	It runs ok on one of my 5 yr old laptops
[15:28]	lufpleh Obstreperous:	thought LL viewer did that for you, just rendered everything grey!! lol
[15:28]	Jacek Antonelli:	What are some of the other problems, besides bandwidth usage? High CPU usage? High memory usage? Needing to buy an expensive graphics card?
[15:28]	Geneko Nemeth:	That si also a possibility, only decoding enough textures to provide a rough color for the surfaces. Now, it will look like Nintendo 64 games....
[15:29]	Charlette Proto:	Second Life™ arcade view haha
[15:29]	lufpleh Obstreperous:	biggest problem is the future of computing is mobile, the desktop is going to disappear so either Sl has to get nimble or mobile power has to grow rapidly
[15:29]	Morgaine Dinova:	Before we leave bandwidth usage .... it should be possible to switch off what we don't want to be sent to us at the server's interest list. Currently even a text only client gets sent all the graphics. It's nuts.
[15:29]	Geneko Nemeth:	There might be also a compressing proxy like Opera Mini/Opera Turbo.
[15:30]	Geneko Nemeth:	Filter out bandwidth hogs like Aklo's silly title.
[15:30]	Jacek Antonelli:	Decoding textures only partially would help with bandwidth, yep. Maybe also use less graphics card memory, I'm not sure.
[15:30]	Morgaine Dinova:	Jacek: what happens if you send no ACKS for graphic data received?
[15:31]	Geneko Nemeth:	No, like prim profile params.
[15:31]	Boroondas Gupte:	it gets resent, I guess
[15:31]	Geneko Nemeth:	You could certainly not request textures.
[15:31]	Morgaine Dinova:	But it slows down, right?
[15:32]	Jacek Antonelli:	I think it may be possible to stop (or pause) requesting texture data after partially downloading it
[15:32]	Jacek Antonelli:	So you'd have blurry textures all the time, but use less bandwidth
[15:33]	Geneko Nemeth:	Just add the bias factor a bit.
[15:33]	Charlette Proto:	actually considering the low-load Second Life™ viewer 2 load on my machine is hidieous in comparison with the legacy viewer crop
[15:33]	Morgaine Dinova:	Can you simply not request any textures at all? Or does the sim populate the interest list automatically?
[15:33]	Jacek Antonelli:	I think surely the viewer can neglect to request any textures, or change the order
[15:33]	Wut Moorlord:	Jacek, true, although AV's tend to look like they just splatted lipstick all over their face while drunk until that last level decodes :)
[15:34]	Morgaine Dinova:	Jacek: depends if it's a pull operation, of it the interest list is just streamed to you
[15:34]	Jacek Antonelli:	After all, Snowglobe modified the order textures were requested, and that didn't require any server side changes (I think)
[15:34]	Morgaine Dinova:	or if*
[15:35]	Charlette Proto:	are textures so much of the problem (besides bandwidth), in my view the rendering FPS throttle control would be the way to go
[15:35]	Jacek Antonelli:	Textures are mostly a bandwidth issue
[15:35]	Morgaine Dinova:	Charlette: just talking about bandwidth reduction at this point
[15:35]	Jacek Antonelli:	Throttling the render loop would certainly help with CPU usage
[15:35]	Charlette Proto:	bandwidth is 'cheap' CPU load is another thing
[15:35]	Jacek Antonelli:	Bandwidth is not cheap everywhere
[15:35]	Boroondas Gupte:	Assuming you can control what you request, the more you cache, the less bandwidth you need.
[15:35]	Charlette Proto:	ah OK sorry, I missed the point
[15:36]	Geneko Nemeth:	Do you think using a top-down view instead of a followcam would mean distant objects will not need to be sent from server?
[15:36]	Charlette Proto:	Mini Opera serverside compression solution like Gen mentioned would be your best friend
[15:36]	Wut Moorlord:	Geneko, that would definately require less bandwidth unless you wanted a really really high camera
[15:36]	Charlette Proto:	WTF?
[15:36]	Geneko Nemeth:	Fix the camera then (?)
[15:37]	Wut Moorlord:	if you're happy rigging it you can save a bunch of stuff
[15:37]	Jacek Antonelli:	But would changing the camera angle still allow you to user SL in a meaningful way?
[15:37]	Jacek Antonelli:	*to use
[15:37]	Wut Moorlord:	although, how would it work if you were in room with a low ceiling?
[15:37]	Geneko Nemeth:	Don't draw the ceiling then.
[15:37]	Wut Moorlord:	how?
[15:38]	filipe Soup shouts: parei ^^
[15:38]	Wut Moorlord:	cull plane everything above the AV?
[15:38]	Wut Moorlord:	clip plane*
[15:38]	Geneko Nemeth:	Figure out it's in the camera's way to the avatar and show it semitransparent.
[15:38]	Wut Moorlord:	oooh, yeah. mmo stylie, good point
[15:39]	Charlette Proto:	what about sculptie map textures - the shapes would break-up if you don't load those 100%
[15:39]	Jacek Antonelli:	Geneko: Would you want to use SL from a top-down perspective? Seems like you would miss out on a lot.
[15:39]	Wut Moorlord:	yeah partially rezzed sculpties are a disaster
[15:39]	Boroondas Gupte:	actually, free camera movement is what sets SL apart from most "games" so I think that should be preserved even for low-bandwidth systems. Better reduce draw distance or something.
[15:39]	Nimra Irata:	what about updating the 3d view only once every 10 seconds?
[15:40]	Geneko Nemeth:	I don't know.
[15:40]	Wut Moorlord:	Nimra: then SL would no longer be a VW, it would be a slideshow.
[15:40]	Geneko Nemeth:	Maybe it's worth trying.
[15:40]	Nimra Irata:	yes, wut, but more accessible than with a text only viewer
[15:41]	Wut Moorlord:	i came in late, bandwidth constrained systems, or generally graceful degradation
[15:41]	Jacek Antonelli:	I don't think the top-down camera would be a good solution, especially if it's main benefit is just to reduce the number of objects that must be downloaded. Lower draw distance would do just as well
[15:41]	Charlette Proto:	Nimira is right, like I said a user controlled rendering load would be great
[15:41]	Boroondas Gupte:	I'd like to be able to modify how the LoD changes with distance.
[15:41]	Jacek Antonelli:	Wut: We're talking in general about ways to reduce the viewer's footprint -- bandwidth, memory use, CPU use, etc.
[15:41]	Wut Moorlord:	if i can trim it down to something like my phone, for e.g., that is bandwidth and crappy OpenGL ES limited
[15:41]	Morgaine Dinova:	Wut: we're taking apart all the components that might help in reducing the viewer footprint. One at a time hopefully
[15:42]	Wut Moorlord:	in situations like that i would love the ability to have a top down view, or maybe 0.1fps update because it's a pretty chat client away from home then.
[15:42]	Charlette Proto:	cache (HD/RAM) footprint besides the running app RAM is huge in Second Life™
[15:43]	Wut Moorlord:	yup :/
[15:43]	Nimra Irata:	:P 0.1 fps is once in 10 seconds
[15:43]	Morgaine Dinova:	There's one other thing that can reduce the bandwidth requirement, which I'll just mention in passing because we've discussed it before: (i) a working cache, and (ii) a vastly larger cache.
[15:43]	Geneko Nemeth:	That can't be helped much... Google Earth has a large cache too.
[15:43]	Wut Moorlord:	yeah, that's what you suggested :)
[15:44]	Charlette Proto:	eg my phone plays games OK (Arm 11) but not enough of anything to even imagine Second Life™ on that
[15:44]	Wut Moorlord:	it was intentional. crap on the desktop because i suspect you would find it felt aweful. but for constrained devices could vaguely render you a scene eventually...
[15:44]	Morgaine Dinova:	I would like world caches to be entirely separate from viewers/clients. Any world cache should be able to benefit.
[15:44]	Wut Moorlord:	i would like caches to work
[15:45]	Wut Moorlord:	they are fundamentally braindamaged
[15:45]	Morgaine Dinova:	Any world *client* should be able to bebefit
[15:45]	Wut Moorlord:	if we decide, for e.g. to all tp to the next sim
[15:45]	Wut Moorlord:	i will have to wait for your textures to rez in again
[15:45]	Morgaine Dinova:	Oh cripes, typing useless today. Getting a coffee
[15:45]	Wut Moorlord:	== retarded.
[15:45]	Charlette Proto:	but Google earth and other satellite viewers have persistent data (to a point) and Second Life™ cache is obsolete in a few hours
[15:45]	Geneko Nemeth:	Depends.
[15:46]	Wut Moorlord:	Charlette, yes, but a crapload of stuff in SL *is* static
[15:46]	Geneko Nemeth:	Depends if you're someone who stays mostly in a few places or if you are a gridanaut.
[15:46]	Nimra Irata:	what about the youtube approach to save bandwith through rendering as text?
[15:47]	Boroondas Gupte:	Even if you're a gridnout, the baked thextures for your most used outfits could be cached to avoid unneeded rebakes.
[15:47]	Charlette Proto:	your inventory is static Wut (about 200MB) but the rest seems to be downloaded each time you log
[15:47]	Wut Moorlord:	Nimra: :)
[15:47]	Geneko Nemeth:	You do realize it's April the 1st, right? It's a client side effect.
[15:47]	Morgaine Dinova:	That's one of the biggest misconceptions in SL, that "everything changes, it's all dynamic". Look around you. How much has changed since your last visit?
[15:47]	Geneko Nemeth:	... well, maybe the joke's on me. T_T
[15:47]	Wut Moorlord:	Charlette: yeah, there's a bunch of fail in the SL caching mechanisms
[15:48]	Morgaine Dinova:	"Everything CAN change" is not the same thing as "Everything DOES change".
[15:48]	Wut Moorlord:	and. specifically, it is eating up a lot of bandwdth.
[15:48]	lufpleh Obstreperous:	the date has changed since Iwas last here!
[15:48]	Morgaine Dinova:	lol
[15:48]	Jacek Antonelli:	The way content is delivered is also suboptimal in many cases, precisely because of the assumption that everything can change.
[15:48]	Nimra Irata:	i saw a video a prototype of the Imprudence viewer doing it http://blip.tv/file/3429688
[15:48]	Wut Moorlord:	yeah, we store dates as unsigned 200gb bigints these days
[15:49]	Charlette Proto:	I have a feeling the Second Life™ cache rarely 'finds' other avatars' assets even sim builds appear to load every time you log
[15:49]	Wut Moorlord:	not going to take THAT risk again =]
[15:49]	Wut Moorlord:	yeah, charlette it sure feels that way doesn't it?
[15:49]	Geneko Nemeth:	Don't think it caches objects.
[15:49]	Nimra Irata:	seems more conveniant to me than viewer 2
[15:50]	Jacek Antonelli:	If there were a way to designate certain objects as "static"/"permanent", they could be delivered once, in a compressed form, and re-used many times
[15:50]	Morgaine Dinova:	LOL @ that video :-)
[15:50]	Wut Moorlord:	i think the needing to reload textures when you move across sims is your skin is baked and stored temporarily on the sim?
[15:50]	Jacek Antonelli:	E.g. you TP to Loco Pocos, and the viewer downloads a compressed package of the main things in the sim, and textures
[15:51]	Boroondas Gupte:	better yet, a compressed diff since you last visited it
[15:51]	Wut Moorlord:	yeah
[15:51]	Wut Moorlord:	extrayeah
[15:51]	Boroondas Gupte:	make every sim a git repo :-P
[15:51]	Jacek Antonelli:	hehehe
[15:51]	Charlette Proto:	you could have a point Wut, the assets are copied to the sim and may even be limited in persistence this way - no idea
[15:51]	Wut Moorlord:	stopnow
[15:51]	Wut Moorlord:	xD
[15:51]	Morgaine Dinova:	Not as good as this video of mine --- this is the REAL future! http://www.blip.tv/file/3407015 :-)
[15:52]	Wut Moorlord:	Charlette, yeah, i'm hunching here we need to borrow someone who understands the cache better
[15:52]	Jacek Antonelli:	Let's switch gears and talk about rendering/graphics cards
[15:53]	Wut Moorlord:	I have nothing nice to say on this one. so i'll get coffee. :)
[15:53]	Morgaine Dinova:	Bor: I wonder if it would actually work to make each sim a Git repo, seriously :-)
[15:53]	Geneko Nemeth:	Maybe a different way of tessalacting prims would make them more optimized?
[15:54]	Geneko Nemeth:	*tessellacting
[15:54]	Boroondas Gupte:	ok. I'd like the settings that affekt FPS (draw distance, enabled shader types) to be set dynamically, depending on FPS, rather than manually
[15:54]	Charlette Proto:	OMG Morgaine, I kept thinking the kids would go in the hopper too
[15:54]	Jacek Antonelli:	There seems to be a widespread idea that post-Windlight viewers are more demanding of graphics cards, even when all the Windlight features are disabled
[15:54]	Boroondas Gupte:	so I can automagically keep FPS in a certain range
[15:54]	Morgaine Dinova:	Charlette: ahahaha
[15:55]	Jacek Antonelli:	Bor: That's an interesting idea. I wonder how it would affect the user experience, to have draw distance and graphics settings constantly adjusting themselves
[15:55]	Geneko Nemeth:	I think that would rock.
[15:55]	Charlette Proto:	anyone considered not rendering sculptie assets till the sculpt map loads and only if one wants them to load?
[15:56]	Charlette Proto:	sculpts can cause a huge load on the client side and sim bandwidth
[15:56]	Boroondas Gupte:	yeah. actual meshes would probably be cheaper than this hack.
[15:56]	Wut Moorlord:	Charlette: that would be so lul
[15:57]	Morgaine Dinova:	In my stupid GL play clients, I already fire off frame processing on a timer, so that my FPS is guaranteed instead of determined by scenegraph load. Of course that doesn't work if you process everything on every frame though.
[15:57]	Wut Moorlord:	okay. my $0.02 on the renderer. the bottleneck primarily seems to be the vertex pipeline, fillrate we have to burn, but can't because the vertex pipe is clogged
[15:57]	Charlette Proto:	mesh is for sure more pracitcal than poorely defined sculpt standard considering most sculpt maps are too big because hardly anyone understands what is good enough
[15:58]	Wut Moorlord:	culling is so broken that if you move the frustum in a crowded scene, it seems to invest more time trying to trim the scene down to size than it would if it hadn't bothered
[15:59]	Morgaine Dinova:	Wut: YES !!!!!!!!!
[15:59]	Charlette Proto:	your $0.02 could be right on the money Wut
[15:59]	Wut Moorlord:	and. yeah. meshes are better. prims solution to detail is THROW MORE VERTICES AT IT
[15:59]	Morgaine Dinova:	I've been examining Fast Timers, and noticed that more time is spent in sorting than in rendering for example
[16:00]	Wut Moorlord:	but i dunno if you noticed the geometry detail hasn't gone up in games for a WHILE, but the fillrate is so plentiful people throw in DoF blurring because they feel they OUGHT too because it's practically free
[16:00]	Charlette Proto:	the 32x32 pixel sculpt was a 'good' idea but people upload 128 and bigger sculpt textures all the time
[16:00]	Wut Moorlord:	it was still a facepalm
[16:01]	Jacek Antonelli:	Qarl is probably sitting in his laboratory screaming "Oh, God, I've created a monster!"
[16:01]	Wut Moorlord:	:D
[16:01]	Morgaine Dinova:	lol
[16:01]	Wut Moorlord:	pretty much. it was a cheap hack at the time, but meshes were needed from the outset
[16:01]	Charlette Proto:	just look at advanced > Rendering > Info Displays > Sculpt to see what I mean
[16:01]	Wut Moorlord:	although- problem, normal mapping is HARD. specular mapping is barable, mesh geometry means blender or such
[16:01]	Jacek Antonelli:	"I never intended for them to be used this way! It's not my fault! Whhyyyyyy???"
[16:02]	Wut Moorlord:	prims are easy.
[16:02]	Charlette Proto:	lump's head has a 256x256 sculpt
[16:02]	Charlette Proto:	my hair is full of 128x128 maps
[16:02]	Wut Moorlord:	noobies have to get results and not be sent away with a giant maya book and told to figure out how to make a cube.
[16:02]	Charlette Proto:	all waste of data since the renderer only uses 32x32
[16:02]	Boroondas Gupte:	prims are easy because they have an inworld editor. Make one for meshes and meshes will be easy, too.
[16:03]	Morgaine Dinova:	I don't see Advanced > Rendering > Info Displays > Sculpt in Imprudence
[16:03]	Wut Moorlord:	Charlette: lol, quite.
[16:03]	Charlette Proto:	old viewer
[16:03]	Wut Moorlord:	Baroondas: patch or GTFO. modelers are a bitch. i'm not writing one.
[16:03]	Wut Moorlord:	=]
[16:03]	Wut Moorlord:	i think Quarl would bring a gun into work if he was asked
[16:04]	Boroondas Gupte:	:-\
[16:04]	Wut Moorlord:	ideally, yeah, brilliant. but y'know
[16:04]	Geneko Nemeth has a partial patch with raycasting and shadow frustra, but seem to have forgotten sculpts.
[16:04]	Wut Moorlord:	ouch
[16:04]	Charlette Proto:	same problem with avs like Tech - hundreds of 64x64 sculpt textures on one av
[16:05]	Boroondas Gupte:	I doubt those are all distinct textures.
[16:05]	Wut Moorlord:	Geneko, wait. sculpts just don't occlude? or they occlude as spheres to the stencil buffer. that would look so um... :)
[16:05]	Geneko Nemeth:	Wut: I have no idea...
[16:05]	Wut Moorlord:	achully, to stencil shadow technically you need a closed orientable manifold
[16:05]	Wut Moorlord:	sculpts aren't
[16:05]	Geneko Nemeth:	I'm only talking about the menus that Jacek moved to XUI
[16:05]	Techwolf Lupindo:	They way I understand it, you have to upload 64x64 to get the 32x32 sculpte map to work properlly.
[16:06]	Nimra Irata:	is quarl == q ?
[16:06]	Morgaine Dinova:	No
[16:06]	Morgaine Dinova:	Qarl's a graphics dev, Q is a product manager
[16:06]	Nimra Irata:	ah
[16:07]	Wut Moorlord:	Techwolf, oh so they did the old double sided quad trick on 64x64 scults to get the shadows working?
[16:07]	Wut Moorlord:	sculpts*
[16:07]	Morgaine Dinova:	Qarl's the guy who did sculpties and now is working on meshies. Q's the guy who I'm fighting on client-side scripting issues.
[16:08]	Charlette Proto:	Tech you are right I think, since the 32x32 isn't interpreted right for what i remember (0 to 31 was never coded right or something like that)
[16:09]	Nimra Irata:	sometimes confusing to keep track of all the lindendes who appear out of nothing and also keep disappearing there
[16:09]	Wut Moorlord:	ooooh no, tey don't shadow sculpts the problem is that you need 64x64 sculpts to get 32x32 sculpt data. back on rtack
[16:09]	Wut Moorlord:	track*
[16:09]	Nimra Irata:	-typos
[16:10]	Wut Moorlord:	sorry. miles away. up for 25 hours or something now. it's interesting wut time ;)
[16:10]	Morgaine Dinova:	Wut: in the end, short of a Nobel Prize winner who makes everything "closed orientable manifolds", everything will become voxels :-)
[16:10]	Wut Moorlord:	naw, we cracked the infinite detail hack
[16:10]	Morgaine Dinova:	Or better, pure particles :-)
[16:11]	Wut Moorlord:	it's real, but it has horrible artefact issues. it's a design flaw
[16:11]	Charlette Proto:	no kidding - I heard that 32x32 map rows and collumns were never coded prperly to read as 0-31 or something like that
[16:12]	Charlette Proto:	isn't voxel a crappy old english GM car morgaine HAHA
[16:12]	Wut Moorlord:	and. why? you can't do multiple passes on particles most of the time. we already have geometry pipelines to tesselate nurbs/catmull-rom splines to whatever detail you fancy. i think we solved that already tbh
[16:12]	Morgaine Dinova:	Haha
[16:13]	Morgaine Dinova:	Wut: won't scale
[16:13]	Wut Moorlord:	does scale
[16:13]	Wut Moorlord:	voxels fail
[16:13]	Morgaine Dinova:	Crowds
[16:13]	Wut Moorlord:	occlusion buffers
[16:13]	Charlette Proto:	the weird thing is that LL have a need to reinvent the wheel when open GL has so much in it anyway
[16:13]	Boroondas Gupte:	are we just talking shadows?
[16:13]	Morgaine Dinova:	Voxels have trivial LOD, scale arbitraily
[16:14]	Boroondas Gupte:	or geometry in general?
[16:14]	Wut Moorlord:	um. you mean like subdivision serfaces?
[16:14]	Wut Moorlord:	surfaces*
[16:14]	Wut Moorlord:	or nurbs? or any order spline you might like?
[16:14]	Wut Moorlord:	beats making pictures out of rice and glue every frame in my book.
[16:15]	Boroondas Gupte:	if you use voxels for everything, you'll get a problem with lightning, as the normals will be all wrong.
[16:16]	Morgaine Dinova:	The "every frame" thing is why SL rendering is so laggy.There doesn't seem to be a meme for only processing things that change
[16:16]	Wut Moorlord:	boroondas not necessarily, you can store normal data along with it
[16:16]	Wut Moorlord:	i mean, there's gigavoxels, they look great
[16:17]	Morgaine Dinova:	Getting back a little closer to home, how else can we reduce rendering costs?
[16:17]	Wut Moorlord:	http://www.icare3d.org/blog_techno/gpu/gigabroccoli_the_mandelbulb_into_gigavoxels.html
[16:17]	Charlette Proto:	agree with Morgaine, the frame based loop for everything has to go one day
[16:17]	Geneko Nemeth just stopped paying attention to the chat for a while
[16:17]	Charlette Proto:	radical Gen
[16:18]	Charlette Proto:	you such a rebel
[16:18]	Morgaine Dinova:	Charlette: if for no other reason that multicore. We HAVE to stop the single rendering loop madness.
[16:18]	Morgaine Dinova:	than* multicore
[16:19]	Charlette Proto:	precisely Morg, especially poor for natwork and rendering to be locked into the frame loop
[16:19]	Boroondas Gupte:	I'd even like distinct loops for UI rendering and content rendering, so neither can lock up the other.
[16:19]	Charlette Proto:	I'm sure a single core CPU would do threaded applications better than trying to thread a single loop
[16:19]	Morgaine Dinova:	Yeah Bor
[16:20]	Morgaine Dinova:	Wut: is there anything going on in the GL world on multicore realtime scene rendering?
[16:21]	Morgaine Dinova:	Wut's forehead may currently be embedded in the keyboard
[16:21]	Charlette Proto:	GPUs are multicore by nature so GL tasks would take advantage of that I think
[16:21]	Morgaine Dinova:	Charlette: no no, I don't mean GP-side
[16:22]	Morgaine Dinova:	Erk, GPU
[16:22]	Nathan24 Romano:	black and blue and broken bones
[16:22]	Boroondas Gupte:	?
[16:22]	Wut Moorlord:	sorry nipped out for a smoke
[16:22]	Nathan24 Romano:	hey pig
[16:22]	Wut Moorlord reads
[16:23]	Charlette Proto:	good work Nathan
[16:23]	Charlette Proto:	we really needed this
[16:23]	Boroondas Gupte:	finally some rain to cool down!
[16:23]	Charlette Proto:	maybe you could attach a self-rezzing garden to your ass as well nathan
[16:24]	Wut Moorlord:	yeah, i think there's a lot of scope for some clever parallel rendering loops in the future, i will ask some eggheads if there's anything out there that's workable
[16:24]	Nathan24 Romano:	what is this
[16:24]	Wut Moorlord:	for now i keep it as a stupid loop because i do as the games programmers do
[16:24]	Wut Moorlord:	because they do it that way and it works :)
[16:24]	Charlette Proto:	a meeting Nathan
[16:25]	Nathan24 Romano:	between?
[16:25]	Charlette Proto:	could you pleasse pay attention Nathan or leave
[16:25]	Nathan24 Romano:	my old pal geinko bolaba is here
[16:25]	Wut Moorlord:	OpenCL and similar gpu insanity will likely lead to some extremely cool stuff. but i'm just not familiar enough with that area to comment, sorry.
[16:26]	Morgaine Dinova:	Next area for footprint reduction, maybe?
[16:26]	Nathan24 Romano:	haha wow this is an advanced meeting of intellectuals huh?
[16:26]	Jacek Antonelli:	I think we'll wrap it up here. Any final comments?
[16:26]	Nathan24 Romano:	yall look like the x-men though
[16:27]	Wut Moorlord:	that's sl for you, nathan.
[16:27]	Charlette Proto:	OpenCL is Nvidia-only I think and ATI have a thing like it as well
[16:27]	Nathan24 Romano:	cept beast is black....was black
[16:27]	Wut Moorlord:	CUDA is nvidia only, OpenCL is kronos, works on both
[16:27]	lufpleh Obstreperous:	came across this impressive claims about graphics and rendering but way above my head http://www.vizworld.com/2010/03/unlimited-detail-technology/
[16:27]	Wut Moorlord:	amd even have an OpenCL emulator
[16:27]	Morgaine Dinova:	Jacek, are we planning to get together in OSgrid again, or in any of the other open grids?
[16:27]	Charlette Proto:	ah OK, i've been confused about these things Wut
[16:28]	Nathan24 Romano:	yall like scripters ?
[16:28]	Wut Moorlord:	lufpleh, rigged demo noone will buy because they couldn't be bothered to make any asset import tools, and their culling system isn't correct enough
[16:28]	Wut Moorlord:	cool bit of engineering though
[16:29]	Nathan24 Romano:	yo smarties whys my type sound still playing
[16:29]	Jacek Antonelli:	Morgaine: I'm not sure, at the moment.
[16:29]	Wut Moorlord:	Charlette: yeah, it's acronym soup :)
[16:29]	Morgaine Dinova:	Jacek: whether UXIG or not, I'd like to get together in other worlds. Whatever the topic
[16:30]	Nathan24 Romano:	just gonna ignore me huh....why the hell you even say leave or stay
[16:30]	Wut Moorlord:	we already did, if you remember.
[16:30]	Charlette Proto:	I said pay attention Nathan - that is the way people find out what meetings are about
[16:31]	Nimra Irata:	Morgaine: I think with all that confusing TVPs, TOSeses, WTFs, FUDs a more convenient meeting point would be great
[16:31]	Nathan24 Romano:	oh snap i figured it out
[16:31]	Wut Moorlord:	i'm sorry, the universe doesn't revolve around you.
[16:31]	Nathan24 Romano:	whos smarter than a 5th garder?
[16:31]	Nathan24 Romano:	haha grader
[16:31]	Jacek Antonelli:	Mkay, I'm heading out. Take care, everyone
[16:31]	Wut Moorlord:	yeah, tc jacek.
[16:31]	Morgaine Dinova:	Take care Jacke
[16:32]	Morgaine Dinova:	Take care Jacek