| 00:12.17 | *** join/#brlcad DarkMaster (n=Apathy@pcp0011677997pcs.waldrf01.md.comcast.net) | |
| 00:22.33 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/src/conv/Makefile.am: renamed g-wave to g-obj to reflect the conventional dominance of .obj's referring to the wavefront format and our converters using conventioned extensions. |
| 00:23.20 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/src/conv/ (g-obj.1 g-obj.c g-wave.1 g-wave.c): renamed g-wave to g-obj to reflect the conventional dominance of .obj's referring to the wavefront format and our converters using conventioned extensions. |
| 00:24.38 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/NEWS: renamed g-wave converter to g-obj |
| 00:34.04 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/src/ (10 files in 3 dirs): use bu_setlinebuf instead of the old block that checked system 'types' and was replicated where needed |
| 01:01.29 | Twingy | +e |
| 01:10.38 | Twingy | grumph |
| 01:18.38 | Twingy | aha |
| 01:19.03 | Twingy | I'm gonna have to write a manual for this thing soon |
| 01:19.12 | Twingy | developers manual anyway |
| 01:19.45 | DarkMaster | "And if this screen comes up, always type 'No'" |
| 01:20.27 | Twingy | "If you get the message "Core Dumped", "Bus Error" or "Segmentation Fault" then you've made a mistake." |
| 01:20.36 | Twingy | :) |
| 01:21.55 | DarkMaster | "Those are bad, and it's your fault." |
| 01:23.22 | Twingy | hrm, halfway done kd cache |
| 01:25.10 | DarkMaster | the developers manual has two pages :-) |
| 01:25.25 | Twingy | the second one reads, "This page intentionally left blank." |
| 01:29.33 | Twingy | grumph |
| 01:30.52 | DarkMaster | title page and blank page |
| 01:31.04 | DarkMaster | sounds like the manual for battlefield 2 |
| 02:02.40 | CIA-5 | BRL-CAD: 03brlcad * 10brlcad/regress/Makefile.am: ignore exp_air.pl and adj_air.pl and .density too |
| 02:04.06 | Twingy | hooray for contiguous memory |
| 02:22.42 | DarkMaster | hooray for Internet |
| 02:25.08 | Twingy | I hear they have the internet on computers now |
| 02:25.28 | DarkMaster | its called... "Internet" |
| 02:25.48 | DarkMaster | it does things like let you connect to bulletin boards and compuserve |
| 02:37.21 | Twingy | mmm, getting close |
| 02:40.12 | Twingy | sean, I think I'll just put kd-cache builder in adrt_bench for now |
| 02:40.22 | Twingy | and leave it out of g-adrt for time being |
| 02:40.32 | Twingy | this makes things a bit less icky |
| 02:41.43 | Twingy | actually, I'll probly leave it there indefinetly |
| 02:42.08 | learner | k |
| 02:42.08 | Twingy | cause you may want to use adrt_bench to tune the tree |
| 02:42.15 | Twingy | to get optimal tree building |
| 02:42.29 | Twingy | ultimately it should be completely turn-key "big red button" |
| 02:42.42 | Twingy | but right now it's 50/50 in terms of getting optimal and super-optimal |
| 02:43.33 | learner | 49/50/1 optimal, super optimal, badass |
| 02:43.37 | Twingy | heh |
| 02:43.54 | Twingy | in my opinion a kd-tree builder is not badass |
| 02:44.06 | Twingy | until you can completely remove the "depth" limit termination criteria |
| 02:44.13 | learner | okay, then just ass |
| 02:44.17 | Twingy | yep :) |
| 02:44.25 | Twingy | I've get to see somone present at siggraph |
| 02:44.35 | Twingy | that doesn't still rely on that depth criteria |
| 02:44.39 | Twingy | err yet |
| 02:44.50 | Twingy | and I'm damn close to being able to remove it |
| 02:44.56 | learner | you shall be the first |
| 02:45.04 | learner | then you still wouldn't get to see it |
| 02:45.12 | learner | unless you watch yourself on reply on the dvd |
| 02:45.19 | Twingy | heh |
| 02:51.11 | Twingy | okay, now to parse the cache |
| 02:52.52 | Maloeran | What is this cache briefly? |
| 02:53.04 | Twingy | a binary dump of my kd-tree |
| 02:53.18 | Maloeran | Ah, right |
| 02:53.26 | Twingy | I have a highly aggressive algorithm that will probly take 2 - 3 hours to generate a kd-tree for the 15 mil polygon stryker |
| 02:53.46 | Twingy | if the user does not supply it then it uses the mid-split algorithm and takes about 3 minutes |
| 02:54.18 | Twingy | and that'll be packed into my benchmark utility |
| 02:54.26 | Maloeran | Hum. I think this aggressive algorithm could be sped up |
| 02:54.29 | Twingy | so if the user wants to play with settings they can to learn how it works |
| 02:54.40 | Twingy | my algorithm hasn't been optimized yet |
| 02:54.53 | Twingy | I have one more piece of logic to add |
| 02:54.58 | Twingy | then I plan to write a paper on it |
| 02:55.21 | Maloeran | Right, okay. I mentionned before, the first pass of my scene analysis builds a tree that should be fairly close to the optimal construction for BSP rendering |
| 02:55.26 | Twingy | I can make the algorithm run for 3 days if I wanted to |
| 02:55.28 | Maloeran | It takes some 5-6 seconds on the usual tank scene |
| 02:55.34 | Twingy | but the improvement would by about 2% more |
| 02:55.49 | Twingy | I take about 3 seconds on that tank with my mid-split |
| 02:55.52 | Maloeran | There are often ways to converge towards the solution much faster |
| 02:55.58 | Twingy | but the aggressive one will spend about 30 - 60 seconds |
| 02:56.36 | Twingy | so the idea will be the analysis converts their geometry during the day |
| 02:56.53 | Maloeran | If I'm going to open-source, would you be willing to make the jump to sector based ray-tracing? |
| 02:56.54 | Twingy | and at night they can crank out a tight kd-tree |
| 02:57.11 | Twingy | hum |
| 02:57.36 | Twingy | I think at first I'd be inclined to make it an alternative like we'd discussed about a year ago |
| 02:57.41 | Twingy | like have 2 engines in the back-end |
| 02:57.46 | Twingy | user just selects which one to use |
| 02:58.11 | Twingy | did you see what I wrote about contract support? |
| 02:58.16 | Maloeran | I didn't |
| 02:58.44 | Twingy | I said you should talk to those swedes and tell them instead of buying IP, having them buy a support contract and you will integrate your engine into their software |
| 02:58.53 | Twingy | support contracts are typically worth much more money and buying a product |
| 02:59.10 | Maloeran | Ah yes, I read that. I have explored the idea, the "problem" is that I understood they don't want the software for themselves |
| 02:59.24 | Maloeran | They want to buy the IP and sell the licenses and support |
| 02:59.24 | Twingy | find out who they want it for |
| 02:59.31 | Maloeran | I have tried ;) |
| 02:59.32 | Twingy | and sell it to them |
| 02:59.47 | Twingy | look at their website and find out who their customer base is |
| 03:00.07 | Twingy | if they are in sweden |
| 03:00.10 | Twingy | sell them the IP |
| 03:00.16 | Twingy | their laws don't apply to canada |
| 03:00.33 | Twingy | they can't sue you |
| 03:00.37 | Maloeran | Eh, amusing thought |
| 03:00.44 | Twingy | so do it |
| 03:00.48 | Twingy | you're legally safe |
| 03:01.17 | Twingy | you'll get a $16k nasty gram |
| 03:01.18 | Twingy | big woop |
| 03:01.31 | Maloeran | I suppose what I wanted to ask a few seconds ago was : if I'm going to open-source, will you keep working on the BSP engine or switch over to the sector engine? |
| 03:01.47 | Twingy | well, you're the brains behind the sector engine |
| 03:01.55 | Twingy | and I'm the brains behind the kd-tree stuff I been doing |
| 03:02.06 | Twingy | so it's not like me jumping in the middle of your work will benefit you much |
| 03:02.13 | Twingy | atleast in the short term |
| 03:02.19 | Twingy | but we can bounce ideas off each other |
| 03:02.26 | Twingy | and look over each others code |
| 03:02.31 | Twingy | why not have the fastest of both types of engine |
| 03:02.39 | Twingy | just to say we can |
| 03:03.31 | Maloeran | I see, okay. I believe sector based ray-tracing has much more potential, BSP has been explored in depth for decades |
| 03:03.31 | Twingy | it'll be like choosing between phong, blinn, torrence-sparrow lighting models |
| 03:03.41 | Twingy | they all more or less do the same thing, but each have their perks |
| 03:04.03 | Twingy | yep, but I still think it's nice to have both in there even for educational reasons for people getting into ray-tracing |
| 03:04.08 | Maloeran | To speak frankly, the whole appeal of open-sourcing to me is being able to keep working on this with other people |
| 03:04.15 | Twingy | to see what a highly optimized kd-tree engine looks like and what its short comings are |
| 03:04.26 | Twingy | yep yep |
| 03:04.30 | Twingy | and we would |
| 03:04.56 | Twingy | I'm sure at some point yours would probably be the default engine used |
| 03:05.06 | Twingy | and mine would be there for legacy reasons |
| 03:05.31 | Twingy | with possibly an MLRTA choice as well |
| 03:05.35 | Twingy | 3 engines to choose from |
| 03:06.11 | Maloeran | *nods* But should I understand you'll keep working on yours for a while? |
| 03:06.14 | Twingy | it's nice to have several engines that can plug and play so you can bench mark your work versus other approaches for when you goto write a paper |
| 03:06.32 | Twingy | until yours works seemlessly with adrt, yah |
| 03:06.39 | Twingy | because I have a customer base to support at work |
| 03:06.49 | Twingy | the analysts... |
| 03:07.00 | Maloeran | Of course. That shouldn't be too much trouble |
| 03:07.01 | Twingy | once yours took the place of mine then I could drop development |
| 03:07.07 | Twingy | in theory, yah |
| 03:07.13 | Twingy | probly 2 - 3 weeks |
| 03:07.26 | Twingy | so this kd-tree code might be the last stuff I do to this engine |
| 03:07.46 | Maloeran | Good :), okay |
| 03:07.52 | Twingy | I've established to management that high-performance ray-tracing is what we should be leveraging for the near future, i.e next 5 years |
| 03:08.13 | Twingy | so anything that gets the job done faster the better |
| 03:08.50 | Maloeran | For you to be able to work on code that doesn't belong to the ARL, I suppose it has to released under a GPL or related license? |
| 03:08.53 | Twingy | I suspect once your engine plugged in and just "worked" I could be working on the sector stuff while at work |
| 03:08.59 | Twingy | LGPL |
| 03:09.05 | Twingy | GPL is no good |
| 03:09.13 | Twingy | according to sean |
| 03:09.28 | Maloeran | I see, okay |
| 03:09.28 | Twingy | the applications are GPL |
| 03:09.32 | Twingy | but the libraries are LGPL |
| 03:09.56 | Twingy | as is all of my stuff |
| 03:10.22 | Twingy | i.e anyone can download BRL-CAD, snarf libtie out of ADRT and give me no credit or tell me they're using it |
| 03:10.52 | Twingy | but not the case for RISE and ISST, the apps that use it |
| 03:11.26 | Twingy | openRT doesn't support clustered computing yet |
| 03:11.33 | Twingy | I seriously think if we were to act RIGHT now |
| 03:11.38 | Twingy | and make an openRT wrapper for your engine |
| 03:11.43 | Twingy | and add my distributed cluster code |
| 03:11.46 | Maloeran | Yes I get the idea, I can foresee source code and the techniques used being "stolen" abundantely |
| 03:11.51 | Twingy | we're knock them out of the water |
| 03:12.01 | Maloeran | Hehe, right |
| 03:12.04 | Twingy | seriously |
| 03:12.09 | Twingy | the code is done |
| 03:12.16 | Twingy | all that needs to be done is an openRT state machine |
| 03:12.29 | Twingy | that could be done in under a month |
| 03:12.35 | Maloeran | You mean being API compatible to their engine? |
| 03:12.47 | Twingy | an API compatible to openRT |
| 03:12.56 | Twingy | so people with openRT apps (yes they are a few out there already) |
| 03:13.00 | Twingy | could just include our stuff |
| 03:13.03 | Maloeran | I already have a complete API, designed to cover the needs of any ray-tracing library, based on the model of OpenGL |
| 03:13.08 | Twingy | and have BETTEr performance AND distributed computing |
| 03:13.13 | Twingy | I know you do |
| 03:13.18 | Twingy | and it's probly better than openRT |
| 03:13.25 | Twingy | but to get the openRT use community we need to support them |
| 03:13.36 | Maloeran | Okay, merely a wrapper then |
| 03:13.46 | Twingy | more or less |
| 03:13.53 | Twingy | it's a little more than that, but yea |
| 03:14.22 | Twingy | then we port Quake I to this engine |
| 03:14.27 | Twingy | and there you go, instant success |
| 03:14.58 | Twingy | I think that's the direction to take |
| 03:15.11 | Twingy | I could probly get sean to try getting a BZ-FLAG version as well |
| 03:15.26 | Twingy | sean is a bzflag developer |
| 03:15.30 | Maloeran | There's a lot of work ahead to handle dynamic geometry |
| 03:15.34 | Twingy | yah |
| 03:15.39 | Twingy | well you know how wald's doing it |
| 03:15.43 | Twingy | 2 tree's |
| 03:15.47 | Twingy | 1 for static, 1 for dynamic |
| 03:16.00 | Twingy | it's cheap, but it works |
| 03:16.06 | Twingy | if you keep that uglyness under the hood |
| 03:16.13 | Maloeran | Urgh :), yes I read |
| 03:16.16 | Twingy | then you could clean it up into 1 smarter structure later on |
| 03:16.31 | Twingy | time is of the essence |
| 03:16.54 | Maloeran | What I had in mind was a graph of sector graph ( a sector graph per object ), plus some object graphs are designed to be quickly modified if only parts of the geometry change |
| 03:17.13 | Twingy | heirchical graphs, sure |
| 03:17.16 | Maloeran | Anyhow, we can discuss that later |
| 03:17.21 | Twingy | yep |
| 03:17.29 | Twingy | well, you think this over nice and hard |
| 03:17.35 | Twingy | when you're ready I'm in |
| 03:17.45 | Maloeran | :) Understood |
| 03:18.03 | Twingy | my obligation to nurbs help with blender is coming to an end |
| 03:50.50 | Twingy | ok, time to finish this junk up |
| 04:31.07 | Twingy | Gee dubya |
| 04:31.45 | DarkMaster | good movie |
| 04:34.28 | Twingy | aight, I'm finishing this junk tommorrow |
| 04:35.57 | DarkMaster | monday? |
| 04:36.04 | Twingy | sunday |
| 04:36.10 | DarkMaster | tomorrow is monday! |
| 04:36.16 | Twingy | for me it |
| 04:36.19 | Twingy | 's sunday! |
| 04:36.37 | DarkMaster | so thats why you never show up on mondays |
| 08:20.24 | *** join/#brlcad brlcad (n=brlcad@bz.bzflag.bz) | |
| 08:21.23 | *** mode/#brlcad [+o brlcad] by ChanServ | |
| 12:35.01 | *** join/#brlcad docelic (n=docelic@195.246.23.200) | |
| 12:47.11 | Maloeran | Ahah... Really Justin, Reshetov didn't invent anything. It's something you and I had both thought of, the shortcuts of shooting rays in carefully crafted breaking up "beams" or "bundles", only viable for primary rays |
| 12:49.57 | Maloeran | Ah well, I'm just relieved I suppose ;) |
| 15:57.28 | Twingy | he's doing more than that |
| 15:58.01 | Twingy | he's entering the kd-tree mid-level using 0 volume kd-tree nodes... this is the frustum culling aspect of it |
| 18:30.02 | *** join/#brlcad silvap (n=prasad@pool-141-157-107-76.balt.east.verizon.net) | |
| 18:33.01 | silvap | ~hi |
| 18:33.06 | ibot | hello, silvap |
| 18:33.06 | learner | heir silva |
| 18:33.32 | silvap | yo sean |
| 18:33.48 | silvap | did u create the sf project yet? |
| 18:33.51 | learner | hehe |
| 18:34.13 | learner | not yet |
| 18:34.19 | learner | but not a bad idea :) |
| 18:34.26 | learner | not sure about that name though :) |
| 18:34.31 | silvap | aight |
| 18:34.33 | silvap | whaa? |
| 18:34.38 | silvap | whats wrong with it |
| 18:35.41 | learner | something punny |
| 18:36.00 | learner | the posterizer! |
| 18:36.30 | learner | source2bigassimage |
| 18:40.22 | silvap | u and ur puns.. |
| 18:51.08 | Twingy | welp, finished my workout and dropped 7 mice off at the pet store |
| 18:51.22 | learner | donation? |
| 18:51.36 | silvap | to hiss the boa |
| 18:51.43 | Twingy | in exchange for 2lbs of mice food |
| 18:51.56 | learner | ooh, good deal |
| 18:52.05 | Twingy | about the same amount the momma consumed to have the kiddlings |
| 18:52.29 | Twingy | presto chango, 2 lbs of mice food becomes little micelings |
| 18:52.45 | Twingy | now if only I could change 100lbs of mice food into a hot chick... |
| 18:54.06 | Twingy | hrm supposedly some people are showing up to look at my house any minute |
| 18:55.21 | Twingy | Yo momma so lazy she thinks a two-income family is where yo daddy has two jobs. |
| 18:56.27 | silvap | dc improv material right there |
| 18:59.24 | Twingy | amusing how well the mouse tank fits in the top drawer of my filing cabinet |
| 20:08.26 | *** join/#brlcad PKMOBILE (n=Apathy@pcp0011677997pcs.waldrf01.md.comcast.net) | |
| 21:55.09 | *** join/#brlcad CIA-12 (n=CIA@flapjack.navi.cx) | |