| *** Pseudonym has joined #Aqsis | 00:18 | |
| *** joeedh_class has joined #aqsis | 00:26 | |
| *** joeedh_class is now known as joeedh | 00:26 | |
| joeedh | hi. how is deep shadow maps in aqsis going to deal with the fact that pixar patented the method? | 00:27 |
|---|---|---|
| joeedh | I ask this because I have much the same problem with blender, where like aqsis I got a GSoC grant to implement them. my "circumvention" is basically using a raw abuffer instead of visibility functions, on the assumption that since blender doesn't have volumetrics it can get away with such a brute-force method. | 00:28 |
| joeedh | naturally this increases the memory usage, thus why part of my proposal is also implementing a tiling disk cache system. | 00:29 |
| Pseudonym | What exactly did Pixar patent? | 00:30 |
| joeedh | http://graphics.stanford.edu/papers/deepshadows/ | 00:31 |
| Pseudonym | That's not what I asked. | 00:31 |
| Pseudonym | :-) | 00:31 |
| joeedh | deep shadow maps | 00:31 |
| Pseudonym | Exactly what is patented? | 00:31 |
| Pseudonym | There's prior art in deep framebuffers from the 80s. | 00:32 |
| Pseudonym | They can't have patented the whole thing. | 00:32 |
| joeedh | iirc I think its the technique of using piecewise linear visibility functions rather then normal abuffers. | 00:32 |
| * joeedh looks it up | 00:33 | |
| * Pseudonym looks it up too | 00:33 | |
| Pseudonym | Crap. | 00:36 |
| Pseudonym | Here's a patent on something I worked out in 1998. | 00:36 |
| Pseudonym | Filed in 2000. | 00:36 |
| joeedh | what was it? | 00:36 |
| Pseudonym | Crack stitching. | 00:36 |
| Pseudonym | Actually, I think I was at Dot C when I worked it out. | 00:37 |
| Pseudonym | No, it was just before. | 00:37 |
| joeedh | well at least you can claim prior art :) | 00:37 |
| Pseudonym | Except it's not published! | 00:37 |
| Pseudonym | Ha. | 00:38 |
| Pseudonym | The deep shadow patent references the Entropy (now Gelato) sampling algorithm as prior art. | 00:38 |
| joeedh | lol | 00:39 |
| Pseudonym | The Entropy algorithm is pretty clever, actually. | 00:39 |
| * joeedh is going to have fun trying to figure out how to correctly create mipmaps when storing abuffer samples instead of visibility functions :/ | 00:39 | |
| joeedh | how does it work? | 00:39 |
| Pseudonym | Let's ignore motion blur and DoF for a moment | 00:40 |
| Pseudonym | Just consider a bunch of micropolygons which fall over an area. | 00:40 |
| Pseudonym | You want to integrate over the area. | 00:40 |
| Pseudonym | The "right" answer is to do a fully polygon clip. | 00:40 |
| Pseudonym | That way, you know exactly what bits of what polygon fall in the area. | 00:40 |
| Pseudonym | Obviously that's too expensive. | 00:41 |
| Pseudonym | The 1986 Pixar solution is to represent the area as points. | 00:41 |
| Pseudonym | And just sample the points. | 00:41 |
| Pseudonym | Entropy droped a whole bunch of lines over the area. | 00:41 |
| Pseudonym | And represented the "polygons" as piecewise linear functions that fall on the area. | 00:42 |
| Pseudonym | The reason being that piecewise linear functions are easier to clip than piecewise bilinear polygons. | 00:42 |
| joeedh | oh interesting. | 00:43 |
| joeedh | the lines were just the polygon edges? | 00:43 |
| Pseudonym | No, the lines are pre-chosen. | 00:43 |
| Pseudonym | Much like the points in stochastic z buffering. | 00:44 |
| joeedh | oh ok. | 00:45 |
| * joeedh draws a diagram | 00:45 | |
| Pseudonym | Draw a line. | 00:45 |
| Pseudonym | Then draw a polygon which intersects it. | 00:45 |
| Pseudonym | The bit that falls on the line is itself a line. | 00:45 |
| Pseudonym | Possibly more than one if the polygon is concave. | 00:45 |
| joeedh | ah I get it now. I vaguely remember reading about this a while back. | 00:46 |
| Pseudonym | Right.; | 00:49 |
| Pseudonym | OK, now I have no idea how to read patents. | 00:50 |
| Pseudonym | But I think the best answer is "ignore it". | 00:50 |
| Pseudonym | We don't know exactly what's patented. | 00:51 |
| Pseudonym | But almost all of it is a simple extension of previously published materials. | 00:51 |
| Pseudonym | And the paper doesn't declare a patent declaration. | 00:51 |
| Pseudonym | Who's doing DSM? Zac? | 00:53 |
| Pseudonym | Where's he located? | 00:53 |
| Pseudonym | Basically, DSM == shadow map + deep framebuffer. | 00:54 |
| Pseudonym | Neither of which has a current patent on it. | 00:54 |
| Pseudonym | If it's the optimisation that's the deal, then we'll come up with a different optimisation method. | 00:54 |
| Pseudonym | I thought of a few. | 00:54 |
| joeedh | well, what's different is rather then storing multiple samples were pixel, they store a piecewise linear function. | 00:55 |
| joeedh | which (it seems to me) it the core of their patent. | 00:55 |
| Pseudonym | So let's make it piecewise quardratic. | 00:55 |
| joeedh | eh. . .would that really help? | 00:55 |
| Pseudonym | And let's do a global minimisation instead of a local one. | 00:55 |
| Pseudonym | You also have to remember that Aqsis is mostly outside the US. | 00:56 |
| Pseudonym | In places where code isn't patentable. | 00:56 |
| joeedh | they use the term visibility function, and piecewise linear is almost treated like a suggested implementation detail. | 00:56 |
| joeedh | Pseudonym: so is blender. . .yet sadly I live *in* the U.S. | 00:56 |
| Pseudonym | Which project are you working on? | 00:57 |
| joeedh | deep shadow maps in blender. | 00:57 |
| joeedh | :) | 00:57 |
| Pseudonym | Ah. | 00:57 |
| Pseudonym | Clearly Blender should use Aqsis. Duh! | 00:58 |
| Pseudonym | :-) | 00:58 |
| *** cgTobi sets mode: +vvvv blend_dem joeedh Pseudonym timo2 | 01:03 | |
| joeedh | heh yeah. :) lol | 01:04 |
| Pseudonym | The fact is, Pixar doesn't use its patents offensively. | 01:05 |
| Pseudonym | And neither does Disney. | 01:05 |
| Pseudonym | So even on the off chance that we infringe... | 01:06 |
| joeedh | well, I did hear rumours about bmrt | 01:06 |
| Pseudonym | BMRT voluntarily dropped a feature. | 01:06 |
| joeedh | eh that isn't how I hear it happened. | 01:06 |
| Pseudonym | You mean the lawsuit? | 01:07 |
| Pseudonym | I think it's more about trade secret in that case. | 01:07 |
| joeedh | well I mean the many people who tell me pixar killed it. | 01:07 |
| joeedh | though now I remember the details, you might be right. | 01:07 |
| Pseudonym | Yes, I agree | 01:07 |
| Pseudonym | Pixar dropped the patent claims and instead alleged misappropriation of trade secrets. | 01:08 |
| Pseudonym | I suspect that Pixar also waited until Entropy was bought by nVIDIA (the deal was done, just not announced) to sue. | 01:09 |
| Pseudonym | Deeper pockets. | 01:09 |
| joeedh | sounded kindof sleezy to me :/ | 01:11 |
| Pseudonym | Yeah, a lot of people were severely pissed off. | 01:14 |
| joeedh | I heard the bmrt ray tracing core was cool. | 01:18 |
| joeedh | hrm. someone actually managed to combined deep shadow maps and penumbra maps. the papers a little confusing though, as the technique is heavily geared toward real-time use on gpus. | 01:24 |
| joeedh | http://www.iro.umontreal.ca/labs/infographie/papers/St-Amour-2005-PDSM/pdsm.pdf | 01:25 |
| *** joeedh is now known as joeedh_sleep | 01:42 | |
| *** Yomogi-aren has joined #aqsis | 02:10 | |
| |rt| | anyone here ever install cgkit on ubuntu | 02:17 |
| * cgTobi didn't. | 02:17 | |
| cgTobi | But what's the problem? | 02:17 |
| |rt| | just trying to figure out the dependancies | 02:17 |
| |rt| | was hoping that someone would know a repository that had cgkit in it :) | 02:18 |
| cgTobi | phew...good question | 02:19 |
| cgTobi | I don't know one, sorry | 02:20 |
| *** cgTobi sets mode: +v Yomogi-aren | 02:20 | |
| |rt| | there's alot of boost releated packages if I do apt-cache boost | 02:20 |
| |rt| | er apt-cache seach boost | 02:20 |
| |rt| | any idea what would satify the boost requirement for cgkit? | 02:21 |
| |rt| | libboost-dev maybe? | 02:21 |
| |rt| | there's also libboost-python-dev and libboost-python1.33.1 | 02:22 |
| cgTobi | I think libboost-python | 02:22 |
| cgTobi | maybe dev, but I am now sure | 02:22 |
| cgTobi | if you want to build it yourself probably dev | 02:22 |
| |rt| | well i don't see a way to get it on linux without building it | 02:23 |
| cgTobi | probably now | 02:23 |
| |rt| | if I could find a bin somewhere that would probably simplify things | 02:23 |
| cgTobi | not | 02:23 |
| cgTobi | building cgkit isn't that bad IIRC | 02:23 |
| cgTobi | packaging is another story | 02:23 |
| cgTobi | I tried to do cgkit rpms | 02:24 |
| cgTobi | and it was quite a pain | 02:24 |
| |rt| | bummer | 02:24 |
| |rt| | cgkit is much easier to deal with in windows | 02:24 |
| |rt| | which is odd | 02:24 |
| cgTobi | is it? I never used it on windows | 02:25 |
| |rt| | there's an installer for windows | 02:25 |
| |rt| | binary | 02:25 |
| cgTobi | well, that's the old linux problem ;) no unified way to install except by (make; make install) ;-) | 02:26 |
| |rt| | yeah i guess so | 02:27 |
| |rt| | i was spoiled by gentoo which has cgkit in their repos | 02:27 |
| |rt| | cgkit works for gentoo since everything is from source i guess | 02:27 |
| *** joeedh_sleep has quit IRC | 02:33 | |
| *** Compaq_Owner has joined #aqsis | 02:39 | |
| *** joeedh_sleep has joined #aqsis | 02:39 | |
| |rt| | libboost-dev got me a bit farther | 02:40 |
| |rt| | looks like it's complaining about missing gl libs now | 02:40 |
| *** blend_dem has quit IRC | 02:49 | |
| cgTobi | gl libs? let me take a look at my config | 02:53 |
| |rt| | i think i got it all now | 02:53 |
| |rt| | compling | 02:53 |
| cgTobi | oh, ok than. good luck | 02:53 |
| |rt| | seems to be working :P | 03:02 |
| cgTobi | cool | 03:02 |
| |rt| | well this seems to work | 03:16 |
| |rt| | #!/usr/bin/python | 03:16 |
| |rt| | import cgkit | 03:16 |
| |rt| | print cgkit.cgkitinfo.version | 03:16 |
| |rt| | so i think i'm good to go | 03:17 |
| *** cgTobi sets mode: +v joeedh_sleep | 03:17 | |
| cgTobi | sweet :) | 03:17 |
| *** Yomoyomo has joined #aqsis | 03:26 | |
| *** Yomogi-aren has quit IRC | 03:27 | |
| *** Yomogi-aren has joined #aqsis | 03:42 | |
| *** Yomoyomo has quit IRC | 03:43 | |
| cgTobi | I'm off to bed | 03:45 |
| cgTobi | cya later | 03:45 |
| Pseudonym | Night. | 03:45 |
| *** cgTobi sets mode: +v Yomogi-aren | 03:45 | |
| *** Yomogi-aren has quit IRC | 04:01 | |
| *** joeedh_sleep is now known as joeedh | 04:39 | |
| *** joeedh is now known as joeedh_sleep28 | 05:23 | |
| *** render has joined #aqsis | 07:50 | |
| *** AlexK has joined #aqsis | 08:08 | |
| *** AlexK has joined #aqsis | 08:08 | |
| *** Pseudonym has quit IRC | 08:12 | |
| tcolgate | mornin' all | 10:19 |
| *** cgTobi has quit IRC | 10:30 | |
| *** Anteru has joined #aqsis | 10:59 | |
| *** Anteru has quit IRC | 11:03 | |
| *** Anteru has joined #aqsis | 11:14 | |
| *** Anteru has quit IRC | 11:17 | |
| *** Anteru has joined #aqsis | 11:24 | |
| *** Joron has joined #Aqsis | 11:33 | |
| Anteru | Hi Michel | 11:36 |
| Anteru | Your latest changes to bake2tif broke VC8, mixed code and declaration ;) | 11:36 |
| Joron | vc8 could not tolerate in a C file; oh well I will change. | 11:41 |
| Anteru | it's fixed already | 11:42 |
| Anteru | no worries | 11:42 |
| Joron | it missed stdlib.h ? | 11:43 |
| Anteru | no, the char* pt = NULL; came after some code | 11:43 |
| Joron | thanks | 11:44 |
| Joron | I don't normally use vc8, only mingw, and gcc on macosx. | 11:44 |
| Joron | somehow vc8 version of aqsis doesn't work for me. | 11:44 |
| Anteru | I thought GCC captures it with -ansi or -strict ... | 11:45 |
| Anteru | ^ What's the problem, I just built successfully with VC8 | 11:45 |
| Joron | It doesn't run at all; it crashes at start-up. | 11:45 |
| Anteru | Hmm, what error? | 11:45 |
| Anteru | Can you hook a debugger, or does it crash before user code starts? | 11:46 |
| Joron | something like libaqsis.dll is not legal manifest.... | 11:46 |
| Anteru | That should be fixed | 11:46 |
| Anteru | we're embedding the manifests | 11:46 |
| Anteru | This was a problem but should not appear any more | 11:47 |
| Joron | I did not have the problem than I synced I got the problem... then I went to a rage; I delete all the directories ; resync same thing (I remvoe all my examples on nt without realizing)... recompiling, sync svn, still the same thing. | 11:48 |
| Anteru | which revision did you try? | 11:49 |
| Joron | something like 2 weeks ago. | 11:49 |
| Anteru | Hmm, maybe you just got a revision before the manifest stuff was fixed | 11:50 |
| Anteru | After all, I have it /working here/ with OpenEXR etc. | 11:50 |
| Anteru | running >90% of the RTS | 11:50 |
| Joron | when the manifest was fixed I got the problem; I did not have any problem before... somehow the fix included something from svn to get the manifest created... I needed at that time to sync svn too but it did not solve anything... I switch back to mingw. | 11:51 |
| Anteru | Hmm. | 11:52 |
| Joron | which version of svn did you have ? | 11:52 |
| Anteru | I have checked it with 980 a few minutes ago | 11:52 |
| Anteru | C:\Dev\Aqsis\aqsis\output\bin>aqsis --version | 11:53 |
| Anteru | aqsis version 1.3.0 | 11:53 |
| Anteru | compiled Apr 24 2007 12:21:21 | 11:53 |
| Joron | It runs without problem ? | 11:53 |
| Anteru | Obviously ^ | 11:53 |
| Joron | I will try again this morning later; I still did not like the fact you remove fltk vc8 libraries. | 11:54 |
| Anteru | Just a moment | 11:55 |
| Anteru | ^^ The FLTK binaries are different to mine ... | 11:55 |
| Anteru | http://anteru.net/temp/aq.png | 11:55 |
| Anteru | I'm not sure they are 100% compatible | 11:56 |
| Joron | configure you aqsis/scons to use to a different directory then. | 11:57 |
| Anteru | or we could check them in with the full compiler version | 11:57 |
| Anteru | i.e. vc8_14.00.72 | 11:57 |
| Joron | What I don't get it is you remove the fltk library for teh rest of us which are not interessted to compile even the fltk. | 11:58 |
| Joron | vc8_sp1, vc8_sp2, vc8,... | 11:58 |
| Anteru | I wanted to avoid problem with SP1 users using FLTK from the repository and having a possibly broken build | 11:58 |
| Anteru | for now, you can simply check out an old revision with libs | 12:01 |
| Anteru | I could check in OpenEXR libs with SP1 then | 12:02 |
| Joron | I could | 12:03 |
| Anteru | I'd just suggest we use something like vc8_build number at least to avoid any problems | 12:04 |
| Joron | vc8_sp1 | 12:04 |
| Joron | ? | 12:04 |
| Anteru | Do you have a warning free OpenEXR build? | 12:05 |
| Anteru | I can check in the OpenEXR SP1 hand made one now | 12:05 |
| Joron | That I don't remember since it has been awhile I touch exr. | 12:05 |
| Anteru | I'll check in VC8_SP1 libs then | 12:05 |
| Anteru | openexr is 26.4 mib ... | 12:08 |
| Anteru | can take a while | 12:08 |
| Anteru | 6 of 24 mib checked in | 12:12 |
| Anteru | done | 12:13 |
| Anteru | feel free to try with OpenEXR and FLTK | 12:13 |
| Joron | Is VC8 always faster than the el-cheapo mingw/gcc ? | 12:14 |
| Anteru | Think so | 12:15 |
| Anteru | I did just a quick try and on that particular scene, MSVC was faster, but YMMV | 12:15 |
| Joron | what does mean Ymmv ? | 12:16 |
| Anteru | Your mileage may vary | 12:17 |
| Joron | The only thin I know is vc8 is faster than vc6 but necessary than vc7.1 and vc8 the math are faster than anything I could come up. | 12:19 |
| Joron | s/necessary/not necessary/ | 12:19 |
| Anteru | the interprocedural optimization step gives a 30% boost over VC8 stock :) | 12:20 |
| Anteru | cya, battery empty | 12:33 |
| *** Anteru has quit IRC | 12:34 | |
| *** cgTobi has joined #Aqsis | 12:45 | |
| *** ChanServ sets mode: +o cgTobi | 12:45 | |
| cgTobi | hi all | 12:45 |
| *** cgTobi sets mode: +vvv AlexK Joron render | 12:46 | |
| Joron | Hello Tobias | 12:46 |
| Joron | I got to go to work bye bye. | 12:46 |
| *** Joron has quit IRC | 12:47 | |
| *** ShortWave is now known as ShortWaveWrk | 14:00 | |
| *** renderguy has joined #aqsis | 14:45 | |
| renderguy | Afternoon all. | 14:45 |
| *** cgTobi sets mode: +v renderguy | 14:57 | |
| cgTobi | hi Leon | 14:57 |
| *** Anteru has joined #aqsis | 15:31 | |
| renderguy | Further to Andrew's and Joe's DSM 'Patent' discussion earlier, the following resource may be interesting - http://www.freepatentsonline.com/result.html?query_txt=Pixar%20(Emeryville,%20CA) | 16:03 |
| cgTobi | Hi Leon | 16:04 |
| cgTobi | any success with scons yet? | 16:04 |
| renderguy | cgTobi: Didn't get a chnace to do anything last night, but should tonight. | 16:05 |
| renderguy | ^^ Actually, that patent search is quite interesting. | 16:07 |
| renderguy | ... in general, not just the Pixar stuff. | 16:09 |
| * renderguy is looking forward to seeing some MakeHuman 'Skin' tests from the MH team. | 16:44 | |
| * Anteru is looking forward for the test human ;) | 16:46 | |
| renderguy | Anteru: Didn't we, once, have MSVC6 libs for OpenEXR? | 16:49 |
| *** AlexK has quit IRC | 16:49 | |
| Anteru | Not really | 16:50 |
| Anteru | OpenEXR does not compile with VC6 | 16:50 |
| renderguy | Anteru: K, just my bad memory then. | 16:51 |
| Anteru | Might be that it did compile some years ago | 16:52 |
| renderguy | Gotta shoot, laterz. | 17:22 |
| *** renderguy has left #aqsis | 17:22 | |
| *** AlexK has joined #aqsis | 17:51 | |
| *** ShortWave has joined #aqsis | 18:48 | |
| *** Anteru has quit IRC | 18:54 | |
| *** Anteru has joined #aqsis | 18:59 | |
| *** AlexK has quit IRC | 20:04 | |
| Anteru | ShortWaveWrk: Around? | 20:38 |
| Anteru | ShortWave: Around? | 20:38 |
| ShortWave | I'm aroud | 20:48 |
| ShortWave | 'sup? | 20:48 |
| ShortWave | err | 20:48 |
| ShortWave | around | 20:48 |
| ShortWave | sorta | 20:48 |
| *** cgTobi sets mode: +vv Anteru ShortWave | 21:04 | |
| cgTobi | hi Bobby | 21:05 |
| Anteru | Ahh! | 21:15 |
| Anteru | ShortWave: Hiho | 21:16 |
| Anteru | ShortWave: Query | 21:16 |
| ShortWave | Hey tobi | 21:21 |
| ShortWave | long time no talk | 21:21 |
| ShortWave | 'sup? | 21:21 |
| |rt| | hey ShortWave | 21:35 |
| cgTobi | yeah, quite some time. Well, not much. Still in London. Still working. So everything is fine from my side. How about you? | 21:37 |
| *** Anteru has quit IRC | 22:33 | |
| *** ShortWave has quit IRC | 22:37 | |
| *** cgTobi has quit IRC | 22:54 | |
| *** cgTobi has joined #Aqsis | 23:04 | |
| *** ChanServ sets mode: +o cgTobi | 23:04 | |
Generated by irclog2html.py 2.5 by Marius Gedminas - find it at mg.pov.lt!
