#jogamp @ irc.freenode.net - 20141031 05:06:25 (UTC)

20141031 05:06:25 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20141030050625.html
20141031 05:06:25 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20141031050625.html
20141031 07:32:09 * eclesia (~husky@anon) has joined #jogamp
20141031 07:32:16 <eclesia> good morning
20141031 07:35:31 * xranby (~xranby@anon) Quit (Ping timeout: 255 seconds)
20141031 07:36:29 * xranby (~xranby@anon) has joined #jogamp
20141031 08:00:47 * monsieur_max (~maxime@anon) has joined #jogamp
20141031 08:00:47 * monsieur_max (~maxime@anon) Quit (Client Quit)
20141031 08:02:12 * monsieur_max (~maxime@anon) has joined #jogamp
20141031 08:02:41 * jvanek (jvanek@anon) has joined #jogamp
20141031 08:12:55 * zzuegg (~zzuegg@anon) has joined #jogamp
20141031 08:35:39 * xranby (~xranby@anon) Quit (Ping timeout: 245 seconds)
20141031 08:37:46 * xranby (~xranby@anon) has joined #jogamp
20141031 09:12:11 * doev (~doev@anon) has joined #jogamp
20141031 09:44:20 * doev (~doev@anon) Quit (Quit: Verlassend)
20141031 12:45:43 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20141031 12:47:06 * zzuegg (~zzuegg@anon) Quit (Ping timeout: 258 seconds)
20141031 12:48:26 * zzuegg (~zzuegg@anon) has joined #jogamp
20141031 13:37:16 * Nosezeichen (~Nosezeich@anon) has joined #jogamp
20141031 14:36:31 <Nosezeichen> can anyone help me with this problem ?
20141031 14:36:33 <Nosezeichen> http://forum.jogamp.org/Offscreen-Rendering-with-GLCanvas-which-wasn-t-created-with-this-capability-td4033474.html
20141031 14:37:13 * jvanek (jvanek@anon) Quit (Quit: Leaving)
20141031 14:40:58 <rmk0> interesting problem...
20141031 14:42:04 <rmk0> the third party code... do you know if it does anything with framebuffer objects internally?
20141031 14:58:26 * eclesia (~husky@anon) Quit (Ping timeout: 264 seconds)
20141031 15:11:37 * eclesia (~husky@anon) has joined #jogamp
20141031 15:24:22 <Nosezeichen> i have to look for it
20141031 15:24:53 <Nosezeichen> when debugging, can I see this in the GLContext ?
20141031 15:25:38 <rmk0> hm, not sure what the easiest way would be to check
20141031 15:25:59 <rmk0> the third party code... what sort of control do you have over it?
20141031 15:26:11 <rmk0> is there any way to make it render with a specific GL interface that you give it?
20141031 15:26:53 <rmk0> i can think of quite a few ways to achieve what you're trying to do, but it's a matter of exactly how much you're able to influence the third party code
20141031 15:28:35 * eclesia (~husky@anon) Quit (Ping timeout: 272 seconds)
20141031 15:28:58 <Nosezeichen> hm I didnt want to change thwe third party code. I decompiled it and I think I could just modify the GLCapabilities for Offscreen rendering and then create that canvas
20141031 15:29:10 <Nosezeichen> but I would prefer not modifying it
20141031 15:29:16 <Nosezeichen> if theres a way
20141031 15:29:21 <rmk0> oh, it's not a matter of modifying it
20141031 15:29:50 <Nosezeichen> what do you mean ?
20141031 15:29:56 <rmk0> what i was thinking was... every time it thinks it's rendering to "the screen", it's actually rendering to a framebuffer object that you've created
20141031 15:30:42 <rmk0> if the third party code never explicitly does anything with framebuffer objects, then by simply creating a new framebuffer object, binding it, then running the third party code, you automatically get it to render to your framebuffer
20141031 15:30:52 <rmk0> you can then use the framebuffer contents as a texture in the way you want
20141031 15:31:24 <rmk0> if the third party code creates its own framebuffer objects and switches between them during rendering, it's a bit harder to deal with
20141031 15:31:40 <Nosezeichen> that was my idea too. but for this mustn't the offscreen rendering to be activated for the GLCanvas ?
20141031 15:32:16 <rmk0> not that i know of
20141031 15:32:31 * rmk0 prods sgothel
20141031 15:33:01 <Nosezeichen> ok I'll have a look to the 3rd party code and if its using framebuffers internally
20141031 15:33:21 <Nosezeichen> and then write here again
20141031 15:33:33 <rmk0> the reason i asked about getting the third party code to use a GL instance that you control is that you can obviously log the calls
20141031 15:33:44 <rmk0> see TraceGL3, TraceGL4, etc
20141031 15:33:49 <rmk0> that'd tell you what you need to know
20141031 15:34:27 <rmk0> otherwise you'd probably have to use something like http://apitrace.github.io to be sure (if you can't read the source code to the third party code!)
20141031 15:36:54 <Nosezeichen> wouldn't something like:
20141031 15:37:14 <Nosezeichen> glacanvas.getGL().setTrace... work ?
20141031 15:39:06 <rmk0> not familiar with that method
20141031 15:39:13 <rmk0> are you sure it's part of the GL* interfaces?
20141031 15:39:27 <rmk0> ah... i see
20141031 15:39:42 <rmk0> glcanvas.setGL(new TraceGL3(glcanvas.getGL()))
20141031 15:39:48 <rmk0> that'd most likely do it
20141031 15:40:00 <Nosezeichen> I'll try it
20141031 15:40:10 <rmk0> you'd need to pick a specific TraceGL*
20141031 15:40:17 <rmk0> not sure which opengl profile you'd be using
20141031 15:40:30 <Nosezeichen> the 3rd party app uses GL2 profile
20141031 15:40:36 <rmk0> right
20141031 15:41:15 <Nosezeichen> If you want to take a look into the 3rd party app: http://vhg.cmp.uea.ac.uk/tech/jas/local/jas.zip
20141031 15:42:05 <Nosezeichen> there is a jar file in it: jas.zip\jas\095s\jars\jarpb.jar
20141031 15:43:28 <Nosezeichen> you can use the online java-decompiler to look into the class jarp.View (online java decompiler at the bottom of http://jd.benow.ca/)
20141031 15:43:43 <Nosezeichen> there is the canvas related code
20141031 15:43:49 <xranby> http://vh.cmp.uea.ac.uk/index.php/JASigning_Conditions
20141031 15:44:02 <xranby> The JASigning software accessible via JASigning is Copyright UEA (2005-2014) and may be downloaded and used for evaluation purposes only. For enquires about public and commercial use of the software and the virtual characters Anna and Marc, please email CMP_VirtualSigning@uea.ac.uk.
20141031 15:45:09 <Nosezeichen> yes I've stayed in contact with the authorithies from the university
20141031 15:45:19 <Nosezeichen> but for another topic
20141031 15:45:23 * eclesia (~husky@anon) has joined #jogamp
20141031 15:45:53 <xranby> the licese makes it difficult for a third party to help you i am afraid
20141031 15:47:48 <Nosezeichen> I think thats something where the code has to be changed from the authors. Im evaluating it right now. but real support you only get if you are getting a licence from them. at that time I dont know for sure if it can work for me
20141031 15:48:28 <Nosezeichen> It would be a part of my master thesis
20141031 15:49:19 <Nosezeichen> if I can't solve this problem "on my own" (you guys are helping :)) I'll have to contact them again
20141031 15:50:16 <Nosezeichen> Or did you mean something else ?
20141031 15:52:39 <xranby> i would put myself into trouble by helping you due to the licence of the code
20141031 15:53:26 <xranby> we gladly help with free software integration
20141031 15:54:37 <Nosezeichen> ok than back to theory
20141031 15:54:50 <xranby> if the university needs help working on commercial support inquiries then send an email to : http://jogamp.org/wiki/index.php/Maintainer_and_Contacts#Contact_Maintainer
20141031 15:56:35 <Nosezeichen> right now its just me working on my master thesis
20141031 15:57:59 <Nosezeichen> but when I don't get it working until the weekend I'll contact the authors of the 3rd party app anyway again
20141031 16:01:12 <xranby> what are the currect state of free software 3D virtual characters to perform signs libraries?
20141031 16:01:46 <zubzub> my guess would be close to none existant
20141031 16:06:37 <Nosezeichen> yes
20141031 16:06:57 <Nosezeichen> the jasigning project is an EU funded project
20141031 16:08:20 * eclesia (~husky@anon) has left #jogamp
20141031 16:20:01 <Nosezeichen> I only see glBindTexture calls in the GLTrace, so is this an indicator that framebuffers are not used ?
20141031 16:20:11 <rmk0> i'd assume so
20141031 16:20:48 <Nosezeichen> otherwise I would expect something like glBindBuffer, glBindFramebuffer ...
20141031 16:23:22 <Nosezeichen> ok than I'll try to bind a framebuffer to the canvas
20141031 16:42:55 * monsieur_max (~maxime@anon) has joined #jogamp
20141031 17:35:47 * Nosezeichen (~Nosezeich@anon) Quit (Ping timeout: 265 seconds)
20141031 17:42:14 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20141031 17:45:07 * Eclesia (~eclesia@anon) has joined #jogamp
20141031 17:54:01 * Eclesia (~eclesia@anon) Quit (Quit: Leaving.)
20141031 18:03:16 <rmk0> trying to find information on graphics algorithms that almost everyone uses makes me want to gouge out eyes and stomp on throats
20141031 18:03:33 <rmk0> cannot believe how awful and/or nonexistent information online is on commonly implemented algorithms
20141031 18:03:46 <rmk0> is like everyone's just copying and pasting from everyone else with no understanding or explanation
20141031 18:03:54 <rmk0> this time it's FXAA
20141031 18:03:59 <rmk0> the original paper is okish
20141031 18:04:11 <rmk0> there have been at least four major revisions to the algorithm since then, completely undocumented
20141031 18:04:32 <rmk0> the author apparently published details on his blog in the past, but cleverly decided to "spring clean" his blog a couple of years ago and threw them all away
20141031 18:04:43 <sgothel> fxaa is also in graph shader .. not good for this purpose (whole AA testbed :)
20141031 18:05:02 <rmk0> i'm not interested in code, i want a full description of the current algorithm
20141031 18:05:07 <rmk0> one just outright doesn't exist online
20141031 18:05:33 <rmk0> there's the original paper, and a pathetic set of 15 slides explaining how great the newest version is, with no detail at all
20141031 18:06:18 <rmk0> the only information available for version 3.11 of the algorithm is ~800 lines of #ifdef-ridden nonsensical garbage
20141031 18:06:25 <sgothel> all papers about AA are like 'art' ..
20141031 18:06:44 <sgothel> will email you my link list :)
20141031 18:06:49 <rmk0> as far as i can tell, all papers on graphics techniques might as well be printed on toilet paper
20141031 18:06:59 <sgothel> relax :)
20141031 18:07:11 <rmk0> i'm relaxed, just been fighting this shit for years
20141031 18:07:42 <rmk0> am really sick of getting a paper that claims to describe an algorithm... turns out to be five pages, three of which are explaining how great it is and the last one is a few percentages about how much faster it is than x
20141031 18:07:58 <rmk0> might get a few equations, with no explanation
20141031 18:08:18 <rmk0> wondering if this is due to researchers in the US being forced to eject N papers for year to keep their positions
20141031 18:08:31 <sgothel> most of them .. :)
20141031 18:09:29 <rmk0> will implement FXAA version 1
20141031 18:09:42 <rmk0> i can claim that later versions don't exist any nobody will be able to prove me wrong!
20141031 18:09:45 <rmk0> *and
20141031 18:09:58 <sgothel> FXAA is only good for .. err .. photorealistic stuff .. not so much 'lineart' like text ..
20141031 18:10:10 <rmk0> urhur
20141031 18:11:11 <rmk0> i'm using it because it's a postprocess
20141031 18:11:32 <rmk0> hehehe
20141031 18:11:47 <rmk0> just saw a comment online... "indeed, the FXAA whitepaper makes for good toilet paper"
20141031 18:11:52 <rmk0> so i'm not the only one
20141031 18:12:20 <rmk0> it bothers me how poor the available information is, and how quickly it's lost
20141031 18:12:34 <rmk0> in a decade or so, we won't have a clue how most things were implemented now
20141031 18:12:47 <sgothel> brute-force (weight all pixels) best, then flipquad3, .. etc
20141031 18:13:16 <sgothel> quincunx :)
20141031 18:15:30 <sgothel> jogl/src/jogl/classes/jogamp/graph/curve/opengl/shader <- many are impl. there
20141031 18:15:55 <rmk0> like i said... i don't want code
20141031 18:16:03 <rmk0> i could copy and paste any number of implementations
20141031 18:16:06 <rmk0> this is exactly the problem!
20141031 18:16:11 <sgothel> yup
20141031 18:16:29 <rmk0> for this renderer... each time a technique's completed, i've documented it fully
20141031 18:16:41 <rmk0> explained everything, equations in haskell, diagrams, etc
20141031 18:17:05 <rmk0> i can't do that if i don't understand why somethin works
20141031 18:17:06 <rmk0> g
20141031 18:17:10 <sgothel> its a psychedelic art to make something rendered in lower frequency look as-if done in high-frequency, that is my take on it
20141031 18:17:19 <sgothel> and it surely depends on what you render
20141031 18:17:20 <rmk0> hehe
20141031 18:17:25 <sgothel> (like mp3 etc)
20141031 18:24:13 <rmk0> i only care about smoothing the pseudo-photorealistic output of a deferred renderer
20141031 18:24:33 <rmk0> wouldn't be applying it to a ui or anything like that
20141031 18:25:34 <sgothel> why not impl. them all .. and see for yourself ? brute-f, fliquqad, quincunx, fxaa, mlaa, ..
20141031 18:25:38 <sgothel> laundry ..
20141031 18:25:44 <rmk0> don't have time to do that
20141031 18:25:51 <rmk0> i know fxaa works
20141031 18:26:03 <sgothel> I did .. and my eyes were hurting afterwards :)
20141031 18:26:16 <sgothel> Qun said: looks all the same :)
20141031 18:26:30 <rmk0> heh
20141031 18:29:37 <rmk0> think blog software has done a massive amount of harm to the net
20141031 18:30:11 <rmk0> people put important information on them, everyone links to them... single point of failure
20141031 18:30:19 <rmk0> a single upgrade and all the content is destroyed
20141031 18:30:23 <rmk0> hideously fragile
20141031 18:30:51 <rmk0> most of the really OLD graphics stuff was researchers or amateurs producing simple static html pages
20141031 18:31:00 <rmk0> and they all still exist and look the same (if the original server still exists)
20141031 18:31:08 <rmk0> they're mirrored easily too
20141031 18:32:09 <rmk0> looks like i've seen all of your links
20141031 18:32:18 * rmk0 (~rmk0@anon) Quit (Quit: leaving)
20141031 18:32:40 * rmk0 (~rmk0@anon) has joined #jogamp
20141031 18:32:41 * rmk0 kicks irssi
20141031 19:05:17 * Eclesia (~eclesia@anon) has joined #jogamp
20141031 19:05:24 <Eclesia> hi
20141031 19:05:30 <rmk0> lo
20141031 19:06:09 <Eclesia> Question : what could cause a texture to be translucent on one pc and not on another ?
20141031 19:06:51 <Eclesia> texture is ok on a GTX480 but is translucent on a 660Ti
20141031 19:07:24 <rmk0> ...
20141031 19:08:52 <rmk0> don't know where to start with that one
20141031 19:10:22 <Eclesia> it's a transparent frame too, on windows 8
20141031 19:10:52 <rmk0> what makes you say the texture is "transparent"?
20141031 19:12:13 <Eclesia> the model did load correctly, otherwise the store would have raised an exception. and it's displayed too whithout gl exceptions. so shader and texture loading didn't crash either
20141031 19:13:11 <Eclesia> I'm clueless. only thing I might think of is some blending issue, or maybe color value range clipping
20141031 19:16:29 <rmk0> what i mean is...
20141031 19:16:37 <rmk0> textures aren't transparent or opaque
20141031 19:16:39 <rmk0> obviously
20141031 19:17:08 <rmk0> how you interpret their alpha channel determines how their colour data influences the scene... what makes you say the texture appears transparent?
20141031 19:18:06 <Eclesia> 1 sec, I upload the screenshot he sent.
20141031 19:18:10 <rmk0> ok
20141031 19:20:15 <Eclesia> erf ftp in maintenance. you know a free image hosting site ?
20141031 19:20:49 <rmk0> imagebin.org
20141031 19:21:36 <Eclesia> there : http://imagebin.org/322933
20141031 19:22:45 <rmk0> interesting...
20141031 19:23:26 <Eclesia> he is using my GamePhases class and obvious he just copied one of my demos. I reconize my grid ground at the bottom
20141031 19:24:27 <rmk0> does look like he's corrupting the blending state somehow
20141031 19:24:28 <Eclesia> look at the model hairs, it's as if the colors where there, there is some yellow and blue
20141031 19:25:22 <Eclesia> it means the bug is in my engine :/
20141031 19:28:55 <Eclesia> just to be sure, do I need to clip the color values in my fragment shaders, between 0 and 1 ?
20141031 19:29:27 <rmk0> i haven't to date... the behaviour is dependent on the format of the texture attached to the framebuffer
20141031 19:29:46 <rmk0> RGBA8 = "unsigned normalized fixed point"
20141031 19:29:53 <rmk0> anything you write to it will be clamped to [0, 1]
20141031 19:30:01 <rmk0> floating point formats don't have that restriction
20141031 19:30:04 * Eclesia checking
20141031 19:31:16 <Eclesia> InternalFormat.RGBA, Format.RGBA, Type.UNSIGNED_BYTE
20141031 19:31:33 <Eclesia> that's GL4.GL_RGBA
20141031 19:31:41 <rmk0> right
20141031 19:31:50 <Eclesia> should I use RGBA8 ?
20141031 19:32:09 <rmk0> i think what you have is equivalent
20141031 19:32:22 <rmk0> GL_RGBA is unsigned normalized fixed point
20141031 19:32:30 <rmk0> RGBA8 just mandates that it be 8-bits per channel
20141031 19:32:51 <Eclesia> could it be more efficient, to use it ?
20141031 19:33:06 <rmk0> unlikely
20141031 19:33:38 <rmk0> GL_RGBA is going to give the implementation freedom to pick whichever it likes best
20141031 19:33:44 <rmk0> is unlikely to give you a slower option
20141031 19:34:35 <Eclesia> maybe the driver picked a higher type ? I don't know for HDR perhaps
20141031 19:34:58 <rmk0> is possible, i suppose
20141031 19:35:07 <rmk0> the result will still be unsigned normalized fixed point
20141031 19:35:25 <rmk0> i mean... if you ask it for GL_RGBA, you're not going to get floating point, or signed integers, or whatever
20141031 19:36:21 <Eclesia> could be a driver issue on windows 8 ?
20141031 19:36:37 <rmk0> is opengl... it could always be a driver issue
20141031 19:36:41 <Eclesia> lol
20141031 19:37:14 <Eclesia> looks painfull to make gl work correctly as expected on all gpu ^^
20141031 19:37:31 <rmk0> the only safe approach seems to be to pick some tiny subset
20141031 19:52:34 <Eclesia> changing topic, I started working on the GLSL parser and optimizer, do you have somekind of optimization tips for shader ? replacement patterns, stuff like that
20141031 19:53:46 <rmk0> not really... the language is pretty anti-optimization
20141031 19:55:09 <Eclesia> your tool parasol doesn't make any ?
20141031 19:55:18 <rmk0> nope
20141031 19:55:30 <rmk0> there are things about it that make it produce code that the existing glsl compiler will cope better with
20141031 19:55:42 <rmk0> they're not "optimizations"
20141031 19:56:04 <Eclesia> let's call them 'recommandations' then
20141031 19:56:07 <sgothel> note: strongest reason for using source code w/ GLSL is to allow the vendor to perform all possible optimizations for the running GPU
20141031 19:56:37 <sgothel> performance optimizations .. that is
20141031 19:56:37 <Eclesia> sgothel: hi
20141031 19:56:40 <sgothel> hi
20141031 19:57:21 <sgothel> me just in-between .. throwing 'wisdom' between diapers :)
20141031 19:57:34 <rmk0> that's not a great thing to put in diapers
20141031 19:57:35 * Eclesia take notes
20141031 19:58:00 <sgothel> you can read alot out of full diapers .. seriously
20141031 19:58:32 <rmk0> assume you've found graphics papers in them before
20141031 19:59:17 <sgothel> graphic analysis can help to indicate insufficiencies here, yup
20141031 20:01:12 <Eclesia> just vocabulary : diapers are for babies, does it have a different meaning here ? :x
20141031 20:01:22 <sgothel> nope :)
20141031 20:02:05 <Eclesia> Well you really used it as if it was something awesome ^^
20141031 20:02:48 <sgothel> newborn .. ~3 weeks old (already ?) - regained birth weight + 1.3 kg .. yes, its awesome! :)
20141031 20:03:32 <sgothel> usually my off-topic is about politics .. guess that changes it a bit :)
20141031 20:04:17 <Eclesia> sgothel: good luck, I hope you sleep enough :D
20141031 20:04:55 <sgothel> thx .. yup, even though I reduced my input here a bit, but will recover in this regard in the next days/weeks :)
20141031 20:06:03 <Eclesia> no problem. life is life, you prepare the next jogamp maintener for next geek generation ^^
20141031 20:09:01 <Eclesia> I hope you read him stories. Once upon a time, little vertex entered the pixel shader ... and so on...
20141031 20:09:12 <Eclesia> to train him :)
20141031 20:09:59 <sgothel> '.. and then there came D3D HLSL along, from the dark side of a tiny soft moon ..'
20141031 20:11:59 <Eclesia> '.. little vertice, I am your dad said float buffer. Noooo you can't be, I'm an integer! "
20141031 20:13:18 <sgothel> then there came Mark, contributing std float bit representation for integer types to make them all getting along
20141031 20:17:34 <Eclesia> And they had many tesselated fragments and lived happily ever after in their fbo.
20141031 21:17:15 <Eclesia> sgothel: will the jogamp package change ?
20141031 21:17:43 <Eclesia> removing the javax.*
20141031 21:23:49 <Eclesia> good night +++
20141031 21:23:58 * Eclesia (~eclesia@anon) Quit (Quit: Leaving.)
20141031 21:42:22 * zzuegg (~zzuegg@anon) Quit (Quit: Nettalk6 - www.ntalk.de)
20141101 01:59:24 * xranby (~xranby@anon) Quit (Ping timeout: 245 seconds)
20141101 02:01:27 * xranby (~xranby@anon) has joined #jogamp
20141101 05:06:26 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20141101050626.html