#jogamp @ irc.freenode.net - 20160314 05:05:29 (UTC)

20160314 05:05:29 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20160313050529.html
20160314 05:05:29 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20160314050529.html
20160314 07:15:32 * odinsbane (~matt@anon) Quit (*.net *.split)
20160314 07:16:07 * odinsbane (~matt@anon) has joined #jogamp
20160314 07:24:15 * elect (~GBarbieri@anon) has joined #jogamp
20160314 07:24:47 <elect> hey
20160314 08:40:26 * monsieur_max (~maxime@anon) has joined #jogamp
20160314 09:01:26 * zubzub_ is now known as zubzub
20160314 09:24:19 * elect (~GBarbieri@anon) Quit (Read error: Connection reset by peer)
20160314 09:24:30 * elect (~GBarbieri@anon) has joined #jogamp
20160314 11:24:43 * elect_ (~GBarbieri@anon) has joined #jogamp
20160314 11:24:43 * elect (~GBarbieri@anon) Quit (Read error: Connection reset by peer)
20160314 11:27:55 * jvanek (jvanek@anon) has joined #jogamp
20160314 13:24:41 * elect_ (~GBarbieri@anon) Quit (Read error: Connection reset by peer)
20160314 13:24:48 * elect_ (~GBarbieri@anon) has joined #jogamp
20160314 13:37:37 * philjord (599cc172@anon) has joined #jogamp
20160314 14:39:40 * dwbrite (4474ac1d@anon) has joined #jogamp
20160314 16:51:27 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20160314 16:57:51 * jvanek (jvanek@anon) Quit (Quit: Leaving)
20160314 17:43:24 * monsieur_max (~maxime@anon) has joined #jogamp
20160314 18:23:35 * Eclesia (~eclesia@anon) has joined #jogamp
20160314 18:23:47 <Eclesia> o/
20160314 18:56:02 * elect_ (~GBarbieri@anon) Quit (Ping timeout: 244 seconds)
20160314 19:24:52 <Eclesia> rmk0: hi, may I borrow your expert eye for a second ? mine's are defective :/
20160314 19:26:01 <Eclesia> when rendering 2 spheres I obtain this in the normal attachement : http://unlicense.developpez.com/temporaire/test.png
20160314 19:26:53 <Eclesia> seems like somekind of blending or mask issue
20160314 19:31:16 * rmk0 eyes it
20160314 19:31:53 <rmk0> are back faces culled?
20160314 19:32:13 <rmk0> also... what's in the stencil buffer, and is testing/writing enabled/disabled?
20160314 19:32:56 <Eclesia> depth testing is enable at least
20160314 19:33:00 <Eclesia> checking the others
20160314 19:34:19 <rmk0> it looks suspiciously like the problem is precisely in the intersection of the two meshes, which suggests an interaction with the depth/stencil buffer
20160314 19:34:29 <Eclesia> backface culling is active
20160314 19:35:21 <rmk0> are you using apitrace?
20160314 19:35:41 <Eclesia> :x nope
20160314 19:35:41 <rmk0> there's an incredibly useful setting that tells you all the bits of state that aren't set to the defaults
20160314 19:35:51 <rmk0> really useful for tracking down this sort of nonsense
20160314 19:36:30 <Eclesia> apitrace... I always forget this one ^^
20160314 19:36:42 <rmk0> i don't think i could write opengl without it
20160314 19:37:37 <rmk0> there are other oddities like glLogicOp
20160314 19:37:43 <rmk0> but i'm guessing you've not touched the defaults there
20160314 19:37:49 <rmk0> not sure why it's still in the API at all
20160314 19:47:22 <Eclesia> hm... deth test is on, stencil test on, masks are at 255...
20160314 19:48:15 <Eclesia> trace is here : http://unlicense.developpez.com/temporaire/java.trace
20160314 19:49:11 <Eclesia> ha, GL_STENCIL_WRITEMASK = 0. but i don't use stencils so it should no be of any use
20160314 19:59:30 <rmk0> i suspect i won't be able to run that trace (only have GL3 here)
20160314 19:59:42 <rmk0> have you found the "only show non-defaults" state inspection thing?
20160314 19:59:48 <rmk0> in qapitrace, that is
20160314 20:00:07 <Eclesia> yes
20160314 20:01:40 <Eclesia> rmk0: http://unlicense.developpez.com/temporaire/test2.jpeg
20160314 20:02:33 <Eclesia> I don't see anything 'not normal' :/
20160314 20:02:34 * rmk0 eyes it
20160314 20:03:28 <rmk0> i.. wonder
20160314 20:03:38 <rmk0> you're rendering a g-buffer pass?
20160314 20:03:43 <Eclesia> yes
20160314 20:03:47 <rmk0> not... sure why blending would be enabled
20160314 20:04:17 * Eclesia testing without, 1 sec
20160314 20:05:34 <Eclesia> :(
20160314 20:05:44 <rmk0> no luck?
20160314 20:05:48 <Eclesia> nope, same
20160314 20:05:55 <rmk0> didn't think so
20160314 20:06:22 <Eclesia> I'll confirm it in 1min, to check with apitrace
20160314 20:09:45 <Eclesia> !!! still at true
20160314 20:10:02 * rmk0 bursts
20160314 20:10:17 <Eclesia> must have something overriding my value somewhere... bug hunting time
20160314 20:10:34 <rmk0> global
20160314 20:10:35 <rmk0> mutable
20160314 20:10:36 <rmk0> state
20160314 20:10:49 <rmk0> wasting developer time and sanity since 1972
20160314 20:12:50 <Eclesia> I should attached those property on my FBO class, and reconfigure everything only when fbo.enable() is called
20160314 20:14:26 <bruce-> Eclesia: what is the state of your fxaa code?
20160314 20:15:45 <Eclesia> more or less working, but as I told you last time, there is a doubt on the license. since the original blog post has been removed it's hard to know for sure
20160314 20:16:37 <rmk0> the author is contactable
20160314 20:17:05 <rmk0> he told me not to bother with fxaa because there are better alternatives (i actually asked him what improvements had been made since the original version, because the code is totally different)
20160314 20:17:13 <rmk0> i used fxaa anyway
20160314 20:17:45 <Eclesia> I like old code :) it's often public domain ^^
20160314 20:18:00 <bruce-> there is SMAA
20160314 20:18:10 <rmk0> that's what he recommended
20160314 20:18:20 <bruce-> http://www.iryoku.com/smaa/
20160314 20:18:22 <bruce-> ah, es
20160314 20:18:23 <bruce-> yes
20160314 20:19:45 <rmk0> Eclesia: you probably could get a public domain dedication out of the author
20160314 20:20:17 <rmk0> of fxaa, that is
20160314 20:20:33 <rmk0> he's already done it once, he just did it poorly
20160314 20:20:43 <Eclesia> that would be great. you have his email ?
20160314 20:21:01 <rmk0> timothy.lottes@epicgames.com
20160314 20:21:10 <Eclesia> thanks
20160314 20:29:56 <Eclesia> Arg! fount it
20160314 20:30:13 <Eclesia> I even had a TODO beside it ... shame on me
20160314 20:30:15 <Eclesia> //enable blending : TODO should it really be done here ? find a more appropriate place
20160314 20:30:16 <Eclesia> GLUtilities.configureBlending(gl, AlphaBlending.DEFAULT);
20160314 20:30:31 <rmk0> hehe
20160314 20:37:17 <Eclesia> rmk0: you saved me once again. _^-o_ (ain't easy to make something like a bowing stickman)
20160314 20:37:29 <rmk0> you saved yourself!
20160314 20:38:49 <Eclesia> perfect example of 'Help Yourself And God Will Help You' lol
20160314 20:40:26 <Eclesia> back to the question : should the FBO class contains all the 'states/masks' ?
20160314 20:44:12 <rmk0> don't think so
20160314 20:45:51 <Eclesia> might be specific to each mesh rendering ...
20160314 20:49:08 <Eclesia> meshes with alpha values are messed up now :/
20160314 21:02:27 <Eclesia> rmk0: how do you some it ? no transparency in meshes ? or 2 pass rendering ?
20160314 21:02:31 <Eclesia> solve*
20160314 21:03:45 <rmk0> i render all opaque instances into a g-buffer, then do SSAO on that g-buffer, then render all light volumes into an l-buffer, then combine the g-buffer, l-buffer, and SSAO map into a lit image
20160314 21:03:50 <rmk0> then i render translucent instances
20160314 21:04:53 <rmk0> no blending for the g-buffer or SSAO, pure additive blending for the light buffer
20160314 21:05:13 <rmk0> arbitrary blending for translucent instances
20160314 21:06:15 <Eclesia> translucent instance must kind a 'pop up' a bit with this pipeline no ?
20160314 21:06:26 <rmk0> how d'you mean?
20160314 21:06:41 <Eclesia> since they don't benefeat from lights and effects
20160314 21:07:16 <rmk0> they can still be rendered with lighting, it's just that they do that via forward rendering
20160314 21:08:04 <Eclesia> I see, that must be a more intense fragment shader
20160314 21:08:22 <rmk0> i only support very limited lighting for translucent instances
20160314 21:08:45 <Eclesia> sure I can imagine that
20160314 21:08:49 <rmk0> translucent instances tend to either be glass or some sort of environmental effect like fire
20160314 21:09:00 <rmk0> in the latter case... lighting obviously doesn't matter
20160314 21:10:28 <rmk0> http://waste.io7m.com/2016/02/29/timelapse.avi
20160314 21:10:46 <rmk0> you might enjoy that. is a timelapse trace of a single frame from deus ex human revolution
20160314 21:11:23 <rmk0> it follows a similar approach, except that it renders normals, then lighting, then lit opaque surfaces in three passes
20160314 21:11:32 <rmk0> then finally does all the translucent instances, then bloom
20160314 21:14:10 <Eclesia> nice vid
20160314 21:14:54 <rmk0> those are obviously individual draw calls
20160314 21:17:39 <Eclesia> doesn't seem like there is a proper way to deal with translucent objects. things like stained glass or thin curtains must be the biggest problem
20160314 21:18:23 <rmk0> it is doable, but it's expensive
20160314 21:19:16 <rmk0> i mean, if those things are supposed to attenuate light to some extent, or result in coloured shadows
20160314 21:20:22 <rmk0> we need hardware that can do easy realtime GI
20160314 21:20:24 <rmk0> problem solved!
20160314 21:20:58 <Eclesia> ^^
20160314 21:22:26 <Eclesia> rmk0: we need human clones a 1/10 scale and we make them 'play' in the screen. of course you would need multiple of them for an fps :D
20160314 21:22:37 <rmk0> /o\
20160314 21:22:40 <Eclesia> like a small theater
20160314 21:23:23 <Eclesia> ho there is blood bleeding out of the screen !
20160314 21:24:12 <rmk0> the cleanup costs might be prohibitive
20160314 21:25:06 <Eclesia> yeah stage loading might be a bit long
20160314 21:54:20 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20160314 23:25:18 <Eclesia> good night ++
20160314 23:25:24 * Eclesia (~eclesia@anon) Quit (Quit: Leaving.)
20160315 00:12:35 * philjord (599cc172@anon) Quit (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client)
20160315 05:05:30 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20160315050530.html