#jogamp @ irc.freenode.net - 20150923 05:05:12 (UTC)


20150923 05:05:12 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20150922050512.html
20150923 05:05:12 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20150923050512.html
20150923 05:46:28 * monsieur_max (~maxime@anon) has joined #jogamp
20150923 05:48:49 * jvanek (jvanek@anon) has joined #jogamp
20150923 06:33:07 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150923 07:13:36 * monsieur_max (~maxime@anon) has joined #jogamp
20150923 07:23:47 * eclesia (~husky@anon) has joined #jogamp
20150923 07:23:52 <eclesia> good morning
20150923 07:39:24 * doev (~doev@anon) has joined #jogamp
20150923 07:41:27 * doev (~doev@anon) Quit (Client Quit)
20150923 08:42:56 <xranby_f22> eclesia: good morning, i like to inform you that a raspberry pi user has published a super minimal EGL GLES2 binding with keyboard and mouse tracker under public domain https://www.raspberrypi.org/forums/viewtopic.php?f=81&t=121400
20150923 08:43:54 <xranby_f22> The emerge of raspberry pi touch screens has uncovered several cornercases at least for me when reciving absolute screen coordinates using the touch events. Thus I enjoy seeing that this persons mouse handler contain some code to remedy that.
20150923 09:19:41 <xranby_f22> i have filed a new bug for this touch input issue https://jogamp.org/bugzilla/show_bug.cgi?id=1227
20150923 10:30:37 * gouessej (5ee4b442@anon) has joined #jogamp
20150923 10:31:23 <gouessej> Hi
20150923 10:31:52 <gouessej> xranby_f22: I see that I really need to work on the NEWT Input API
20150923 10:34:42 <xranby_f22> gouessej: hi
20150923 10:35:10 <xranby_f22> gouessej: i have used your libgdx work to test the new 2.3.2 jogamp release on the raspberry pi
20150923 10:35:59 <xranby_f22> gouessej: yes it would be good to have NEWT Input API handle gaming devices
20150923 10:36:08 <xranby_f22> such as USB joysticks
20150923 10:36:31 <gouessej> but I have to spend a lot of time in maintaining the engines :s
20150923 10:37:03 <xranby_f22> KUDOS for that
20150923 10:37:22 <gouessej> Thanks
20150923 10:38:00 <xranby_f22> it is hard for some users to use your work only because we do not provide builds that they can download and use
20150923 10:38:06 <gouessej> There are often some small changes that force me to spend much time than I thought to update them
20150923 10:38:07 <xranby_f22> use our work
20150923 10:38:16 <gouessej> yes
20150923 10:38:31 <gouessej> but I do so for JogAmp's Ardor3D Continuation :)
20150923 10:38:46 <xranby_f22> thank you!
20150923 10:39:01 <xranby_f22> do you have a raspberry pi now to test Ardor3d on?
20150923 10:39:13 <gouessej> and it's easy to test with JMonkeyEngine 3, there are just default 2 settings to modify
20150923 10:39:19 <gouessej> yes
20150923 10:39:41 <gouessej> but I want to fix the support of OpenGL ES 1 before trying it on a Pi 2
20150923 10:40:08 <xranby_f22> using ImmModeSink should make it easy to get ES 1 working
20150923 10:40:37 <xranby_f22> gouessej: i used ImmModeSink one day ago to update your JOGLQuad demo to work on the raspberry pi
20150923 10:41:22 <gouessej> cool :) PMWMatrix is harder to use...
20150923 10:41:36 <xranby_f22> https://gist.github.com/xranby/ccbcfcaef341f5bd1ba8 <- JOGLQuadES1
20150923 10:42:11 <xranby_f22> yes PMWMatrix will be needed for JOGLQuadES2
20150923 10:42:35 <xranby_f22> the above code may contain bugs,, at least it render a quad however it is always white hmm
20150923 10:42:49 <xranby_f22> on both desktop and on my pi
20150923 10:43:11 <xranby_f22> thus i guess i have misunderstood how to enable correct light using ImmModeSink
20150923 10:43:31 <gouessej> It shouldn't be white
20150923 10:43:46 <xranby_f22> yes i realized that after i posted the code :/
20150923 10:44:01 <gouessej> You should call glColor3f as many times as you call glVertex, maybe it helps
20150923 10:44:22 <xranby_f22> GL2 allow code such as
20150923 10:44:29 <gouessej> the creation of the ImmModeSink object seems to be wrong
20150923 10:44:41 <gouessej> 0, GL.GL_FLOAT, // color ???
20150923 10:44:59 <xranby_f22> glColor3f(0.0f, 1.0f, 1.0f); glBegin(glSink.GL_QUADS); glVertex3f(-1.0f, 1.0f, 0.0f);
20150923 10:45:11 <xranby_f22> however the ImmModeSink only allow
20150923 10:45:12 <gouessej> 3, GL.GL_FLOAT, // texCoords ??? why?
20150923 10:45:26 <xranby_f22> glBegin(glSink.GL_QUADS); glColor3f(0.0f, 1.0f, 1.0f); glVertex3f(-1.0f, 1.0f, 0.0f);
20150923 10:46:00 <xranby_f22> gouessej: why is that i have not understood the api :/ and used cut and paste from other initialization
20150923 10:46:16 <xranby_f22> let me take a look and read how to correctly initialize it
20150923 10:46:17 <gouessej> there is no texture coordinates
20150923 10:46:29 <gouessej> there are no texture coordinates
20150923 10:46:54 <xranby_f22> kudois, that is likely the issue
20150923 10:47:11 <gouessej> You're welcome. We'll mention this example in the wiki :)
20150923 10:47:37 <xranby_f22> let me update it first to work
20150923 10:47:41 <gouessej> yes
20150923 10:47:42 <xranby_f22> with color
20150923 10:48:06 <gouessej> sgothel: Why is AWTAnimatorImpl called even when I use GLWindow?
20150923 10:48:42 <gouessej> xranby_f22: JogAmp's Ardor3D Continuation will become usable with an headless JRE
20150923 10:48:53 <xranby_f22> gouessej: very good!
20150923 10:49:16 <gouessej> I have moved the AWT craps into separate subprojects
20150923 10:49:37 <xranby_f22> i have color now !
20150923 10:49:58 <gouessej> It's a good piece of news
20150923 10:50:15 <gouessej> I'm impatient to see JoglQuadES2
20150923 10:50:40 <xranby_f22> and you are right need to have a glColor3f before each glVertex3f else you get black
20150923 10:51:01 <gouessej> com.jogamp.opengl.util.glsl.fixedfunc.FixedFuncUtil
20150923 10:51:01 <xranby_f22> let me update and push this gist then i can start on the ES2 verision :)
20150923 10:51:41 <gouessej> There is no hurry
20150923 10:52:14 <gouessej> If we understand better how to emulate the fixed pipeline, I'll be able to fix JogAmp's Ardor3D Continuation before the end of the year
20150923 10:52:45 <xranby_f22> our most advanced example is Jake2
20150923 10:52:54 <xranby_f22> it can emulate fixed function using es2
20150923 10:53:12 <gouessej> it's a bit broken, there is a problem in the key management
20150923 10:53:24 <gouessej> yes I should look at Sven's code
20150923 10:55:44 <gouessej> If I succeed, JogAmp's Ardor3D Continuation will become the only scenegraph API supporting both ES 1, ES 2, OpenGL 1, 2, 3 and 4 :)
20150923 10:56:37 <xranby_f22> https://gist.github.com/xranby/ccbcfcaef341f5bd1ba8 <- JOGLQuadES1 now with color
20150923 10:57:29 <gouessej> big kudos :)
20150923 10:59:28 <xranby_f22> wget http://jogamp.org/deployment/archive/master/gluegen_891-joal_617-jogl_1442-jocl_1087-signed/fat/jogamp-fat.jar
20150923 10:59:28 <xranby_f22> wget https://gist.githubusercontent.com/xranby/ccbcfcaef341f5bd1ba8/raw/6617c1f19c4feac26a28b24756e0983c65b212b3/JOGLQuadES1.java
20150923 10:59:28 <xranby_f22> javac -cp jogamp-fat.jar JOGLQuadES1.java
20150923 10:59:28 <xranby_f22> java -cp jogamp-fat.jar:. JOGLQuadES1
20150923 10:59:49 <xranby_f22> the above will test the code on all unix platforms we support :)
20150923 11:00:05 <xranby_f22> gouessej: KUDOS
20150923 11:00:20 <xranby_f22> i will take a look at the ES2 this evening
20150923 11:00:23 <xranby_f22> work to do
20150923 11:00:32 <gouessej> I have added your example into the wiki
20150923 11:00:43 <xranby_f22> superb!
20150923 11:01:37 <gouessej> I have some questions for you. Do you have a few minutes to answer?
20150923 11:01:48 <xranby_f22> please ask
20150923 11:02:01 <gouessej> ok it concerns the startup time and your attempts
20150923 11:02:21 <gouessej> Does the use of uncompressed files instead of JARs noticeably improve the situation?
20150923 11:02:27 <xranby_f22> not much
20150923 11:02:43 <xranby_f22> using AOT is currently the best way to improve startup time
20150923 11:02:57 <gouessej> ok. Does it start faster without AWT?
20150923 11:03:04 <xranby_f22> my finding is that by using robovm you get very good startup times
20150923 11:03:19 <xranby_f22> for example i get sub 0.5 second statup times using robovm
20150923 11:03:44 <gouessej> Can Robovm use the OpenJDK class library? Impressive short startup time :)
20150923 11:03:57 <xranby_f22> unfortunally robovm can not
20150923 11:04:03 <xranby_f22> use openjdk class librarys
20150923 11:04:38 <gouessej> What about Avian? Does it help?
20150923 11:06:15 <xranby_f22> i do not know yet. avian can be used in two ways combining your app with avion into a minimal executable, this is untested by me or using avian in combinatio with the openjdk class librarys by using avian as a hotspot libjvm.so replacement i have only tested avian by using it as a libjvm.so replacement but that do not reduce startup times
20150923 11:07:00 <gouessej> ok
20150923 11:07:21 <gouessej> I plan to use one of these packages to make my bundle for ARM: https://github.com/libgdx/packr/issues/35#issuecomment-135903158
20150923 11:07:22 <xranby_f22> http://oss.readytalk.com/avian/ <- the best documentation how to do the minimal executable setups
20150923 11:07:52 <zubzub> I'd like to see a jvm/jdk that analyses all your runtime deps, compiles aot and produces a single absolute minimal stand alone executable
20150923 11:08:08 <zubzub> and it must support java 8 :p
20150923 11:08:11 <xranby_f22> zubzub: then do read the avian tutorial above
20150923 11:08:15 <bruce-> and run on iphone
20150923 11:08:20 <xranby_f22> it uses proguard to cut away unrelated parts
20150923 11:08:21 <zubzub> arm is fine
20150923 11:08:26 <xranby_f22> and it work on iphone
20150923 11:08:28 * gouessej (5ee4b442@anon) Quit (Quit: Page closed)
20150923 11:08:32 <zubzub> xranby_f22: java 8 support too?
20150923 11:08:50 <zubzub> jdk/java 8
20150923 11:08:56 <xranby_f22> maybe, avian have added openjdk 8 support when using avian as a libjvm.so replacement
20150923 11:09:12 <xranby_f22> but i do not know if it work when you use avian in the embedded mode
20150923 11:09:33 <xranby_f22> thus it is definately on their road map
20150923 11:10:01 <xranby_f22> please checkout avian and report sucess
20150923 11:10:12 <zubzub> will do
20150923 11:10:33 <zubzub> can it handle jni/jna?
20150923 11:10:47 <xranby_f22> yes
20150923 11:11:01 <xranby_f22> i have used avian to run jogamp and jogamp uses jni
20150923 11:11:09 <xranby_f22> hence it work
20150923 11:11:23 <xranby_f22> at least when i use it as a libjvm.so replacement inside openjdk
20150923 11:12:42 * a11mad11 (~marc-anto@anon) has joined #jogamp
20150923 11:18:52 * a11mad11_ (~marc-anto@anon) has joined #jogamp
20150923 11:21:59 * a11mad11 (~marc-anto@anon) Quit (Ping timeout: 265 seconds)
20150923 11:23:12 * gouessej (5ee4b442@anon) has joined #jogamp
20150923 11:23:54 <gouessej> zubzub: If you still need a Mac to target iOS, I won't give it a try
20150923 11:24:45 <gouessej> xranby_f22: Does the use of a compact JRE without AWT reduce the startup time?
20150923 11:25:34 <zubzub> gouessej: I don't need/want ios
20150923 11:25:36 <gouessej> xranby_f22: I have modified JogAmp's Ardor3D Continuation so that a subset can be used with the "compact2" or "compact1" profile.
20150923 11:25:38 <zubzub> I do have a mac
20150923 11:25:44 <zubzub> (my first & last)
20150923 11:26:06 <zubzub> gouessej: I don't think so
20150923 11:26:46 <zubzub> I just lowers the overal size of your jre
20150923 11:26:49 <gouessej> I hope we'll get Wayland support soon by the way...
20150923 11:27:15 <zubzub> my java wayland compositor can run on compat1 (or 2) profile :p
20150923 11:27:47 <gouessej> you should reassure him: http://forum.jogamp.org/LWUIT-NEWT-JOGL-Enlightenment-Wayland-Tizen-Phone-tp4035335p4035367.html
20150923 11:28:13 <zubzub> what I would really like to see is a some basic widgets for java that can run on wayland
20150923 11:29:01 <gouessej> You can probably use Unlicense, Scene2D LibGDX or Ardor3d-UI
20150923 11:29:32 <zubzub> those all sound overkill
20150923 11:29:41 <zubzub> but I haven't investigated tbh
20150923 11:30:27 <gouessej> Scene2D can be used to build complicated UIs similar to Swing
20150923 11:30:45 <gouessej> Maybe you can use NiftyGUI
20150923 11:31:10 <zubzub> https://github.com/void256/nifty-gui
20150923 11:31:11 <zubzub> yeah
20150923 11:31:41 <gouessej> I'm almost sure that there are a few things to fix to make it work with ES
20150923 11:31:52 <xranby_f22> nifty-gui do work with ES2, me and void256 demonstated it on siggraph 2013
20150923 11:32:10 <xranby_f22> i used my toshiba ac100 es 2 hardware running the demo
20150923 11:32:16 <gouessej> yes but JoglCore in NiftyGui doesn't
20150923 11:32:20 <gouessej> look at the code
20150923 11:32:29 * xranby_f22 has to take a look
20150923 11:33:08 <gouessej> https://github.com/void256/nifty-gui/blob/1.4/nifty-renderer-jogl/src/main/java/de/lessvoid/nifty/renderer/jogl/render/JoglCoreGL.java
20150923 11:33:35 <zubzub> gouessej: for wayland support you don't need es(2)
20150923 11:33:59 <zubzub> just a wayland compatible egl implementation
20150923 11:34:39 <gouessej> xranby_f22: It's not done "my way", they should use the most compatible interface instead of putting GL3 everywhere
20150923 11:36:21 <xranby_f22> aha.. ok only nifty gui 1.3 has suppoer for ES2 "JOGL: add ES2 support to the core batched renderer" https://github.com/void256/nifty-gui/commits/1.3
20150923 11:36:52 <xranby_f22> that was the version i demonstrated
20150923 11:37:50 <xranby_f22> void256 runs three parralel branches and only one branch have ES2 support
20150923 11:42:31 <gouessej> xranby_f22: they just need to replace GL2 and GL3 by GL2ES2 in JoglCore...
20150923 11:43:36 <gouessej> Maybe Brian can do it
20150923 11:43:46 <eclesia> \o/ displaying bbox of a unity mesh
20150923 11:44:12 <gouessej> eclesia: good job :)
20150923 11:44:49 <gouessej> xranby_f22: I can fix it when I have some time after fixing JMonkeyEngine
20150923 11:44:57 * gouessej (5ee4b442@anon) Quit (Quit: Page closed)
20150923 11:48:53 * a11mad11_ (~marc-anto@anon) Quit (Ping timeout: 246 seconds)
20150923 11:55:57 <zubzub> there, replied to the wayland newt therad
20150923 12:07:31 <xranby> gouessej: sgothel: a question: is it possible to specify that you want an GL2ES2 context that uses a specific GLSL version?
20150923 12:08:11 <zubzub> isn't that specified inside glsl itself
20150923 12:08:15 <zubzub> with the version directive?
20150923 12:08:29 <zubzub> and then you need to pray what happens
20150923 12:09:25 <xranby> i have seen projects with a lot of legacy GLSL 1.1-1.3 shaders however if they request an GL2ES2 profile they run into issues they they get a context that require the GL3 core and more modern GLSL causing the shaders to fail to compile
20150923 12:09:30 <zubzub> https://www.opengl.org/wiki/Core_Language_(GLSL)#Version
20150923 12:09:44 <zubzub> To specify which version of GLSL should be used to compile/link a shader, use this directive:
20150923 12:09:47 <zubzub> #version 150
20150923 12:10:07 <zubzub> maybe certain driver implementations are bad
20150923 12:10:15 <xranby> the only workaround is to first try request an GL2 profile and if that fail then request a GLES2 profile... by doing so you avoid getting a context that only accept new shaders
20150923 12:10:18 <zubzub> or maybe certain profiles are not compatible with certain glsl versions?
20150923 12:15:21 <xranby> yes that is the current situation hence best would be to have a way to explicit request a profile that is compatible with the shaders inside your project
20150923 12:16:13 <xranby> becasue there may be three profiles available that is compatible with GL2ES2 however only two of them are compatible with the GLSL used by your projects shaders
20150923 13:06:18 <sgothel> @zuzub: use our graph ui for widgets .. that should be 1st advice anyways
20150923 13:06:49 <zubzub> sgothel: that's pretty much my plan
20150923 13:07:01 <zubzub> but there doesn't seem to be alot of documentation(?)
20150923 13:07:11 <zubzub> how to set it up, how to use it etc
20150923 13:07:27 <sgothel> copy/paste .. its all in flux :)
20150923 13:09:33 <sgothel> GPUUISceneGLListener0A <- this piece of horrible code :)
20150923 13:10:37 <zubzub> gpu ui scene listener oa
20150923 13:10:54 <zubzub> or gupujiscenegulistenerooaa
20150923 13:11:07 <zubzub> 1000 lines
20150923 13:11:20 <zubzub> I already feel bad at 500 :p
20150923 13:11:54 <sgothel> programmatic assembling ui :)
20150923 13:57:56 <xranby> Q: what to we say to people who wait for the 2.3.2 release? A: it is done when i tis done
20150923 13:58:19 <xranby> collaborate to make it happen faster
20150923 13:58:22 <sgothel> yes .. target date end of this month
20150923 13:58:34 <sgothel> well .. it takes its time ..
20150923 13:58:48 <sgothel> so when crucial issues occur like Bug 1219 .. its a showstopper
20150923 13:59:34 <sgothel> only guarantee within October 2015 I would say .. but trying to make it this month
20150923 13:59:54 <sgothel> but yeah .. test test test
20150923 14:00:21 <sgothel> then .. we have quite a few new bug reports due to heavy testing, need to walk them all .. ofc
20150923 14:00:40 <sgothel> i.e. at least to see whether they are bugs or simply user mistakes
20150923 14:00:56 <sgothel> Julien is helping out here big time .. KUDOS
20150923 14:01:06 <sgothel> and you Xerxes as well, KUDOS
20150923 14:01:50 <sgothel> and in the meantime: pls use the RC builds!
20150923 14:02:00 <sgothel> .. almost daily ..
20150923 14:02:22 <sgothel> ok?
20150923 14:02:42 <xranby> ok!
20150923 14:10:50 <zubzub> so xranby I tried avain
20150923 14:10:51 <zubzub> embedded
20150923 14:11:14 <zubzub> (had to make a few small changes to avoid using unimplemented jdk classes)
20150923 14:11:17 <zubzub> now I simply get
20150923 14:11:19 <zubzub> Aborted
20150923 14:11:22 <zubzub> :p
20150923 14:13:49 <xranby> zubzub: hmm , odd that you have to do any changes
20150923 14:14:00 <zubzub> well I use the embedded thing
20150923 14:14:04 <zubzub> without the openjdk classes
20150923 14:14:18 <zubzub> which only provides a small subset of classes
20150923 14:14:26 <zubzub> http://oss.readytalk.com/avian-web/javadoc-1.2.0/index.html
20150923 14:15:03 <xranby> in that case you should not need the openjdk classes
20150923 14:15:17 <xranby> avian cn be used with its own minimal classes
20150923 14:15:21 <zubzub> I know
20150923 14:15:23 <xranby> or android harmony classes
20150923 14:15:25 <zubzub> that's what I do
20150923 14:15:27 <xranby> or openjdk classes
20150923 14:15:34 <zubzub> that's why I had to remove slf4j dep
20150923 14:15:37 <zubzub> which was no biggy
20150923 14:15:44 <zubzub> now it simply
20150923 14:15:45 <zubzub> aborts
20150923 14:15:49 <xranby> i know that avian aborts with such simple error like Abort as soon as it encounter nay error
20150923 14:16:01 <zubzub> will see if I can add more debugging
20150923 14:16:18 <xranby> yes crank up debugging and asserts
20150923 14:16:39 <xranby> zubzub: have you written down the steps you performed?
20150923 14:16:53 <xranby> i will try see if i can get avian in embedded mode running
20150923 14:17:00 <xranby> do hellow world work?
20150923 14:17:04 <xranby> hello world
20150923 14:17:52 <zubzub> yes
20150923 14:18:02 <zubzub> well all I did was
20150923 14:18:04 <zubzub> build avian
20150923 14:18:06 <zubzub> with
20150923 14:18:06 <zubzub> make
20150923 14:18:12 <zubzub> then try to run my jar
20150923 14:18:24 <zubzub> the problem is that the exampel at the bottom isn't very clear
20150923 14:18:33 <zubzub> and the dld links dont work
20150923 14:18:44 <zubzub> zubzub@zubnix ~/Downloads/avian $ build/linux-x86_64/avian -cp build/linux-x86_64/test Hello
20150923 14:18:47 <zubzub> hello, world!
20150923 14:19:15 <xranby> ok great that this worked
20150923 14:19:58 <xranby> did you use the 1.2.0 release or the current git?
20150923 14:20:09 <xranby> my personal experience is that only the git is reliable
20150923 14:20:11 <zubzub> 1.2
20150923 14:20:13 <zubzub> ok
20150923 14:20:17 <zubzub> will try git
20150923 14:20:33 <xranby> the releases only work if you use them in managed environments
20150923 14:20:46 <xranby> talking about personal experience
20150923 14:21:27 <xranby> ok now we have two gits to choose from
20150923 14:21:29 <xranby> https://github.com/bigfatbrowncat/avian-pack
20150923 14:21:58 <xranby> and https://github.com/readytalk/avian
20150923 14:22:17 <xranby> avian-pack: This project is a blend of Avian (http://oss.readytalk.com/avian/) and Android classpath that is (at the moment) far more compatible with the original proprietary JDK Classpath.
20150923 14:22:36 <xranby> Avian all-inclusive. Everything needed to build Avian with (or without) Android classpath.
20150923 14:23:10 <xranby> I will test git clone --recurse-submodules https://github.com/bigfatbrowncat/avian-pack
20150923 14:23:43 <zubzub> ah
20150923 14:23:49 <zubzub> the git version gives me a stacktrace
20150923 14:24:03 <zubzub> caused by: java/lang/NullPointerException at com/sun/jna/Native.getTempDir
20150923 14:24:34 <xranby> what code did you test to get this stacktrace?
20150923 14:24:56 <zubzub> the readytalk avian
20150923 14:24:59 <xranby> your wayland binding?
20150923 14:25:04 <zubzub> my wayland compositor
20150923 14:25:09 <xranby> ok
20150923 14:25:44 <xranby> joel dice if usually very fast to six issues if you can make him observe them
20150923 14:26:28 <xranby> https://github.com/readytalk/avian/issues?q=is%3Aissue+JNA+is%3Aclosed <- interesting all JNA should have been fixed
20150923 14:26:34 <xranby> JNA issues that is
20150923 14:27:15 <xranby> wonder what your JNA does differently
20150923 14:27:27 <zubzub> I use direct mode
20150923 14:27:33 <zubzub> maybe that's the problem?
20150923 14:29:01 <xranby> if you file a bug for avian ans show him how you use it then i think it will be sorted
20150923 14:29:33 <xranby> zubzub: btw take a look at this part https://github.com/ReadyTalk/avian/pull/454
20150923 14:29:40 <xranby> support AOT-compilation of Java 8 lambda expressions
20150923 14:29:59 <xranby> 11 days ago
20150923 14:32:26 <zubzub> huray \o/
20150923 14:32:29 <zubzub> now only I needs to work
20150923 14:32:31 <zubzub> :p
20150923 14:34:21 <zubzub> *it
20150923 14:35:27 <xranby> zubzub: which compile options did you pass to the avian makefile?
20150923 14:35:48 <zubzub> none
20150923 14:36:10 <xranby> aha.. i recall you had to pass quite a fiew
20150923 14:36:11 <xranby> http://labb.zafena.se/?p=630
20150923 14:36:16 <xranby> to make things work
20150923 14:36:37 <xranby> maybe these is unrelated, since I compiled against openjdk
20150923 14:37:03 <zubzub> yeah
20150923 14:37:08 <zubzub> and cross compile it seems?
20150923 14:37:24 <zubzub> or directly on the pi itself?
20150923 14:37:51 <xranby> this was done directly on a raspberry pi
20150923 14:42:21 <xranby> todays discovery: "The semi transparent silver static shield bag's you get sensitive electronics in can be used to generate touch on capacitive touch screens!"
20150923 14:44:16 <xranby> ls
20150923 14:44:25 <xranby> hmm wrong window focus
20150923 14:44:41 <zubzub> :p
20150923 14:44:47 <zubzub> dmesg
20150923 15:00:54 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150923 15:35:37 <xranby> gouessej: on a raspberry pi jogamp startup time of com.jogamp.newt.opengl.GLWindow using the default JDK 8-b132 is 8 seconds using a jogamp-fat.jar and 7 seconds using unpacked classes.
20150923 15:36:20 <xranby> thus you gain one second by pre unpack it
20150923 15:39:34 <xranby> startup time using the most minimal java opengl es 2 binding on the same system an platform dependednt all static JNI binding takes 1.6s
20150923 15:39:55 <xranby> such as the Jgles2pi
20150923 15:40:25 <xranby> https://github.com/chriscamacho/Jgles2pi <- public domain
20150923 15:40:49 <xranby> hence i am thinking about ways to make startup faster
20150923 15:41:10 <xranby> Jgles2pi is tutally unsecure however it is my reference of what is possible
20150923 15:42:38 <xranby> https://www.raspberrypi.org/forums/viewtopic.php?p=819563#p819563 <- it is public domain because chris said so here
20150923 15:44:50 <xranby> my brainstorming go towards if jogl may use its pluggable architecture to do probing if opengl es library methods can be done lazy during runtime instead of probing them all at once during startup
20150923 15:46:18 <xranby> 1.4s using jgles2 with unpacked jars
20150923 15:47:34 * jvanek (jvanek@anon) Quit (Quit: Leaving)
20150923 16:04:51 <xranby> btw would it be evil if jogl used the mesa3d insternals http://dri.freedesktop.org/wiki/GLdispatch/ by fetching the per-thread dispatch pointer using _glapi_get_dispatch() ?
20150923 16:10:18 <xranby> i found that page usefull as well to clarify the misunderstanding that "the opengl api is static" while it is not
20150923 16:11:19 <xranby> http://forum.jogamp.org/Why-does-JOGL-use-Instances-of-GLContext-GL-instead-of-exposing-a-Static-API-td4034144.html
20150923 16:17:43 * eclesia (~husky@anon) has left #jogamp
20150923 16:55:44 * Eclesia (~eclesia@anon) has joined #jogamp
20150923 17:10:27 <sgothel> lazy gl ptr fetching: this would be very bad, since it would require to raise permissions for each action
20150923 17:10:55 * monsieur_max (~maxime@anon) has joined #jogamp
20150923 17:10:57 <sgothel> further more .. you can perform the initSingleton from another thread, while initializing your application
20150923 17:11:22 <sgothel> to enhance proc-address fetching, there is an EGL extension supporting that
20150923 17:11:29 <sgothel> we can earmark this -> bug entry
20150923 17:11:46 <sgothel> i.e. a burst mode .. to fetch many function pointer
20150923 17:12:30 <sgothel> due to the other bindings lazy init, we cannot compare performance properly
20150923 17:17:46 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150923 17:18:10 * monsieur_max (~maxime@anon) has joined #jogamp
20150923 17:28:05 <Eclesia> unity mixes all datas in a single byte array, vertice,normal,uv,... for multiple meshes. Grrrr
20150923 17:28:33 <sgothel> that is good .. i.e. stored interleaved .. allows streaming 1:1
20150923 17:28:50 <sgothel> bad for animation of data though ..
20150923 17:28:53 <Eclesia> 40 values wtf !
20150923 17:29:18 <sgothel> quite some vertex attribute stream .. :)
20150923 17:29:35 <sgothel> *streams*
20150923 17:31:34 <Eclesia> My engine doesn't support such streams :( 1 vbo for each vertices,normals,uvs ...
20150923 17:32:21 <sgothel> looks like the unity engine does more a 'kicking the can' approach .. w/o editing the data
20150923 17:32:40 <sgothel> i.e. just a replay mode .. hence unsuitable format for dynamic tasks
20150923 17:34:48 <Eclesia> hmm I could cheat, use 3 vbo with the same buffer but with a different offset and a large stride.
20150923 17:35:02 <Eclesia> would that have a bad impact on performances ?
20150923 17:35:07 <sgothel> while saturating bandwidth ..
20150923 17:35:14 <sgothel> sure .. bad
20150923 17:35:33 <sgothel> well .. all depends on the target platform
20150923 17:35:45 <Eclesia> gtx970 :D
20150923 17:35:46 <sgothel> hence .. I assume the format can be configured .. now?
20150923 17:35:52 <sgothel> *no* ?
20150923 17:36:04 <sgothel> i.e. for embedded .. etc
20150923 17:36:34 <Eclesia> I guess unity sdk has some option to configure how the data are structured
20150923 17:36:47 <Eclesia> but I don't use unity ^^ just the model files hehehe
20150923 17:40:59 <Eclesia> there is big buzz around unity3d, we start to find a lot of models in this format this year
20150923 17:41:27 <Eclesia> likely to become a new 'de-facto'
20150923 17:41:31 <sgothel> hmm .. dunno, don't care - not open, not free
20150923 17:42:20 <sgothel> also .. it uses mesh data, i.e. preset quality and stuff .. but well, most do
20150923 17:42:59 <sgothel> (i.e. useless for UI and CAD)
20150923 17:43:51 <Eclesia> I don't care, as far as I can read it and convert it to my internal model, and then output it in an format wanted
20150923 17:43:56 <Eclesia> any*
20150923 18:20:15 <Eclesia> okay, I think I get it, located stream, offsets,size,masks
20150923 18:20:45 <Eclesia> looks like the format only handle triangles, no point or lines.
20150923 18:27:56 <Eclesia> what kind of information for a mesh has a size of 4 for each vertice ?
20150923 18:28:54 <Eclesia> could be a quaternion, but for what purpose ?
20150923 18:29:58 <Eclesia> maybe bone weights
20150923 19:49:29 <Eclesia> ho cool, unitypacakge == plain old tar archive
20150923 20:18:05 * a11mad11 (~marc-anto@anon) has joined #jogamp
20150923 20:41:05 * Eclesia (~eclesia@anon) Quit (Quit: Leaving.)
20150923 20:52:09 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150924 00:52:59 * xranby_f22 (~familjen@anon) Quit (Ping timeout: 256 seconds)
20150924 00:54:10 * xranby_f22 (~familjen@anon) has joined #jogamp
20150924 02:08:03 * a11mad11 (~marc-anto@anon) Quit (Ping timeout: 240 seconds)
20150924 05:05:12 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20150924050512.html