#jogamp @ irc.freenode.net - 20141011 05:05:55 (UTC)


20141011 05:05:55 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20141010050554.html
20141011 05:05:55 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20141011050555.html
20141011 07:19:53 * monsieur_max (~maxime@anon) has joined #jogamp
20141011 09:11:52 * gouessej (5279cbab@anon) has joined #jogamp
20141011 09:12:09 <gouessej> Hi
20141011 09:12:20 <sgothel> Hi Julien
20141011 09:12:23 <sgothel> Good morning
20141011 09:12:25 <gouessej> sgothel: "Julien: "and keep cool, we're here to help" <- now _he_ made my day, he - the most emotional of us - I love it :)" You're right
20141011 09:12:34 <sgothel> yeah .. that was great
20141011 09:12:41 <sgothel> 1st time you calm me down :)
20141011 09:13:08 <gouessej> sgothel: Sometimes I just get angry in myself when I see that some developers don't even read the examples and the documentations that we show them
20141011 09:13:18 <sgothel> hehe .. same here
20141011 09:13:53 <sgothel> and if its 'late' .. sometime .. I can't hold it back :)
20141011 09:13:54 <gouessej> sgothel: we spend some time in helping them, the minimum politeness would consist in reading the stuff we point out to them instead of demanding us to "feed" them
20141011 09:14:14 <sgothel> 'I'm wishing for a star .. '
20141011 09:14:45 <sgothel> so how does 2.2.4 work for you ?
20141011 09:15:15 <gouessej> I haven't had time to test it at work, I left a few minutes or hours before you put it online
20141011 09:15:27 <gouessej> no regression so far in my personal projects
20141011 09:15:34 <sgothel> great
20141011 09:15:43 <gouessej> but we have another trouble...
20141011 09:16:20 <gouessej> Brian is working on NiftyGUI and he complains a bit about a difference of performance between JOGL and its main competitor under GNU Linux
20141011 09:16:51 <gouessej> I looked at his code, I found nothing particularly offending
20141011 09:16:56 <sgothel> ah .. you mean Void's (Jens) NiftyGUI ?
20141011 09:17:09 <sgothel> dunno 'Brian'
20141011 09:17:23 <sgothel> however .. so perf. regressions are claimed ?
20141011 09:18:07 <gouessej> https://github.com/bgroenks96/coregl
20141011 09:18:46 <sgothel> and ?
20141011 09:19:27 <gouessej> I'm looking at his figures...
20141011 09:19:40 <sgothel> I mean .. maybe this intermediate layer is not optimal ? I assume using that w/ both impl. gives diff perf ..
20141011 09:20:36 <sgothel> well .. IMHO this is a bit ridiculous
20141011 09:20:43 <sgothel> public int GL_ONE_MINUS_DST_ALPHA() { return GL.GL_ONE_MINUS_DST_ALPHA; }
20141011 09:20:43 <gouessej> both backends are extremely similar despite the difference of OpenGL bindings, JOGL is a bit faster most of the time than its competitor under Windows but not under GNU Linux
20141011 09:20:44 <sgothel> etc
20141011 09:21:06 <sgothel> we would need native comparison here ..
20141011 09:21:21 <sgothel> i.e. one demo written using LWJGL .. and one w/ NEWT/JOGL
20141011 09:21:29 <sgothel> then compare ..
20141011 09:21:42 <sgothel> plus .. diff features etc etc
20141011 09:21:59 <sgothel> bringup time, runtime .. many things to consider for sure
20141011 09:22:10 <gouessej> I have just sent you the figures
20141011 09:22:27 <gouessej> he measures the amount of time to render
20141011 09:22:31 <sgothel> now I see them .. looks horrible :)
20141011 09:22:51 <sgothel> but I only trust stats I have fiddled with :)
20141011 09:23:22 <sgothel> VM comparison .. etc etc
20141011 09:23:29 <sgothel> and I dunno the tests themselves
20141011 09:24:08 <sgothel> 1st Q would be: are those tests using core profiles ?
20141011 09:24:17 <sgothel> or tons of FFP calls ? :)
20141011 09:24:41 <gouessej> yes, core profile
20141011 09:25:10 * Eclesia (~eclesia@anon) has joined #jogamp
20141011 09:25:14 <sgothel> sure it would be nice to tackle such comparison .. but need all the data ofc
20141011 09:25:21 <Eclesia> good morning
20141011 09:25:24 <sgothel> tests source .. and so on
20141011 09:25:28 <sgothel> goody
20141011 09:25:59 <sgothel> is there a perf test bed ?
20141011 09:26:12 <sgothel> i.e. git pull .. and run ?
20141011 09:26:33 <gouessej> I'm looking for the test source but I don't find it
20141011 09:27:14 <sgothel> well, I have a perf test on my mind as well (updating an old one from jogl-demos) .. but this tests the GL performance, not our API
20141011 09:27:51 <sgothel> since I cannot see where there is any difference possible, but surely I am open for elucidation :)
20141011 09:28:21 <gouessej> Please ask Brian to show you where the tests are
20141011 09:29:06 <gouessej> Personally, I have never found noticeable differences of performance between JOGL and its main competitors, especially since I use NEWT
20141011 09:29:09 <sgothel> dunno whether I have resources or priority
20141011 09:29:44 <gouessej> We can let him investigate a bit but I would like to know where those tests are, I'll ask him
20141011 09:30:07 <sgothel> perf should be measure: setup/bringup time, and then indeed avr frame _PLUS_ fps deviation (i.e. GC load etc)
20141011 09:30:27 <sgothel> the latter to test whether JOGL corrupts fluent animation for example
20141011 09:30:49 <gouessej> I have noticed that JogAmp's Ardor3D Continuation calls glGetError at least once per draw, is it a bad idea?
20141011 09:31:04 <sgothel> if he like to cooperate .. and all is open source, I can help and assist - sure. but only if I am asked, me not driving this thing.
20141011 09:31:18 <sgothel> yes sure
20141011 09:31:41 <sgothel> https://jogamp.org/bugzilla/show_bug.cgi?id=1066 <- I already removed a few calls
20141011 09:32:00 <gouessej> Renanse calls this method just one time before swapBuffers() (we don't use auto swap buffer mode)
20141011 09:33:01 <sgothel> http://jogamp.org/git/?p=jogl.git;a=commit;h=7749f0567fb592ee4b3f698bd2424a5687b02f47
20141011 09:33:06 <sgothel> http://jogamp.org/git/?p=jogl.git;a=commit;h=4d736527930511cabf2c6604aa831544da917198
20141011 09:33:16 <sgothel> see comments on other GetError calls ..
20141011 09:33:27 <sgothel> (in bug report)
20141011 09:33:42 <sgothel> surely a GetError kills performance, and should only be there if must
20141011 09:33:46 <sgothel> even .. swapBuffers
20141011 09:34:11 <sgothel> so my idea here is: for resource creation, GetError is OK, since it shall not happen every frame
20141011 09:34:22 <sgothel> but GetError for swap - is killing
20141011 09:34:44 <sgothel> we have DebugGL to test application
20141011 09:35:08 <gouessej> Ok. We already provide the option of using DebugGL and TraceGL
20141011 09:35:28 <sgothel> https://jogamp.org/bugzilla/show_bug.cgi?id=1066#c2
20141011 09:35:30 <sgothel> https://jogamp.org/bugzilla/show_bug.cgi?id=1066#c3
20141011 09:35:52 <sgothel> i.e. validate .. whether any of the remaining GetError are called w/ high-frequency ..
20141011 09:36:14 <sgothel> (all that is in 2.3.* master .. and only goes in there)
20141011 09:37:19 <gouessej> Ok I will have to give it a try as soon as possible but is there a chance that you put it into 2.2.5? Does it impact onscreen rendering?
20141011 09:37:34 <sgothel> again, those perf stats are useless w/o proper environment spec, FOSS framework etc
20141011 09:37:57 <sgothel> I don't want to put anything which can cause instability to 2.2.*
20141011 09:38:19 <sgothel> 2.2.* _was_ for bugfix only, and should now be done :)
20141011 09:38:28 <sgothel> so .. 2.3.0 is our next thing
20141011 09:38:29 <gouessej> Ok
20141011 09:38:48 <sgothel> currently doing https://jogamp.org/bugzilla/show_bug.cgi?id=1068, which will speed up offscree-drawables based on FBO
20141011 09:39:16 <gouessej> Yes, no drawable when using makeCurrent :)
20141011 09:39:20 <sgothel> at least .. its creation .. and will ease resources (no more dummy window if supported)
20141011 09:39:37 <sgothel> but .. not all driver comply w/ spec
20141011 09:40:19 <sgothel> then making EGL working w/ desktop GL profiles .. then Wayland .. etc etc
20141011 09:40:41 <sgothel> and on the side .. OculusVR .. GraphUI .. so many things :)
20141011 09:41:48 <gouessej> I need to put OpenJDK into my packages, create MSI packages too and then maybe I'll have to work on NEWT Input API... finally
20141011 09:42:11 <sgothel> first thing w/ such perf testbed .. using a perf analyzer and compare
20141011 09:43:06 <gouessej> which perf analyzer?
20141011 09:43:30 <sgothel> jvisualvm ?
20141011 09:43:45 <sgothel> or instrumented code in perf testbed
20141011 09:43:45 <gouessej> of course
20141011 09:44:06 <gouessej> I have another question for you about the bug 419
20141011 09:44:35 <sgothel> go on
20141011 09:44:49 <Eclesia> are you talking about the bug that profiling crash with jogl ?
20141011 09:44:59 <gouessej> I will probably not provide RPM and DEB packages for Linux distros as your explanations were correct
20141011 09:45:21 <gouessej> it's too difficult to comply with all distro policies
20141011 09:45:39 <gouessej> Some of them use patched versions of native packagers :s
20141011 09:45:53 <sgothel> yeah, it is a full time job for sure
20141011 09:46:08 <sgothel> @Eclesia: ? No I guess
20141011 09:46:22 <gouessej> I'll just do my best to help developers to make their own packages with or without a JRE/JDK
20141011 09:46:32 <gouessej> with some precise instructions
20141011 09:46:45 <gouessej> and I have to deal with MSI too (Windows)
20141011 09:47:01 <sgothel> I mean .. we already went over the line .. a lot, supporting native JAR handling .. fat JARs .. etc
20141011 09:47:23 <gouessej> Yes but Java Webstart is mostly unusable
20141011 09:47:27 <sgothel> maybe the only missing thing here .. is combining it with one other tool, I guess you mentioned that
20141011 09:47:42 <sgothel> not talking webstart here ..
20141011 09:47:52 <sgothel> i.e. you can already do one fat jar w/ our stuff
20141011 09:48:02 <sgothel> not .. just make it executable
20141011 09:48:05 <sgothel> *now*
20141011 09:48:16 <gouessej> Yes but Firefox under Windows wraps JARs into ZIPs :(
20141011 09:48:24 <sgothel> cannot be too hard
20141011 09:48:35 <gouessej> Tons of archivers open the JARs instead of letting the OS running the programs
20141011 09:48:35 <sgothel> then make an executable zip ..
20141011 09:48:47 <gouessej> LibGDX PackR does that
20141011 09:49:02 <sgothel> you .. or others mentioned a fat jar tool ..
20141011 09:49:04 <sgothel> open source
20141011 09:49:07 <gouessej> Yes
20141011 09:49:34 <sgothel> so maybe tweak that to produce something like executable zip, which contains our fat jar and a java invocator .. done
20141011 09:49:43 <gouessej> I just want something that allows to run my application with one click or two
20141011 09:49:50 <sgothel> yup
20141011 09:50:25 <sgothel> isn't there a cross platform shell script based archive package format ?
20141011 09:50:29 <gouessej> and when I use "native" packages, no web browser will get in my way
20141011 09:50:40 <sgothel> well - if so - I surely would disable it ofc (security) .. hehe
20141011 09:50:48 <gouessej> lol
20141011 09:51:28 <sgothel> yeah, so it's clear .. a startup native stub extracting the zip, and launching java w/ the zip contained fat jar .. done
20141011 09:51:42 <sgothel> I sort of explained that in JiGong Runtime
20141011 09:51:58 <sgothel> but surely not time nor interest to implement that now
20141011 09:52:02 <gouessej> Why not using already known native packages?
20141011 09:52:13 <sgothel> go for it
20141011 09:52:31 <sgothel> if you need it .. I won't hold you back
20141011 09:52:38 <sgothel> just brainstorming here
20141011 09:52:49 <gouessej> Do you think it is a bad idea? I would like to know if I'm missing something obvious
20141011 09:53:21 <sgothel> I don't really know .. security vs convenience, a bitch
20141011 09:53:46 <gouessej> Native packages can be signed with PGP too as far as I know
20141011 09:53:47 <sgothel> just can tell you, I would not DL such things .. but Windows users do that all the time
20141011 09:53:54 <sgothel> now we are talking
20141011 09:54:17 <gouessej> You wouldn't download a signed DEB package, would you?
20141011 09:54:25 <sgothel> hehe :)
20141011 09:54:32 <sgothel> yeah so that works
20141011 09:54:36 <sgothel> ring of trust
20141011 09:54:39 <sgothel> and so forth
20141011 09:54:47 <sgothel> a launchable DEB package ?
20141011 09:54:52 <gouessej> yes
20141011 09:54:57 <sgothel> sounds good
20141011 09:55:26 <sgothel> and what about Windows, OSX .. - must be possible, tedious work
20141011 09:55:28 <gouessej> the application can be launched in command line or by using a menu in its group
20141011 09:55:36 <gouessej> OSX, done
20141011 09:55:40 <gouessej> :)
20141011 09:56:14 <sgothel> go for it .. curious, pls update
20141011 09:56:34 <sgothel> (if you want it ofc .. for your self, most importantly)
20141011 09:56:45 <gouessej> I need that stuff
20141011 09:56:55 <gouessej> for my game
20141011 09:57:11 <sgothel> then .. maybe something like chroot environment .. like docker :)
20141011 09:57:21 <gouessej> and we won't advise to use Java Webstart
20141011 09:57:22 <gouessej> :s
20141011 09:57:45 <sgothel> since we use java, we don't need the big fat docker stuff
20141011 09:58:01 <sgothel> but the chroot env thingy .. nice
20141011 09:58:34 <sgothel> at least not spamming the host system
20141011 09:59:02 <gouessej> The hardest thing for me consists in making MSI files as this is a bit poorly documented
20141011 09:59:18 <sgothel> gnupg .. for those packs - I am afraid another use-case ordinary user won't understand
20141011 09:59:30 <sgothel> what is MSI ?
20141011 09:59:37 <sgothel> microsoft ?
20141011 09:59:40 <gouessej> Microsoft Software Installer
20141011 10:00:01 <sgothel> just an executable .. putting all in temp folder, own namespace .. exec .. done
20141011 10:00:09 <gouessej> but there is an open source API to manipulate it
20141011 10:00:18 <gouessej> Apache POI
20141011 10:00:37 <sgothel> I see .. it's not something you just do in a week :)
20141011 10:00:58 <gouessej> but DEBs and RPMs will be ready before the end of the month
20141011 10:01:27 <sgothel> download zip, extract zip, and click on .bat or .sh file (matching OS) ain't enough ?
20141011 10:01:29 <gouessej> and it is already possible to make APP (OSX) files without a Mac :)
20141011 10:01:46 <gouessej> It's already doable with PackR
20141011 10:01:53 <gouessej> zip with scripts
20141011 10:01:58 <sgothel> but ?
20141011 10:02:18 <gouessej> It's too much for lots of end users unfortunately :(
20141011 10:02:42 <sgothel> horizontal vs vertical market
20141011 10:03:13 <gouessej> The zipped app file gets automatically extracted and recognized under Mac, it's good :)
20141011 10:03:44 <gouessej> One click to download (or two if unsigned), another one to run
20141011 10:04:13 <sgothel> good
20141011 10:04:18 <gouessej> It gets harder when Gatekeeper is enabled :(
20141011 10:04:43 <gouessej> I put some instructions here: http://tuer.sourceforge.net/play.php?lang=en
20141011 10:05:58 <sgothel> are the 'days of wonder' guys around ?
20141011 10:06:02 <sgothel> ask them maybe .. ?
20141011 10:06:12 <gouessej> I'll let you know my findings. Native packaging will probably be the advised way of packaging applications using JogAmp's Ardor3D Continuation
20141011 10:06:18 <sgothel> maybe Harvey knows a bit as well ?
20141011 10:06:20 <gouessej> Days of wonder? The game?
20141011 10:06:25 <sgothel> the company
20141011 10:06:43 <sgothel> they did the Android/Desktop rail game thingy
20141011 10:06:58 <sgothel> you met one (Brice) at Fosdem AFAIK
20141011 10:07:01 <gouessej> yes
20141011 10:07:27 <sgothel> I mean .. this problem must have been solved a thousand times already :)
20141011 10:07:58 <sgothel> launching tuer
20141011 10:08:00 <gouessej> Actually, only a very few Java games are packaged as DEB and RPM
20141011 10:08:33 <sgothel> slow server 'eh ?
20141011 10:08:55 <gouessej> Yes :s
20141011 10:09:23 <gouessej> FreeCol is packaged as a DEB but I'm sure it's painful to do :s
20141011 10:09:44 <sgothel> my security settings blocked .. blabla .. I love it :-/
20141011 10:10:12 <gouessej> Lol
20141011 10:10:18 <gouessej> Are you actually under Mac?
20141011 10:10:37 <gouessej> If so, you have to follow the instructions on the download page
20141011 10:10:39 <sgothel> linux .. our mac test machine next to me
20141011 10:10:47 <sgothel> tried jnlp
20141011 10:11:02 <gouessej> You're using the proprietary c..p... JRE
20141011 10:11:04 <sgothel> http://tuer.sourceforge.net/very_experimental/tuer.jnlp .. w/ Oracle's java8
20141011 10:11:07 <sgothel> yup
20141011 10:11:18 <gouessej> You see it's unusable
20141011 10:11:41 <sgothel> now I have some lunch .. and a nap
20141011 10:11:55 <sgothel> great chatting Julien .. have a great day & laters
20141011 10:12:01 <gouessej> See you later
20141011 10:12:08 <gouessej> Bye
20141011 10:12:12 * gouessej (5279cbab@anon) Quit (Quit: Page closed)
20141011 10:12:12 <sgothel> bye .. AFK
20141011 10:35:47 * bbbruce (~bx@anon) Quit (Ping timeout: 258 seconds)
20141011 10:50:43 * hija (~hija@anon) has joined #jogamp
20141011 11:22:48 * hija_ (~hija@anon) has joined #jogamp
20141011 11:25:00 * hija (~hija@anon) Quit (Ping timeout: 260 seconds)
20141011 11:25:00 * hija_ is now known as hija
20141011 11:31:16 * hija_ (~hija@anon) has joined #jogamp
20141011 11:31:22 * hija (~hija@anon) Quit (Ping timeout: 244 seconds)
20141011 11:31:22 * hija_ is now known as hija
20141011 12:05:57 * hija_ (~hija@anon) has joined #jogamp
20141011 12:07:46 * hija (~hija@anon) Quit (Ping timeout: 255 seconds)
20141011 12:07:46 * hija_ is now known as hija
20141011 14:05:25 * hija (~hija@anon) Quit (Quit: hija)
20141011 14:50:17 * Eclesia (~eclesia@anon) Quit (Quit: Leaving.)
20141011 15:05:49 * hija (~hija@anon) has joined #jogamp
20141011 15:27:57 * hija (~hija@anon) Quit (Quit: hija)
20141011 15:42:58 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20141011 15:45:30 * hija (~hija@anon) has joined #jogamp
20141011 16:42:52 * hija (~hija@anon) Quit (Quit: hija)
20141011 17:09:08 * bbbruce (~bx@anon) has joined #jogamp
20141011 19:35:04 * void256 (~void@anon) has joined #jogamp
20141011 19:52:26 * hija (~hija@anon) has joined #jogamp
20141011 21:13:44 * [Mike] (~Mike]@anon) has joined #jogamp
20141011 21:14:46 * [Mike] (~Mike]@anon) Quit (Client Quit)
20141011 21:15:10 * [Mike] (~Mike]@anon) has joined #jogamp
20141011 22:05:01 * void256 (~void@anon) Quit (Remote host closed the connection)
20141011 22:15:46 * zzuegg (~zzuegg@anon) Quit (Quit: Nettalk6 - www.ntalk.de)
20141012 00:41:44 * xranby (~xranby@anon) Quit (Ping timeout: 272 seconds)
20141012 00:55:34 * xranby (~xranby@anon) has joined #jogamp
20141012 01:01:22 * xranby (~xranby@anon) Quit (Ping timeout: 250 seconds)
20141012 01:17:09 * xranby (~xranby@anon) has joined #jogamp
20141012 04:02:56 * xranby (~xranby@anon) Quit (Ping timeout: 250 seconds)
20141012 04:21:56 * xranby (~xranby@anon) has joined #jogamp
20141012 04:27:39 * xranby (~xranby@anon) Quit (Ping timeout: 244 seconds)
20141012 04:43:56 * xranby (~xranby@anon) has joined #jogamp
20141012 05:05:55 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20141012050555.html