#jogamp @ irc.freenode.net - 20130129 05:05:28 (UTC)
20130129 05:05:28 -CatOut- Previous @ http://jogamp.org/log/irc/jogamp_20130128050615.html
20130129 05:05:28 -CatOut- This channel is logged @ http://jogamp.org/log/irc/jogamp_20130129050528.html
20130129 05:28:44 * hharrison (~chatzilla@anon) has joined #jogamp
20130129 05:28:57 <sgothel> good morning
20130129 05:29:08 <hharrison> good evening ;-)
20130129 05:37:13 <hharrison> Getting ready for FOSDEM?
20130129 05:39:23 <sgothel> right .. well, more like business as usual, i.e. enhancing and fixing - showing what we have ..
20130129 05:39:43 <sgothel> even though .. JOAL openal-soft should be in by then
20130129 05:40:57 <hharrison> Got some very encouraging test results from Emmanual with August's latest fixes
20130129 05:41:09 <sgothel> great
20130129 05:41:30 <hharrison> And even went a little nuts last night and got the warnings in a default build of j3dcore under 1500
20130129 05:41:38 <sgothel> I also ran unit tests on my OSX w/ Java7 .. well, more errors - but sort of running ok
20130129 05:41:51 <sgothel> wow
20130129 05:42:19 <hharrison> That's down for 5-6k when I took it over
20130129 05:43:04 <sgothel> kudos - the tedious tasks and discipline keeps a vision alive
20130129 05:43:37 <hharrison> I'm actually hopeful to have a small pile of code up for grabs later in February
20130129 05:44:02 <hharrison> The renderer we built using JOGL2 @work, trying to get the go-ahead to offer you the math lib we developed
20130129 05:44:22 <sgothel> sweet
20130129 05:45:20 <hharrison> something that would fit well in com.jogamp.opengl.math
20130129 05:46:34 <sgothel> sounds great .. and I know it's coming from a good source, well tested - as you always do, remembering our last reviews etc .. nice nice
20130129 05:46:34 <hharrison> andmath.geom
20130129 05:46:34 <hharrison> already got the go ahead for GPL3, trying to get release under apache or bsd for that small bit
20130129 05:46:50 <hharrison> Quick highlights:
20130129 05:47:05 <hharrison> - no temp object allocation in all functions
20130129 05:47:16 <hharrison> - 100% unit test coverage
20130129 05:47:36 <sgothel> float ?
20130129 05:47:42 <hharrison> And optimized just enough to still be pretty obvious code (didn't go crazy)
20130129 05:48:08 <hharrison> double...but one s/double/float/ away from a parallel implementation
20130129 05:48:23 <hharrison> (had that in mind from the beginning)
20130129 05:48:55 <sgothel> yeah .. float is still the choice for embedded .. could have a Float/Double dual class we/ that manual build time re-fit
20130129 05:49:23 <sgothel> matrix .. vector .. what kind of math ?
20130129 05:49:33 <hharrison> I went the Vec2d/Vec2f kind of approach
20130129 05:50:18 <sgothel> so OO style, but w/o those temp objects as you say .. good
20130129 05:50:26 <hharrison> I'll get it out tomorrow under GPL3 and you can have a look to see if it's anything you'd actually care about
20130129 05:50:44 <sgothel> how about data presentation in i.e. float arrays to be suitable for GPU ?
20130129 05:51:09 <hharrison> We're still experimenting a bit
20130129 05:51:45 <hharrison> public float[] getFloats() vs public void getFloats(float[] out)
20130129 05:51:46 <sgothel> that was my take on the Mali math package .. too much allocation, and discrete components (float x, ..) -> not suitable
20130129 05:52:41 <hharrison> To be fair, on hotspot, as long as it stays in-scope of the function and is a small object, stack allocation is getting better
20130129 05:52:44 <sgothel> yup - looking fwd to see your code & design
20130129 05:53:07 <sgothel> problem: we don't have hotspot available all the time, you know it
20130129 05:53:24 <hharrison> Yeah, the disclaimer was there for a reason
20130129 05:53:27 <hharrison> ;-)
20130129 05:54:34 <sgothel> great new, Harvey .. just drop the URL later here, then I like to discuss
20130129 05:55:23 <hharrison> It'll be here, but I'll clone it to my regular github right after
20130129 05:55:25 <hharrison> https://github.com/ausencosimulation
20130129 05:56:13 <hharrison> Afraid I've been a bit lax about pushing stuff out, just going to dump over the wall again andf then try and do better going forward
20130129 05:57:48 <sgothel> well, here it's not like many people read the code I push - but pushing more often does keep my discipline a bit higher and also allows better bisecting for regressions and make more atomic commits
20130129 05:58:09 <hharrison> ahem...I read what you push out
20130129 05:58:24 <sgothel> oh oh - thank you :)
20130129 05:58:39 <sgothel> so it's unix style, no feedback == good feedback :)
20130129 05:59:16 <hharrison> I read, and run findbugs every week or so, then do a diff of the findbugs stuff from my last build
20130129 05:59:51 <sgothel> nice - Q&A - so pls don't be shy to nag & complain!
20130129 06:00:11 <hharrison> you just haven't made any major mistakes yet ;-)
20130129 06:00:47 <hharrison> but your testsuite gives you a pretty good safety net from what I've seen
20130129 06:01:30 <sgothel> Real anecdote: Grandfather's apprentice was complaining about him with Grandma why he never tells him he performed good work. She says, well - if he doesn't criticize your work, that means you are doing a great job. :)
20130129 06:02:09 <sgothel> yup .. our CI saves sanity .. and those days when coding w/ sickness / cold etc :)
20130129 06:03:59 <sgothel> When I was asking former colleagues to do help w/ background check for Canadian company and US comp. later, I told them the attribute 'good' more equals 'not so good', hence you need to use a higher one, i.e. 'great' etc etc - German understatement is not understood properly over there :)
20130129 06:04:43 <sgothel> We have the same for negative classification - like 'he is social' -> 'alcoholic' :)
20130129 06:05:22 <sgothel> or 'he tries hard' -> 'he cannot succeed' .. etc
20130129 06:06:46 <sgothel> well .. now I have to finish the GLAutoDrawable 'survive context' feature .. not as easy as it seemed
20130129 07:09:38 * EricB1 (~Adium@anon) has joined #jogamp
20130129 07:35:00 <masterzen> chiming in the discussion: even on hotspot with young generation GC, allocating objects on the stack on an inner loop kind of suck. It's waaaay better than on dalvik, but we still can see it :)
20130129 07:35:58 <sgothel> good to know - and ofc we should follow write-once if possible
20130129 07:38:48 <masterzen> yep
20130129 07:39:15 <masterzen> but frankly, dalvik reminded me the state of the jvm back 10 years ago
20130129 07:40:40 <hharrison> Yeah, I'm more descibing the single 'temp' object you needed in a helper
20130129 07:41:21 <hharrison> By the time you start allocating inside a loop, not much will help, the JVM doesn't seem to do stack-reuse for the temps yet
20130129 07:52:11 * hharrison (~chatzilla@anon) has left #jogamp
20130129 15:02:15 * EricB1 (~Adium@anon) Quit (Quit: Leaving.)
20130129 16:03:10 * EricB1 (~Adium@anon) has joined #jogamp
20130129 17:59:13 * hharrison (~chatzilla@anon) has joined #jogamp
20130129 18:03:44 * EricB1 (~Adium@anon) Quit (Quit: Leaving.)
20130129 19:17:59 * EricB1 (~Adium@anon) has joined #jogamp
20130129 20:50:32 * EricB1 (~Adium@anon) Quit (Quit: Leaving.)
20130129 22:39:33 * EricB1 (~Adium@anon) has joined #jogamp
20130129 23:06:28 * EricB1 (~Adium@anon) Quit (Quit: Leaving.)
20130130 01:36:21 <hharrison> As promised, code pushed out to github
20130130 01:36:22 <hharrison> https://github.com/AusencoSimulation/JaamSim
20130130 01:44:24 <hharrison> Sven, the math-bits I was talking about are in com.jaamsim.math Mat4d and Vec[234]d
20130130 01:44:55 <hharrison> I'm writing unit tests for the rest of the code in there and cleaning it up a bit, but have a look and see if there is anything you might want
20130130 03:32:33 * hharrison (~chatzilla@anon) Quit (Quit: ChatZilla 0.9.89 [Firefox 19.0/20130116072953])
20130130 05:05:42 -CatOut- Continue @ http://jogamp.org/log/irc/jogamp_20130130050542.html