#jogamp @ irc.freenode.net - 20150818 05:05:30 (UTC)


20150818 05:05:30 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20150817050530.html
20150818 05:05:30 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20150818050530.html
20150818 05:09:50 * badshah400 (~badshah40@anon) has joined #jogamp
20150818 06:27:10 * monsieur_max (~maxime@anon) has joined #jogamp
20150818 06:57:21 * elect (~elect@anon) has joined #jogamp
20150818 06:57:24 <elect> hi
20150818 07:34:55 * badshah400 (~badshah40@anon) Quit (Ping timeout: 250 seconds)
20150818 07:40:36 * eclesia (~husky@anon) has joined #jogamp
20150818 07:40:49 <eclesia> good morning
20150818 08:14:56 <xranby_> good morning
20150818 08:52:01 <sgothel> goody
20150818 08:53:13 <zubzub> aloha
20150818 09:19:26 * gouessej (5ee4b442@anon) has joined #jogamp
20150818 09:20:42 <gouessej> Hi
20150818 09:27:56 <sgothel> hi
20150818 09:31:27 * gouessej (5ee4b442@anon) Quit (Ping timeout: 246 seconds)
20150818 09:32:29 <xranby_> goody! we reduced startup time with ~half a second on the raspberry pi by reducing places to look for natives etc
20150818 09:32:48 <sgothel> wow
20150818 09:34:30 <sgothel> maybe the fat-lookup should be placed _after_ the single-slim one? hmm ..
20150818 09:34:41 <xranby_> i am using time to run com.jogamp.newt.opengl.GLWindow as my benchmark
20150818 09:34:42 <sgothel> since single-slim is the official way ..
20150818 09:35:44 * badshah400 (~badshah40@anon) has joined #jogamp
20150818 09:35:50 <sgothel> JNILibLoaderBase line 220
20150818 09:36:28 <sgothel> "final URL nativeLibraryURI = cl.getResource(nativeLibraryPath);" <- even this could take some time .. hmm
20150818 09:36:46 <sgothel> I will reverse order ..
20150818 09:37:53 <xranby_> sgothel: do we have a debug flag that can help us with the timestamping of how much time is spent in the different parts of the loader
20150818 09:37:57 <xranby_> i think not
20150818 09:38:18 <sgothel> not yet .. good one .. adding something here .. hmm
20150818 09:40:18 <xranby_> when i run GLWindow on a low en device i see pause -> wall of text ->pause -> wall of text -> window open -> wall of text -> done
20150818 09:40:50 <xranby_> thus there is obviously these places where the pause is that we are interested in profiling
20150818 09:45:26 <zubzub> you can use jprofiler
20150818 09:45:35 <zubzub> well for 30 days at least :p
20150818 09:45:45 <zubzub> it will tell you how long each method takes
20150818 09:46:04 <zubzub> maybe you can get a free open source license (if they do that?)
20150818 09:46:36 <sgothel> hardcoding counters .. to present accumulated duration in JNILibLoaderBase
20150818 09:47:39 <xranby_> time java -cp gluegen-rt.jar:jogl-all.jar com.jogamp.newt.opengl.GLWindow
20150818 10:08:44 <xranby_> pro tip: if you are running on a raspberyr pi and see a "rainbow square" in the top right corner of the screen. then this is the firmware telling you that it was approacing a low voltage brownout, the firmware tried to save you by reducing clockrate but you may have memory corruption and should switch powersupply
20150818 10:09:56 <xranby_> also thick usb cables are better at delivering current than flexible slim usb cables
20150818 10:14:06 <sgothel> pushed GlueGen changes (jar order + perf-counter)
20150818 10:14:42 <sgothel> on a desktop .. around 3ms per native-jar
20150818 10:15:27 <sgothel> JNILibLoaderBase: addNativeJarLibsImpl.X: gluegen-rt.jar / gluegen-rt-natives-linux-amd64.jar -> ok: true; duration: now 7 ms, total 7 ms (count 1, avrg 7.000 ms)
20150818 10:15:27 <sgothel> JNILibLoaderBase: addNativeJarLibsImpl.X: jogl-all-noawt.jar / jogl-all-noawt-natives-linux-amd64.jar -> ok: true; duration: now 7 ms, total 14 ms (count 2, avrg 7
20150818 10:16:00 <sgothel> 7ms .. ops :)
20150818 10:16:38 <sgothel> so 14ms .. on a raspi .. probably much longer
20150818 10:17:11 <sgothel> hence .. it might be a good idea to offload the initSingleton completely to another thread .. if you can do other things in the meantime
20150818 10:27:45 <xranby_> sgothel: on the pi the time is now 80ms good
20150818 10:28:01 <sgothel> w/ single-slim jars?
20150818 10:28:28 <xranby_> yes -cp gluegen-rt.jar:jogl-all.jar
20150818 10:30:11 * gouessej (5ee4b442@anon) has joined #jogamp
20150818 10:30:53 <gouessej> xranby_: it is a good piece of news for the Raspberry Pi with JogAmp :)
20150818 10:31:32 <xranby_> gouessej: the 80ms if the time spent in the JNILibLoaderBase
20150818 10:31:45 <xranby_> total time on the pi is still ~6s
20150818 10:32:30 <sgothel> the BCM blob init .. we know it takes alot of time as well
20150818 10:32:35 <gouessej> I wonder how much time it takes to start an application under Android with OpenJDK, several seconds would be a no go
20150818 10:32:55 <gouessej> sgothel: I have to talk to you about StreamInput
20150818 10:34:04 <sgothel> https://jogamp.org/bugzilla/show_bug.cgi?id=1042 ?
20150818 10:34:04 <gouessej> sgothel: I have investigated a bit, it seems to be an higher level API, a bit like OpenVX and WebAudio, graph-based
20150818 10:34:23 <rmk0> gouessej: 2.3.2-rc-20150812 should be in the jogamp repository now
20150818 10:34:24 <gouessej> sgothel: no, NEWT Input, etc... sorry
20150818 10:34:44 <gouessej> rmk0: Thanks, I can update JogAmp's Ardor3D Continuation tonight
20150818 10:34:44 <sgothel> ah
20150818 10:35:20 <gouessej> sgothel: In my humble opinion, it would be better to use lower level APIs first and then maybe look at StreamInput
20150818 10:35:46 <gouessej> For the others: https://www.khronos.org/streaminput/
20150818 10:36:01 <sgothel> 'StreamInput' .. err .. ah .. well, not for 2.3.2 for sure :)
20150818 10:36:12 <gouessej> sgothel: Of course :)
20150818 10:36:29 <xranby_> looks good for 2.3.3!
20150818 10:36:31 <sgothel> and also: ofc .. this is your turf .. i.e. ballgame
20150818 10:36:55 <sgothel> if you can do a standalone API/binding .. we may use that for NEWT then as well ..
20150818 10:37:13 <sgothel> but as you pointed out .. this is sort of independent
20150818 10:37:14 <gouessej> sgothel: I try to take into account your suggestions
20150818 10:37:39 <sgothel> i.e. like we stitch other input into NEWT (Linux event driver .. etc)
20150818 10:38:13 <sgothel> I know .. its a big thing .. configure input devices, capabilities, etc
20150818 10:38:18 <gouessej> sgothel: I will keep it small and make it a separate branch but I'm still planning to put it directly into NEWT
20150818 10:38:35 <sgothel> as you wish - ofc, great!
20150818 10:39:31 <sgothel> btw .. is Bug 1042 work done and has no regressions? what else has to be done here?
20150818 10:39:56 <gouessej> sgothel: just the move of the patterns into a separate internal class
20150818 10:40:07 <sgothel> ah ..
20150818 10:40:27 <sgothel> other than that .. it works and the new code path is being used all the time?
20150818 10:40:40 <gouessej> sgothel: it was already working as is in JogAmp's Ardor3D Continuation for several months (years?)
20150818 10:40:43 <sgothel> (i.e. implicitly tested using our TextureIO tests)
20150818 10:41:01 <sgothel> yeah, I know .. still asking :)
20150818 10:41:12 <sgothel> new code path is being used all the time?
20150818 10:41:39 <gouessej> There is a few cases in which it isn't used
20150818 10:41:48 <gouessej> TGA no, no magic number
20150818 10:42:04 <gouessej> "There are"
20150818 10:42:10 <sgothel> ah .. 'great' :)
20150818 10:42:29 <sgothel> right .. its a raw format AFAIK
20150818 10:42:32 <gouessej> It is used in other providers when you read
20150818 10:42:45 <gouessej> I haven't looked at the writers
20150818 10:42:47 <sgothel> ok .. reviewing .. and closing bug afterward. OK?
20150818 10:43:07 <gouessej> Yes but feel free to make any necessary cosmetic changes
20150818 10:43:17 <sgothel> ay ay - thank you
20150818 10:43:18 <xranby_> gouessej: TGA version 2 claims to have a "footer" containting TRUEVISION-XFILE
20150818 10:43:25 <xranby_> https://en.wikipedia.org/wiki/Truevision_TGA
20150818 10:43:50 <xranby_> its kind of more work looking for this
20150818 10:43:57 <xranby_> but it may serve as a magic number
20150818 10:44:05 <xranby_> if you can read the entire file
20150818 10:44:39 <gouessej> the footer is optional :s
20150818 10:46:08 <gouessej> I could improve ImageIOUtil to look for a pattern in the header and in the footer
20150818 10:47:08 <gouessej> sgothel: is there a plan to support GIF?
20150818 10:48:00 <gouessej> sgothel: Maybe you would like to use BitStream in ImageIOUtil, wouldn't you?
20150818 10:48:48 <xranby_> the LZW compression algorithm used in git -> patented
20150818 10:48:51 <xranby_> gif
20150818 10:49:59 <xranby_> "In 1999, Unisys had the following to say about the issue of their patent:Unisys has frequently been asked whether a Unisys license is required in order to use LZW software obtained by downloading from the Internet or from other sources. The answer is simple. In all cases, a written license agreement or statement signed by an authorized Unisys representative is required from Unisys for all use, sale or distribution of any softwar
20150818 10:50:04 <xranby_> http://www.gnu.org/philosophy/gif.html
20150818 10:53:54 <gouessej> You're right: http://burnallgifs.org/ :)
20150818 10:55:50 <gouessej> xranby_: ImageIOUtil could read the header and the footer in the method that takes an input stream in input and pass it to getFileSuffix() that would take a second byte array for the footer
20150818 10:56:38 <gouessej> xranby_: Should I fill a bug report and implement it later (not in 2.3.2)?
20150818 10:58:25 <gouessej> xranby_: The specification discrepancies make it hard to support this format anyway
20150818 11:21:05 <sgothel> @Julien: quite many references to those parsed formats are missing .. do you have more information you could drop in ImageIOUtil?
20150818 11:23:11 <sgothel> ImageIOUtil shall *not* read the footer -> performance .. and it used buffered stream mark/reset!
20150818 11:45:42 <gouessej> sgothel: Then we won't implement TGA support in ImageIOUtil
20150818 11:46:09 <gouessej> sgothel: elect had an interesting reference from MSDN for DDS
20150818 11:46:32 <elect> wut
20150818 11:46:48 <sgothel> I push a little patch in a sec to ImageIOUtul ..
20150818 11:46:57 <sgothel> Q: those formats are all images only?
20150818 11:47:08 <gouessej> sgothel: https://msdn.microsoft.com/en-us/library/windows/desktop/bb943991%28v=vs.85%29.aspx#File_Layout1
20150818 11:47:21 <gouessej> sgothel: yes
20150818 11:47:52 <sgothel> ImageIOUtil -> ImageType .. hmm
20150818 11:48:45 <sgothel> it would be nice if you can add docs to the next version I push ..
20150818 11:50:03 <gouessej> I wouldn't rename this class to ImageType
20150818 11:50:22 <gouessej> because it is similar to IOUtil and it handles images
20150818 11:50:40 <gouessej> but you can create an internal class called ImageType for the patterns
20150818 11:50:48 <gouessej> it would make sense to me
20150818 11:51:35 <sgothel> I would keep the static stuff in ImageType (or make an internal Utils) .. and allow an instance of ImageType for better documentation/source
20150818 11:54:19 <gouessej> What would such instances contain?
20150818 11:54:39 <gouessej> a label and a pattern?
20150818 11:54:40 <sgothel> just a marker .. i.e. the string
20150818 11:54:48 <sgothel> but a better type than String :)
20150818 11:54:57 <sgothel> dinner laters .. 1-2 h
20150818 11:55:10 <gouessej> ok enjoy youself
20150818 12:49:58 <xranby_> zubzub: i am testing your travis-ci scripts on my jogl branch -> https://travis-ci.org/xranby/jogl
20150818 12:53:16 <zubzub> well, looks like it's doing *something* at least :)
20150818 12:57:38 <zubzub> looks like it succeeded in building
20150818 12:57:41 <zubzub> now it's running the unit tests
20150818 12:57:42 <xranby_> indeed
20150818 12:58:40 <xranby_> junit.run usually takes 45 min .. and this build derver will try run them using the mesa3d software renderer
20150818 12:59:05 <xranby_> is the log file accessible?
20150818 13:00:14 <zubzub> not unless you explicitly upload somewhere afaik
20150818 13:02:54 <xranby_> checked what to expect to happen -> https://travis-ci.org/Zubnix/jogl
20150818 13:02:54 <xranby_> /home/travis/build/Zubnix/gluegen/make/jogamp-archivetasks.xml:23: Execute failed: java.io.IOException: Cannot run program "7z" (in directory "/home/travis/build/Zubnix/jogl"): error=2, No such file or directory
20150818 13:03:31 <xranby_> need to add p7zip-full dependency to the travis-ci script
20150818 13:04:45 <xranby_> at least good that it shows a list of test failing on mesa3d using software renderer
20150818 13:05:05 * badshah400 (~badshah40@anon) Quit (Quit: badshah400)
20150818 13:05:49 <xranby_> look at that
20150818 13:06:45 <xranby_> sgothel: the travis build bot found a crash using its mesa3d software rasterizer :/ https://travis-ci.org/xranby/jogl
20150818 13:06:56 <xranby_> looks like the crash may be inside the llvm backend
20150818 13:07:07 <xranby_> of mesa3d
20150818 14:23:11 * mjavka (uid63119@anon) has joined #jogamp
20150818 14:26:09 <xranby_> mjavka: welcome
20150818 14:57:14 * rmk0_ (~rmk0@anon) has joined #jogamp
20150818 14:59:10 <xranby_> well travis has its limitations... or our tests are too verbose " The log length has exceeded the limit of 4 Megabytes " https://s3.amazonaws.com/archive.travis-ci.org/jobs/76115357/log.txt
20150818 14:59:41 * rmk0 (~rmk0@anon) Quit (*.net *.split)
20150818 15:03:08 <zubzub> xranby_: that's why I put it in a seperate file :)
20150818 15:08:50 * gouessej (5ee4b442@anon) Quit (Quit: Page closed)
20150818 15:16:09 * xranby_ (~familjen@anon) has left #jogamp
20150818 15:36:39 * mjavka (uid63119@anon) Quit (Remote host closed the connection)
20150818 15:46:21 * elect (~elect@anon) Quit (Ping timeout: 256 seconds)
20150818 15:51:04 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20150818 16:15:39 * eclesia (~husky@anon) has left #jogamp
20150818 16:16:28 * elect (~elect@anon) has joined #jogamp
20150818 16:19:02 * mjavka (uid63119@anon) has joined #jogamp
20150818 18:21:32 * mjavka (uid63119@anon) Quit (Quit: Connection closed for inactivity)
20150818 18:22:36 * elect (~elect@anon) Quit (Ping timeout: 255 seconds)
20150818 18:45:14 * badshah400 (~badshah40@anon) has joined #jogamp
20150818 19:48:49 * elect (~elect@anon) has joined #jogamp
20150818 20:39:45 * badshah400 (~badshah40@anon) Quit (Quit: badshah400)
20150818 20:52:53 * elect (~elect@anon) Quit (Ping timeout: 265 seconds)
20150819 00:07:03 * badshah400 (~badshah40@anon) has joined #jogamp
20150819 02:07:40 * a11mad11 (4c47eb44@anon) has joined #jogamp
20150819 02:41:17 * badshah400 (~badshah40@anon) Quit (Quit: badshah400)
20150819 04:03:03 * a11mad11 (4c47eb44@anon) Quit (Ping timeout: 246 seconds)
20150819 04:18:25 * badshah400 (~badshah40@anon) has joined #jogamp
20150819 05:05:30 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20150819050530.html