#jogamp @ irc.freenode.net - 20160915 05:05:15 (UTC)


20160915 05:05:15 -jogamp- Previous @ http://jogamp.org/log/irc/jogamp_20160914050514.html
20160915 05:05:15 -jogamp- This channel is logged @ http://jogamp.org/log/irc/jogamp_20160915050515.html
20160915 05:44:38 * SHC (~quassel@anon) has joined #jogamp
20160915 06:59:24 * elect (~elect@anon) has joined #jogamp
20160915 07:01:08 <elect> hi
20160915 07:28:26 * Eclesia (~husky@anon) has joined #jogamp
20160915 07:29:11 <Eclesia> hi
20160915 07:37:44 * elect (~elect@anon) Quit (Read error: Connection reset by peer)
20160915 07:42:52 * elect (~elect@anon) has joined #jogamp
20160915 08:18:41 * monsieur_max (~maxime@anon) has joined #jogamp
20160915 08:41:02 * SHC_ (~quassel@anon) has joined #jogamp
20160915 08:44:19 * SHC (~quassel@anon) Quit (Ping timeout: 248 seconds)
20160915 09:04:16 * elect (~elect@anon) Quit (Quit: Leaving)
20160915 09:06:05 * elect (~elect@anon) has joined #jogamp
20160915 09:18:18 <elect> had some perf issues in vr
20160915 09:18:50 <elect> debugging with vs nsight, at each frame jogl simply throw away over 1ms doing nothing
20160915 09:25:51 * SHC_ (~quassel@anon) Quit (Read error: Connection reset by peer)
20160915 09:28:24 <elect> also the buffer swapping
20160915 09:28:31 <elect> 330 us on jogl
20160915 09:28:35 <elect> 30 on C
20160915 09:28:57 <monsieur_max> ouch
20160915 09:30:47 <elect> but the good new is that in the sample
20160915 09:31:05 <elect> jogl and c takes the same time from the begin of the first call to the end of the last rendering call
20160915 09:56:22 <elect> http://imgur.com/a/JiWkA
20160915 09:56:50 <elect> we get a wglMakeCurrent at 6.689s
20160915 09:57:04 <elect> first gl call comes at 6.690
20160915 09:57:49 <elect> 6.6896 vs 6,6904
20160915 09:58:15 <elect> 0.8 ms
20160915 10:04:59 <rmk0> elect: is this 1ms delay specific to the VR code?
20160915 10:05:06 <rmk0> 1ms delay is rather a lot...
20160915 10:11:22 <elect> no
20160915 10:11:30 <elect> I'd say it's internal to jogl
20160915 10:12:15 <rmk0> nasty
20160915 10:12:27 <elect> I am wondering
20160915 10:12:38 <elect> is it necessary to call a makeCurrent every frame?
20160915 10:12:53 <elect> btw this is the original C sample
20160915 10:12:53 <elect> http://imgur.com/a/i2OFM
20160915 10:13:02 <elect> first call glBindBuffer right there
20160915 10:13:17 <elect> dont get confused by the size, they have different zooming
20160915 10:13:49 <rmk0> struggling to read that
20160915 10:13:56 <elect> I know
20160915 10:14:06 <elect> do you have VS?
20160915 10:14:10 <rmk0> the glBindBuffer call took 6361us?
20160915 10:14:17 <rmk0> visual studio?
20160915 10:14:26 <rmk0> am on linux
20160915 10:14:53 <elect> 6,361 us
20160915 10:15:25 <rmk0> that's ...
20160915 10:15:42 <rmk0> i assume they don't all take that long
20160915 10:15:46 <rmk0> would be unusable
20160915 10:17:11 <elect> the format is called nvact
20160915 10:17:15 <elect> I saved it
20160915 10:17:25 <elect> but I don't think there is anything to open that on linux
20160915 10:17:42 <elect> for the frametime analysis I mean
20160915 10:20:09 <rmk0> probably not
20160915 10:20:32 <elect> although I recall we should be able to play with nvsight also on linux
20160915 10:20:40 <elect> wanna give it a try?
20160915 10:21:18 <rmk0> if it requires nvidia hardware, i don't have any
20160915 10:22:20 <elect> I'd say no
20160915 10:22:26 <elect> but I am not sure
20160915 11:13:21 <zubzub> 12:12 < elect> is it necessary to call a makeCurrent every frame?
20160915 11:13:23 <zubzub> no
20160915 11:13:34 <zubzub> unless your context changed
20160915 11:13:37 <zubzub> gl rendering context
20160915 11:13:38 <zubzub> that is
20160915 11:13:43 <zubzub> surface, thread and errr
20160915 11:13:49 <zubzub> config?
20160915 11:14:36 <rmk0> i think it's fairly commonplace to do it anyway
20160915 11:14:56 <zubzub> it does take some overhead if you do
20160915 11:15:19 <rmk0> urhur
20160915 11:35:35 <elect> who changes the context?
20160915 11:35:44 <elect> who may
20160915 11:35:48 <elect> app and system?
20160915 11:38:38 <zubzub> anybody that wishes to do gl calls
20160915 11:38:52 <zubzub> if he's doing it on a different thread
20160915 11:39:01 <zubzub> or if on a different surface
20160915 11:39:13 <zubzub> or with an another gl config
20160915 11:39:20 <elect> are there any indication how this is handled by jogl?
20160915 11:39:44 <zubzub> no idea
20160915 11:40:04 <rmk0> if i remember correctly, if you attach an animator to a GLWindow, it'll release/makeCurrent each frame
20160915 11:40:22 <elect> interesting
20160915 11:47:05 <elect> worst case scenario
20160915 11:47:22 <elect> what happens if you call a gl call from a thread without context?
20160915 11:47:30 <zubzub> it fails
20160915 11:47:37 <elect> but it doesnt crash?
20160915 11:47:54 <zubzub> not unless you have sucky drivers :p
20160915 11:47:58 <elect> ^^
20160915 11:48:10 <rmk0> it won't necessarily fail in an obvious way either
20160915 11:48:30 <rmk0> on intel, it seems like all gl calls turn into stubs, some of which will print strange messages to stderr
20160915 11:49:08 <zubzub> you should get somethiung like no context current afaik
20160915 11:49:15 <zubzub> but yeah you never know what the driver might do
20160915 11:49:24 <zubzub> don't know if it's specced what should happen
20160915 14:19:26 * SHC (~quassel@anon) has joined #jogamp
20160915 14:20:46 * SHC (~quassel@anon) Quit (Read error: Connection reset by peer)
20160915 14:21:07 * SHC (~quassel@anon) has joined #jogamp
20160915 14:22:09 * SHC (~quassel@anon) Quit (Read error: Connection reset by peer)
20160915 14:22:37 * SHC (~quassel@anon) has joined #jogamp
20160915 14:27:04 * SHC_ (~quassel@anon) has joined #jogamp
20160915 14:27:15 * SHC (~quassel@anon) Quit (Ping timeout: 248 seconds)
20160915 14:31:10 * SHC_ is now known as SHC
20160915 15:31:19 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20160915 15:37:07 * SHC (~quassel@anon) Quit (Ping timeout: 248 seconds)
20160915 16:01:36 * Eclesia (~husky@anon) has left #jogamp
20160915 16:11:50 * elect (~elect@anon) Quit (Ping timeout: 244 seconds)
20160915 16:57:20 * monsieur_max (~maxime@anon) has joined #jogamp
20160915 18:31:03 * monsieur_max (~maxime@anon) Quit (Remote host closed the connection)
20160915 18:33:45 * monsieur_max (~maxime@anon) has joined #jogamp
20160915 20:37:39 * monsieur_max (~maxime@anon) Quit (Quit: Leaving.)
20160916 05:05:15 -jogamp- Continue @ http://jogamp.org/log/irc/jogamp_20160916050515.html