**** BEGIN LOGGING AT Tue Jul 02 02:59:56 2019
Jul 02 06:42:01 <Herrie> Morning!
Jul 02 07:17:10 <Tofe> Morning!
Jul 02 07:26:46 <Herrie> Tofe: What parameters should I pass to top?
Jul 02 07:27:55 <Sho_> top -down
Jul 02 07:29:43 <Tofe> :)
Jul 02 07:29:44 <Tofe> none
Jul 02 07:29:49 <Tofe> term=XTERM top
Jul 02 07:30:00 <Tofe> no TERM=xterm top
Jul 02 07:31:15 <Herrie> Output in my adb is a bit garbled
Jul 02 07:31:18 <Herrie> Let me paste it though
Jul 02 07:31:38 <Herrie> https://bpaste.net/show/wkQc
Jul 02 07:38:34 <Tofe> mmmh
Jul 02 07:38:55 <Tofe> you don't have much ram left, do you
Jul 02 07:48:06 <Herrie> Seems not
Jul 02 07:48:12 <Herrie> But that's expected seeing it's 1GB only
Jul 02 07:48:22 <Herrie> I thought we enabled ZRAM etc?
Jul 02 07:52:50 <Tofe> maybe it's the case, I'm not sure how it appears in standard tools in this case
Jul 02 07:53:20 <Tofe> anyway if it ran out of ram I guess you would get a more dramatic experience
Jul 02 08:08:45 <Herrie> Yeah it's slow as hell and just locks up at some point
Jul 02 15:16:46 <novaldex>  /quit
Jul 02 17:22:37 <Tofe> Herrie: it's swapping
Jul 02 17:23:53 <Tofe> but I'm surprised, I still see
Jul 02 17:24:17 <Tofe> Mem: total 817Mi
Jul 02 17:24:48 <Tofe> (free 5.0Mi , swap 135Mi used, etc )
Jul 02 17:25:00 <Tofe> so we're completely over the edge
Jul 02 17:26:34 <Tofe> luna-next, maliit-server and LunaWebAppManager are the big top 3 (resp. 600Mb, 200Mb and 200Mb, they share a lot, but still...)
Jul 02 17:33:42 <Tofe> I don't think we take advantage of zram at all
Jul 02 18:04:32 <Tofe> damn, even with zram enabled it's quickly painfully slow
Jul 02 18:07:11 <Herrie> Tofe: I read somewhere today that zram support was added in 3.14 kernel
Jul 02 18:07:23 <Herrie> Not sure if that's an accurate statement though
Jul 02 18:11:01 <Tofe> given that I'm using it right now, I'm not sure either
Jul 02 18:29:39 <Tofe> but did we use so much ram before ? it feels like it's using a lot now
Jul 02 18:49:02 <Herrie> Well could be Chromium upgrade?
Jul 02 18:49:31 <Herrie> Chromium is notorious for memory hogging
Jul 02 18:55:58 <Herrie> Not sure why luna-next would take 600mb though
Jul 02 18:56:11 <Herrie> That's way too much for what it should be providing
Jul 02 18:56:36 <Herrie> s/should be/is
Jul 02 19:05:21 <Herrie> In terms of functionality I mean
Jul 02 19:05:38 <Herrie> Qt is used quite a lot on embedded devices as well, so it shouldn't be such a memory hog
Jul 02 19:09:23 <Tofe> it could be the textures or things like tha
Jul 02 19:09:24 <Tofe> t
Jul 02 19:09:37 <Tofe> we should compare with another device
Jul 02 19:09:52 <Herrie> Yes
Jul 02 19:24:49 <Tofe> it can be compared
Jul 02 19:25:15 <Tofe> ah no wait, it's not the same column
Jul 02 19:25:53 <Tofe> 1210 root      20   0 1195936 704060 704040 D   7.3  84.1   5:30.48 luna-next  <-- TP
Jul 02 19:25:54 <Tofe> 1648 root      20   0  535996  87540  49168 S   3.6   4.6   0:08.98 luna-next       <-- N5
Jul 02 19:26:08 <Tofe> Sir, we have a problem
Jul 02 19:27:31 <Herrie> So likely something with textures/graphics?
Jul 02 19:28:54 <Tofe> I don't know...  maybe... but I don't see why; the screen doesn't have more pixels than N5
Jul 02 19:36:57 <Herrie> Maybe kernel somehow
Jul 02 19:37:12 <Herrie> Seems that they tweaked quite a bit for zram in the 7.1 version
Jul 02 19:37:53 <Herrie> https://github.com/Evervolv/android_kernel_htc_msm8960/commits/ng-7.1
Jul 02 19:39:19 <Herrie> Might be worth to give that kernel a try, just we'd need to add our patches there
Jul 02 19:40:45 <Herrie> There's even a 9.0 kernel it seems :S
Jul 02 19:43:56 <Tofe> but still 3.4 :(
Jul 02 19:45:00 <Herrie> True
Jul 02 19:45:13 <Herrie> There are some mainlining efforts as well but not very stable I think
Jul 02 19:45:51 <Herrie> I mean forking, adding a few patches, updating defconfig and buiding the uImage shouldn't take too long to at least making sure it's not our kernel config & patches
Jul 02 19:49:02 <Tofe> I'm not sure I follow you, aren't we using the same kernel as for doppio, more or less?
Jul 02 19:49:55 <Herrie> Tofe: Yeah that's true
Jul 02 19:50:11 <Herrie> I'm also not sure kernel is the problem, but it might be quicker to test v.s. debugging the rest
Jul 02 19:50:12 <Herrie> ?
Jul 02 19:50:22 <Herrie> I'm not even sure where to start to debug other things ;)
Jul 02 19:50:30 <Herrie> I'm sure you have clues though :P
Jul 02 19:51:20 <Tofe> weeeeell..........
Jul 02 19:51:21 <Herrie> Just seems that they have added quite some commits in the kernel for newer Android versions to optimize memory usage etc
Jul 02 19:52:08 <Tofe> it sure can help a bit, we could gain some Mb here and there, but it won't end with a 60% gain
Jul 02 19:52:17 <Herrie> That's true probably
Jul 02 19:52:33 <Herrie> I can have a go at tinkering with a newer kernel tomorrow a bit
Jul 02 19:52:41 <Herrie> Shouldn't take me too long, just to have try
Jul 02 19:52:56 <Tofe> :) can't stop you
Jul 02 19:52:57 <Herrie> Worst thing that can happen is that it won't boot :P
Jul 02 19:53:39 <Tofe> it'll probably boot, but there's also a good chance that halium won't like it
Jul 02 19:54:17 <Herrie> With the right defconfig it should be OK-ish I'd hope
Jul 02 19:54:25 <Tofe> maybe yes
Jul 02 19:55:16 <Herrie> Just mind is not clear enough to start with it now :P
Jul 02 19:55:24 <Tofe> I wonder what could be tp-specific... gralloc ?
Jul 02 19:56:18 <Tofe> we did upgrade hybris a bit
Jul 02 19:56:57 <Herrie> What I understood the graphics were always a bit hacky on Tenderloin
Jul 02 19:57:13 <Tofe> yes, HP patched android's standard API
Jul 02 19:58:14 <Tofe> it's nothing new, but maybe we're not releasing memory when we should
**** ENDING LOGGING AT Wed Jul 03 03:00:28 2019