<_inky>
rafael2k: and thank you too for advices and ideas.
<_inky>
i think eventually we have breakthroughs with pinephone support.
<_inky>
and this is very important: people come to linux phones and operating systems via pinephone.
<_inky>
so maemo will actually get more users via pinephone.
<_inky>
i'll ask these couple of days if i need to update from -devel to get the updates you all made.
tanakian has quit [Quit: tanakian]
BenLand100 has quit [Ping timeout: 268 seconds]
BenLand100 has joined #maemo-leste
BenLand100 has quit [Changing host]
BenLand100 has joined #maemo-leste
Pali has quit [Ping timeout: 240 seconds]
joerg has quit [Ping timeout: 240 seconds]
joerg has joined #maemo-leste
Wikiwide has joined #maemo-leste
macros_ has joined #maemo-leste
macros__ has quit [Ping timeout: 268 seconds]
<tmlind>
hmm my bionic is all updated now, but not accelerated.. i have sgx-ddk-um-ti443x and xserver-xorg-video-omap installed, kernel module is initialized on boot..
<freemangordon>
tmlind: check xorg.conf
<tmlind>
freemangordon: don't seem to have one.. should it be under /etc/X11?
<freemangordon>
no, /usr/share/X11/xorg.conf.d
<tmlind>
ok
<freemangordon>
there should be 99-omap
<tmlind>
i have 99-omap.conf symlinked to 99-omap.conf.leste
<freemangordon>
yeah
<freemangordon>
that's fine
<freemangordon>
could you provide xorg log, from /tmp
<tmlind>
also trying to run glxgears as user user fails
<tmlind>
/dev/dri is group readable/writable for video
<freemangordon>
glxgears?
<freemangordon>
you should run es2_gears
<tmlind>
oh ok
<freemangordon>
OGL is not accelerated
<freemangordon>
brb
<tmlind>
ok es2gears works :)
<tmlind>
stellarium is not accelerated, let me recheck
<tmlind>
starting stellarium as user produces: QXcbIntegration: Cannot create platform OpenGL context, neither GLX nor EGL are enabled
<tmlind>
i can start is as root but it's not accelerated
<tmlind>
fyi
<tmlind>
hmm es2tri triangle is only partially visible somehow
<freemangordon>
does glmark2_es2 work?
<freemangordon>
tmlind: actually, does your device boot to h-d?
<tmlind>
freemangordon: yeah boots now to h-d, looks like stellarium tries to start with sgx with QT_QPA_PLATFORM=eglfs stellarium
<tmlind>
then segfaults
<tmlind>
let me check glmark2_es2
<tmlind>
hmm can't find glmark2 package..
<freemangordon>
h-d is accelerated, right?
<freemangordon>
glmark2 shall be compiled, there is no package
<tmlind>
ok, h-d looks a bit jerky, not sure if it's accelerated
<freemangordon>
hmm, weird
<tmlind>
i'm pretty sure h-d is accelerated, only jerky here and there
<freemangordon>
that's weird, should be smooth
<freemangordon>
but well, i have d4 here, not bionic
<tmlind>
should be about the same
<freemangordon>
maybe lets wait for Wizzup to share his experience
<freemangordon>
but swiping h-d should be smooth
<freemangordon>
well, non-accelerated h-d renders with 1-2 fps, so... :)
<tmlind>
yeah it's mostly smooth, sometimes starting a new app the animation is kind of jerky..
<freemangordon>
ah
<freemangordon>
well, for sure we have issues on omap, there is also tearing and the reported fps looks too high for what is really seens
<freemangordon>
*seen
<freemangordon>
but otherwise is useful
<tmlind>
yup looks way better than earlier :)
<tmlind>
heh looks like gdb shows stellarium segfaults at ssl libcrypto somewhere..
<tmlind>
ok got stellarium running with QT_QPA_PLATFORM=eglfs stellarium
<tmlind>
sorry with QT_QPA_PLATFORM=eglfs stellarium --home-planet=mars
<tmlind>
some location lookup issue it seems :)
<tmlind>
about 9 fps it seems
<freemangordon>
:)
<freemangordon>
hmm, shouldn;t it be faster?
<tmlind>
i'm seeing someting like 12 fps with wayland
<freemangordon>
ah
<freemangordon>
well, given that compositing is non-accelerated...
<freemangordon>
hmm, maybe I shall RE that part too
<freemangordon>
btw, is it fullscreen?
<tmlind>
yeah it's full screen
<freemangordon>
hmm
<freemangordon>
should be the same like WL
<freemangordon>
if not faster
<tmlind>
let me check again on wl
<freemangordon>
with 'mars'
<tmlind>
uh looks like i have some pvr issue with my wl setup right now, need to check the fps later
<tmlind>
switching from stellarium to h-d window list hung xorg
<freemangordon>
hmm
<freemangordon>
some oops?
<tmlind>
no, Xorg.0.log has many (EE) OMAP(0): ERROR: waitForBlitsCompleteOnDeviceMem: PVR2DQueryBlitsComplete failed with error code: -8 (Blit not complete)
<freemangordon>
this should not result in a hang
<tmlind>
keeps producing them about every 500 ms
<freemangordon>
dmesg?
<freemangordon>
I would say there is HW recovery trigered
<tmlind>
no errors
<freemangordon>
weird
<tmlind>
what should i type to restart xorg?
<freemangordon>
nothing
<freemangordon>
I mean - you shoud restart the device
<tmlind>
killall -9 Xorg at least got fb back
<freemangordon>
yeah
<freemangordon>
hmm, we even have tearing with modesetting without glamor
<freemangordon>
this is omapdrm behaving I would say
<tmlind>
weird now it's behaving after a reboot
<freemangordon>
hmm?
<freemangordon>
you mean there are no issues?
<tmlind>
can't make QT_QPA_PLATFORM=eglfs stellarium --home-planet=mars hang after switching between app view and stellarium
<freemangordon>
it should not :)
<freemangordon>
I was hoping modesetting with glamor will hit > 50 fps on d4
<freemangordon>
but unfortunately it hits 50fps and that's it
<freemangordon>
and it suffers from the same tearing issue
<tmlind>
anyways a massive upgrade to get rid of all the old buggy legacy crap with the kernels and ddk-1.9 :)
* tmlind
goes to investigate fridge for the ham
<freemangordon>
tmlind: any advice on how to deal with omapdrm tearing issue? I need upstream support, but I doubt Tomi will care about that
Twig has joined #maemo-leste
Wikiwide has quit [Ping timeout: 268 seconds]
<Wizzup>
rafael2k: once I get home in ~8 hours I'll look at packaging this stuff, ty
<Wizzup>
08:56 < tmlind> heh looks like gdb shows stellarium segfaults at ssl libcrypto somewhere..
<Wizzup>
this is openssl testing instructions (sigill)
<Wizzup>
freemangordon: same for lima bug/video, will do that when I get back
<freemangordon>
ok
<freemangordon>
no hurry
<Wizzup>
:)
<Wizzup>
Just don't have the device here at the moment
<freemangordon>
yay, omapdrm BO patch is in -next :)
<tmlind>
freemangordon: what's the omapdrm tearing issue?
<tmlind>
Wizzup: i guess that openssl issue will get fixed upstream pretty fast
<freemangordon>
this happens more often if your battery is low
<tmlind>
weird
<freemangordon>
but I am almost sure this is either PVR or h-d/clutter bug
<tmlind>
so a variation of the ants problem i see on wayland is a bit larger black rectangles in the same location, top left corner for landscape mode
<freemangordon>
because this corruption is exactly behind and with the size of six squares icon
<freemangordon>
could be
<freemangordon>
but I almost never see that after a reboot
<freemangordon>
only if I play stopping/starting xorg/h-d
<tmlind>
well m-l should revert "drm/omap: Fix omap_gem_dma_sync_buffer() when we already have a dma_addr" if it causes no other issues, wayland is unusable without that patch
<freemangordon>
this should be fixed in wl though
<tmlind>
no idea how it should be fixed :(
<tmlind>
especially if m-l also shows a similar issue on xorg
<freemangordon>
it should use sync
<freemangordon>
xorg uses "wait for blits"
<freemangordon>
but WL can include sync objects in GL commands stream and wait for them, IIUC
<tmlind>
i don't think i've seen the ants with weston
<freemangordon>
ugh, I should have fixed TS buttons udev rules before reboot :(
<freemangordon>
now I have to wait device to autoconnect
<tmlind>
weird how the autoconnect stuff still takes few mins
<freemangordon>
not on pinephone though
<tmlind>
weird
<freemangordon>
so it is something with d4 driver
<tmlind>
i wonder if adding iw dev wlan0 scan on start up would help
<freemangordon>
still tears
<freemangordon>
lemme check fps
<freemangordon>
oh
<freemangordon>
we hit 80 fps with ease
<tmlind>
nice
<tmlind>
maybe check the stellarium on mars rate too?
<freemangordon>
is it in repos?
<tmlind>
yeah apt-get install stellarium
<tmlind>
takes a while for the data files though
<freemangordon>
maep renders with 45 fps though
<freemangordon>
this is with 3-buffer
<tmlind>
need to go for some family time now, ttyl
<freemangordon>
ttyl
<freemangordon>
yay! I know why fps is too low - tmlind broke it :p
Treebeard has joined #maemo-leste
Treebeard has joined #maemo-leste
Treebeard has quit [Changing host]
BenLand100 has quit [Ping timeout: 240 seconds]
Treebeard is now known as BenLand100
<freemangordon>
tmlind: fps ~9
<freemangordon>
but, nothing moves, is that normal?
<freemangordon>
oh, ok, we hit 80 fps without 3-buffer
<freemangordon>
I think we shall disable it by default
<freemangordon>
at least on d4 it doesn;t seem we need 3-buffer
<rafael2k>
Wizzup: tks
<rafael2k>
Wizzup: ofono will be pretty easy
<rafael2k>
Wizzup: kernel wise, I want some help with git, and also some suggestion if I should just add the line in rules to install kernel, initrd and dts to /boot, or may be use a postinst scrit...
<freemangordon>
wow, firefox renders as mad!
<freemangordon>
hmm, why there is no sound?
<rafael2k>
Wizzup: dtbs are being installed in /usr/lib/linux-image-5.15.10/allwinner/ and I copied by hand to /boot/allwinner
<rafael2k>
Wizzup: kernel and initrd are installed, but they dont get copied to ddefault locations specified in boot.txt
<freemangordon>
so, this buffer is output for gpu and input for display controller
<freemangordon>
and this is display controller (so CPU) doing sync
<freemangordon>
but there is nothing to be synced, IIUC, as CPU does not render
<freemangordon>
I think this commit mitigates the issues tmlind has with WL as it adds a delay, during which GPU finishes rendering
<freemangordon>
but not fully, as sometims GPU needs more than 6ms to render
<freemangordon>
does this sound sane?
<freemangordon>
it is absolutely possible that I am talking nonsense here, but the net effect is that reverting that results in stable 80fps being hit on h-d swiping without any visible rendering artifacts, besides the tearing and ants, but they are there with and without that commit
<freemangordon>
and also removing that commit results in drmModePageFlip() behave according to specs
<freemangordon>
(didn;t check xorg logs though)
xmn has joined #maemo-leste
<Wizzup>
freemangordon: ok so I will revert 6dd1aad1d0c4716944098b7b0f2abdd3a26cc0be for now in our -devel kernel
rafael2k has quit [Ping timeout: 256 seconds]
<tmlind>
ack, that patch is a hack
<freemangordon>
Wizzup: :mod:
<freemangordon>
:nod:
<freemangordon>
Wizzup: shall I disable 3-buffer to be enabled by default in omap code?
<freemangordon>
I think this is the sane
<freemangordon>
the other option is to disable it in config
<Wizzup>
I would disable by default if we have no use for it
<freemangordon>
ok, will disable and spin a new build
<Wizzup>
ty
<Wizzup>
maybe wait for kernel to complete
<freemangordon>
for the build?
<freemangordon>
sure
<freemangordon>
tmlind: can you recall any other hack patch that might be related to tearing?
<freemangordon>
hmm, I think I know what the possible fix for the corruption could be, in xorg
<freemangordon>
there is pvr driver call that flushes GPU buffers to memory
<Wizzup>
great @ corruption
<freemangordon>
well, I think this is of low prio
<Wizzup>
mhm
<tmlind>
freemangordon: well are you seeing the same issues on n900? if not, it's somehow related to the tiler possibly