stikonas has quit [Remote host closed the connection]
anarsoul|2 has quit [Ping timeout: 250 seconds]
NeuroScr has quit [Quit: NeuroScr]
NeuroScr has joined #panfrost
memeka has joined #panfrost
<memeka>
hi, how can i build panfrost on exynos?
<memeka>
i got it built with -Dgallium-drivers=panfrost but when running something, dri/panfrost_dri.so doesn't get loaded at all (according to strace), however I get MESA-LOADER failed to load driver: exynos
NeuroScr has quit [Quit: NeuroScr]
<memeka>
i added an exynos dri winsys, and now at least it's doing something, but still fails in the end with error 3001 EGL_NOT_INITIALIZED
<alyssa>
memeka: I don't think anyone has done anything with panfrost on exynos, cool!
<HdkR>
Oh, hello memeka
<HdkR>
Welcome to the party, nice seeing you here
<alyssa>
HdkR: Friend of yours?
<HdkR>
They hang out in the odroid channel
<alyssa>
Ah
<alyssa>
memeka: How'd you setup the exynos winsys? Just copied over the rockchip/meson one (and added a stub for it and added it to the gallium-drivers list etc) and changed the name, I guess?
<memeka>
Yes
<memeka>
Should there be anything else?
<memeka>
alyssa: would exynos be more special?
<alyssa>
Nothing comes to mind..
<Lyude>
alyssa: exynos is on one of the boards I've got
<alyssa>
I don't think it would be special, but I don't know what's up with the exynos display drivers. Normal DRI like the others, I hope?
<Lyude>
It's a T6xx though
<Lyude>
also you might need a exynos dri driver for the winsys I'd think?
<HdkR>
I have like 4 XU4 devices for testing but I haven't been doing anything with them
<memeka>
Lyude: that’s the one I got by copying the code from meson/rk
<Lyude>
memeka: that should be good enough!
<Lyude>
(I think)
<memeka>
Lyude: yeah check the logs posted
<memeka>
there's nothing on syscalls that fails tho`
<memeka>
DRM is opened, /dev/mali0 is opened ...
<Lyude>
(I can't help that much ATM I'm just stopping by)
<HdkR>
and I'm just about to eat a late dinner, so boop
<alyssa>
You know what, never mind, this was silly, off to zzz
<memeka>
ok sorry :)
<alyssa>
memeka: Best of luck :) I'll take a look tomorrow if it's not sorted by then
<memeka>
yeah probably not :)
<memeka>
it fails on exynos with a segm fault ...
<Lyude>
Which GPU is this btw?
<memeka>
mali T628
<memeka>
midgard
<memeka>
(gdb) bt
<memeka>
#0 0x00000000 in ?? ()
<memeka>
#1 0x0040120a in swap_buffers () at drm-gbm.c:102
<memeka>
#2 0x0040135a in draw (progress=0) at drm-gbm.c:120
<Lyude>
Ah, ok lol
<Lyude>
There is your problem, we don't support T6xx atm
<memeka>
eeeeh
<memeka>
isn't it midgard? :) what do you support?
<alyssa>
Lyude: I mean, that's also an issue. But it's not the one memeka is currently dealing with
<HdkR>
(This is why I have a ton of XU4 boards. Need to fix support of Midgard 1st gen)
<Lyude>
alyssa: oh?
<alyssa>
Lyude: This is a winsys issue. Panfrost isn't even being _loaded_ for them
<memeka>
i get segmentation fault in gbm_surface_lock_front_buffer
<alyssa>
Anyway, poof
<Lyude>
alyssa: ahhhh
<Lyude>
Well it is still relevant either way
<memeka>
seems to be is being loaded
<memeka>
exynos_dri.so gets loaded, /dev/mali0 gets opened
<memeka>
can't see how /dev/mali0 gets opened w/o panfrost being loaded
<Lyude>
Yeah panfrost is definitely being loaded
<Lyude>
memeka: also, there were some big changes between T6xx and T7xx
<memeka>
i c
<memeka>
well in that example, setup_opengl actually passes
<Lyude>
But the code was originally written for the T6xx, it just hasnt been run on one for a while and some stuff is known to be broken
<memeka>
so eglCreateWindowSurface passes ...
<memeka>
it fails in swapBuffer -> gbm_surface_lock_front_buffer
<memeka>
anw ...
<memeka>
so i will have to wait for T6xx support .... i guess :(
* Lyude
points to "help wanted" sign
<Lyude>
:)
anarsoul has quit [Remote host closed the connection]
<memeka>
also: ./kmscube: symbol lookup error: ./kmscube: undefined symbol: glDrawArrays
anarsoul has joined #panfrost
<HdkR>
memeka: That'll be because it linked against libGL which now doesn't exist for some reason :P
<memeka>
yeah i LD_PRELOAD the panfrost one and got rid of it
<tomeu>
memeka: have you already found why it crashes in gbm_surface_lock_front_buffer ?
<tomeu>
alyssa: maybe not etnaviv, but maybe we could merge with freedreno :p
<memeka>
tomeu: nope
<tomeu>
anyway, I have a lot to learn before I can decide on what would be best
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
<tomeu>
memeka: are you thinking of fixing panfrost to work on that specific gpu?
<memeka>
tomeu: no, too much of a learning curve for my available time
<memeka>
I was hoping to have it working already :)
<tomeu>
ok, then probably it's better for someone else to come along and do it
<memeka>
Maybe Lyude :)
pH5 has joined #panfrost
BenG83 has quit [Ping timeout: 250 seconds]
<memeka>
alyssa: can you add the exynos patch above? At least to get things started? Or do you want to wait for a complete pull request with t6xx working?
raster has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
<raster>
hurr
robclark has joined #panfrost
afaerber has quit [Quit: Leaving]
afaerber has joined #panfrost
<tomeu>
alyssa: do you have somebody to review !20 and !21? some stuff I'm comfortable to review, but some other stuff I'm not familiar with at all and would be better for somebody else to review
<alyssa>
Lyude: To be clear, /dev/mali0 is opened in the winsys, not in panfrost
<alyssa>
tomeu: lol
<alyssa>
memeka: Could you send an MR to gitlab.freedesktop.org/panfrost/mesa? I guess for this it's simple enough that I can just push it but normally we prefer code review through the proper channels
<alyssa>
tomeu: Not at the moment. I'm of the camp code review can be a good way to get acquainted with the codebase (reducing the bus factor) so if you want to take a read, go for it :)
<tomeu>
alyssa: ok, will do tomorrow
<alyssa>
Thank you!
<tomeu>
yw :)
<alyssa>
Lyude: To be fair there's limited bw compat for t7xx to t6xx iirc
<alyssa>
I just have mostly been unenthused about working on fixing regressions on older hw when there's so much to do on the newer stuff (that will backport easily enough when someone/I decide[s] to work on it)
<alyssa>
Note to self: resource_destroy leaks tiled texture memory, since we're freeing cpu but not the entry[0] copy
<raster>
oooh life!
<raster>
any hints on why cbuf[] array ptrs would all be null?
<raster>
ctx->pipe_framebuffer.cbufs <- that array
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
powerbit has quit [Ping timeout: 272 seconds]
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
jernej has joined #panfrost
anarsoul|2 has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
raster has quit [Remote host closed the connection]
raster has joined #panfrost
pH5 has quit [Quit: bye]
stikonas has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
pH5 has joined #panfrost
raster has quit [Remote host closed the connection]
<shadeslayer[m]>
Is there a talk video anyone can recommend if I wanted to learn the basics of gallium 3d
<anarsoul|2>
:)
<anarsoul|2>
there was a talk on NIR
<shadeslayer[m]>
Gonna start getting my hands dirty with panfrost tomorrow
<shadeslayer[m]>
anarsoul|2 uf that sounds way too specific
<shadeslayer[m]>
I'm looking for something more genric
<anarsoul|2>
shadeslayer[m]: well, you'll need it
<shadeslayer[m]>
Yeah, those videos are from 2009
<anarsoul|2>
the biggest part of gpu driver is compiler
<anarsoul|2>
(well, I'm judging by lima, but I assume it's true for other drivers)
<anarsoul|2>
shadeslayer[m]: beside that, read the code and ask #dri-devel if you have questions
<shadeslayer[m]>
Ok
<shadeslayer[m]>
anarsoul|2 also on the agenda for tomorrow is gonna be modesetting with Mali on the pinebook ;)
raster has joined #panfrost
<shadeslayer[m]>
Let's see how that goes
tgall_foo has quit [Quit: My iMac has gone to sleep. ZZZzzz…]
tgall_foo has joined #panfrost
raster has quit [Remote host closed the connection]
AntonioND has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
<ezequielg>
alyssa: tomeu: afbc under testing.
<ezequielg>
black screen.
<anarsoul|2>
:(
<ezequielg>
first thing i noticed is where are we setting the modifier?
<ezequielg>
ah, we can hardcode in kmscube
<ezequielg>
ok hold.
* Lyude
holds
<ezequielg>
I think we need surface_create
<ezequielg>
as per gbm_surface_create_with_modifiers
<ezequielg>
don't we?
<ezequielg>
kmscube complaints about libgbm not supporting modifiers
<ezequielg>
please, stop holding for now :-)
* Lyude
ceases holding
stikonas has quit [Remote host closed the connection]
stikonas has joined #panfrost
<ezequielg>
which is our dri-driver?
<ezequielg>
not so easy to spot
raster has joined #panfrost
<ezequielg>
dri2 probably
BenG83 has joined #panfrost
<ezequielg>
ah, we probably need resource_create_with_modifiers in pan_screen.c
<ezequielg>
"one step closer to knowing"
raster has quit [Read error: Connection reset by peer]
rhyskidd has quit [Quit: rhyskidd]
rhyskidd has joined #panfrost
afaerber has quit [Quit: Leaving]
<TheKit>
in theory, which kernel subsystems DRM display driver is dependant on? I got 4.9 booting on Gemini PDA with PMIC and eMMC drivers, but that's mostly it for now. Getting display and Panfrost later on feels tempting enough
<Lyude>
i keep forgetting how much different arm is with it's average drm architectures...
<Lyude>
TheKit: yeah you will need rockchip or (I'm fairly sure it's ready but might be wrong? I need to give it a shot myself) meson
<Lyude>
alyssa: it's some mediatek chipset iirc
<TheKit>
MediaTek, yes, so no rockchip or meson
<Lyude>
you will probably need your own DRM driver then
<TheKit>
the question was about DRM display
<Lyude>
I'm not sure I follow?
AntonioND has quit [Quit: Quit]
<TheKit>
they have MediaTek DRM driver in mainline for other SoCs, so I wondered what other parts of kernel should be working before even attempting to tinker with display (kind of offtopic though)
<Lyude>
oh! all you should need is mediatek drm stuff, mali_kbase, and a mediatek dri driver for mesa then I believe
<alyssa>
Gotta say, the winsys arch was really not meant for IP like Mali...