02:00 <alyssa> jenneron[m]: I think I fixed this a few weeks ago

02:00 <alyssa> MR !18115

02:00 <alyssa> It should've been backported appropriately, though maybe a bad release made it into a distro :(

02:01 <jenneron[m]> ok, i still build mesa myself, so i will just build it from master

02:02 <jenneron[m]> alyssa: i've also found this one https://bpa.st/3YBQ

02:03 <jenneron[m]> i suppose for enabling t628 in upstream mesa, i need to fix performance issue on exynos in upstream kernel?

02:03 <alyssa> er that's a new one

02:04 <alyssa> and no? if it works well enough I'll merge the patch?

02:04 <jenneron[m]> well, it works, but on exynos using upstream kernel it's slower than llvmpipe

02:04 <alyssa> Oh, uff

02:04 <jenneron[m]> using kernel hack (which is merged in postmarketOS) it works fine

02:05 <jenneron[m]> on the other hand, upstream device-trees do not enable GPU

02:05 <jenneron[m]> so they would not be affected

02:06 <jenneron[m]> only samsung tablets (but they don't have display support) and odroid boards (but they use mali ddk in their images) have GPU enabled in DT, on chromebooks it's not enabled

02:09 <alyssa> nod

02:10 <alyssa> if merging the 1-line allowlist patch means you can drop a patch in postmarketOS, i'm all for it :)

02:10 <jenneron[m]> actually we do not provide mesa fork, users have to build it themselves

02:11 <jenneron[m]> so, it would be nice to enable it

02:24 <alyssa> Under good conditions I accept MRs :)

03:02 <jenneron[m]> <alyssa> "jenneron: I think I fixed this a..." <- still happens (current main branch of mesa) https://bpa.st/WZ6Q

06:57 <Daanct12> can anyone look at this issue for me? apitrace is attached

06:57 <Daanct12> https://gitlab.freedesktop.org/mesa/mesa/-/issues/7251

11:55 <iohannes> Hey, so what's the current situation with panfrost on the T604? I'm trying to put linux on a 2013 samsung chromebook; I've got sway running but afaict it's all in software. I can only get an eglInitialize failed error when trying to access the hardware. More of a sysadmin than a coder and never worked in low level graphics, but I'm willing to help out if there's anything I could do. If I wanted to work on the t604, how hard would it be and where woul

13:41 <alyssa> jenneron[m]: ^^

13:43 <jenneron[m]> iohannes: unsupported. it might actually work some time ago, i'm still going to try it and bisect if needed

13:43 <jenneron[m]> i may do that when i receive a new laptop (my samsung chromebook now has a broken display)

13:43 <alyssa> note that t600 is likely to be even more broken than t620

13:43 <alyssa> the first midgard ;)

13:43 <jenneron[m]> if you want, you can try this version and patches yourself https://github.com/quarkscript/linux-armv7-xe303c12-only/tree/master/archlinuxarm/some_forked_apps/mesa, it seems that it worked for them

13:44 <jenneron[m]> alyssa: also, should I send MR to enable T628 or xorg issue has to be fixed firstly?

13:47 <alyssa> xorg issue?

13:48 <jenneron[m]> alyssa https://bpa.st/WZ6Q - the same happens on main branch if mesa built this night

13:49 <jenneron[m]> it doesn't happen on wayland and it didn't happen for me on Alpine with xorg, weird

13:49 <jenneron[m]> of*

13:49 <alyssa> jenneron[m]: `p dev->model`

14:42 <jenneron[m]> alyssa: $1 = (const struct panfrost_model *) 0x0

14:55 <greenjustin> Is that Kukui/Jacuzzi or an Hana/Elm?

14:57 <greenjustin> Oh nvm, daisy. That might work ok then.

15:05 <alyssa> jenneron[m]: is this mesa patched or not patched?

15:07 <jenneron[m]> alyssa: this is patched and installed to /usr/local, gnome + wayland works and recognizes t620 on the same installation

15:07 <alyssa> evidently the patch isn't quite right...

15:08 <alyssa> /* Check if we're loading against a supported GPU model. */

15:08 <alyssa> that's an old mesa

15:08 <alyssa> uprev

15:08 <jenneron[m]> alyssa: https://bpa.st/ICUQ

15:09 <alyssa> jenneron[m]: yes that's how it should look today

15:09 <jenneron[m]> that's what i use

15:09 <alyssa> ack

15:09 <alyssa> that should be ok

15:10 <jenneron[m]> -Dgallium-drivers=panfrost,kmsro,swrast -Dplatforms=x11,wayland -Dopengl=true -Dosmesa=true -Dglx=dri -Dshared-glapi=enabled -Dgles1=disabled -Dgles2=enabled -Degl=enabled -Dvulkan-drivers= -Ddri-drivers= -Dgbm=enabled -Dllvm=false

15:10 <alyssa> kmsro is unnecessary

15:10 <jenneron[m]> do flags look ok? i suppose i could miss something and something might been used not from /usr/local

15:11 <alyssa> I don't set any of those

15:11 <alyssa> -Dgallium-drivers=panfrost -Dvulkan-drivers=

15:11 <alyssa> is how i build

15:12 <jenneron[m]> alyssa: are you sure? as i understand exynos uses mali only for rendering, not for displaying

15:12 <jenneron[m]> and as i understand kmsro is needed for this, but i may be wrong

15:13 <alyssa> 70813c1c13b ("meson: Remove kmsro from gallium-drivers")

15:13 <jenneron[m]> oh

15:13 <alyssa> in gdb, could you `p dev->gpu_id` out of curiosity?

15:28 <jenneron[m]> alyssa: $1 = 1568

15:52 <alyssa> hm

16:59 <robclark> https://www.irccloud.com/pastebin/GZ5m9JAn/

16:59 <robclark> alyssa: ^^^

17:00 <robclark> I guess we don't have perfetto enabled in any of the CI builds

17:01 <robclark> oh, wait.. no.. just panfrost uapi header is br0ken

17:03 <robclark> ahh, looks like https://gitlab.freedesktop.org/mesa/mesa/-/issues/7195

17:04 <robmur01> ...and https://lore.kernel.org/dri-devel/5898d7a6-14c2-8e20-abc3-5971a4bcfd85@arm.com/T/#t

17:07 <robclark> yeah, found that