03:29 <starnight> Hi! We have a khadas VIM2 board (S912, Mali-T820). We tried Panfrost on this board, but screen shows artifacts. I have placed a ticket at https://gitlab.freedesktop.org/mesa/mesa/-/issues/3143

04:05 <chewitt> interesting to see @icecream95 talking about power issues potentially being a source of issues that cause artefacts

04:06 <chewitt> I'm wondering if the same underlying "how the GPU is powered" issue we see on Amlogic G31/G52 is also present on T820

04:10 <chewitt> The current power-up code @narmstrong wrote appears to work well on T820, but due to the original lack of Linux mali blobs for T820 the number of people trying to run Linux on GXM devices is probably more limited than say Rockchip boards, so maybe there are still issues but just not seen in enough volume to trigger recognition

07:03 <tomeu> chrisf: I merge this fragment into the arm64 defconfig: https://paste.debian.net/1154318/

07:03 <tomeu> but mmind00 might have a better branch/instructions?

07:10 <mmind00> chrisf: tomeu: don't really have ready instructions ... but looking at my scratch branch, mainline (5.8-rc2 to be specific) only seems to be missing tomeu 's bifrost enablement patches ... other than that, I guess just use 5.8-something + the config fragment from above

07:27 <tomeu> chrisf: yeah, there shouldn't be much about it

07:28 <tomeu> do you get 3 devices in /dev/dri?

07:30 <narmstrong> chewitt: theorically, T820 also need the power-up/reset code

07:35 <tomeu> ah, interesting!

10:09 <icecream95> Panfrost vkcube trace: https://gitlab.freedesktop.org/snippets/1085/raw It doesn't actually draw anything yet, but at least it doesn't crash...

10:09 <HdkR> =o

10:20 <tomeu> almost there :)

14:22 <alyssa> :o

14:57 <alyssa> code for that pushed anywhere btw?

19:39 <alyssa> Something smells funny about writeout deps

19:39 <alyssa> I recall there being issues here and we used to workaround in the compiler

19:39 <alyssa> but I was able to coax the blob into generating asm that is -- according to our old model of the hw -- invalid

19:39 <alyssa> something fishy

22:12 <chrisf> tomeu: still having trouble. i built your slightly hacked 5.7 (on the device), installed it [with `make install`, `make modules_install`] and things look fairly sensible, but it doesn't boot

22:13 <alyssa> chrisf: doesn't boot how?

22:13 <alyssa> also found my issue -- r1 is special, derp

22:14 <chrisf> appears to be stuck somewhere very early, possibly still in uboot. i dont get any logs

22:14 <chrisf> still waiting on parts for serial console

22:18 <alyssa> chrisf: nothing comes on screen?

22:18 <chrisf> alyssa: the uboot splash never goes away

22:19 <alyssa> Okay

22:20 <alyssa> I'd blame the uboot ini(?) file, there's a lot of things that can go wrong there iirc

23:59 <chrisf> alyssa: no doubt, but very hard to figure out what is going wrong