14:45 <steev> oh, i didn't realize panfrost had their own channel :)

14:46 <steev> alyssa: there is other "good" news.... the youtube ad to purchase youtube music or whatever they call it.... rendered perfectly even though the rest of the site doesn't... heh

14:48 <alyssa> giggle

14:52 <steev> you're right though - i noticed some other things - e.g. in my sway setup, when i hit print screen, i run a script and it basically uses rofi to show a menu to choose full screen/app/whatever - and in mesa < 22, it was "blank" aside from whatever the selected option was, whereas with 22, it actually does show the menu

15:08 <steev> alyssa: alacritty is on the right... i have no idea what it's doing though. the config and such worked fine with 21.3.3 https://usercontent.irccloud-cdn.com/file/tTYYDs2x/2022-01-10-090514_screenshot.png

15:35 <steev> alyssa: ah, looks like someone else is having the same/similar issues - https://gitlab.freedesktop.org/mesa/mesa/-/issues/5594

15:38 <alyssa> right that ticket

15:45 <steev> fwiw, i'm running 5.15.13 on that box (haven't tried 5.16 yet)

18:10 <cphealy> Do Mali GPUs have one or two sets of HW perf counters? (I recall there being an embedded GPU vendor that has two sets but not sure if it's ARM Mali or other vendor.)

18:11 <alyssa> Yes

18:13 <cphealy> Does PF use both sets or is one set available for other users outside of Mesa?

18:15 <cphealy> The use case I have in mind is for kernel driver to expose one set via perf PMU events such that one can monitor GPU read/write bytes on bus use kernel perf API independently from Mesa usage.

18:18 <robmur01> IIRC from when I briefly looked at it, the way that Midgard/Bifrost performance counters work didn't seem conducive to doing what the perf API really wants