01:45 karolherbst[d]: oof
01:47 karolherbst[d]: I could check which games looks the most impacted here
02:49 karolherbst[d]: mhh yeah there are a couple of games that are up to 5% impacted, so not surprising to see some of those perf benefits
04:04 mhenning[d]: marysaka[d]: want to review the kernel/mesa patches?
04:04 mangodev[d]: karolherbst[d]: what about "part 1 out of 5" 😟
04:05 mangodev[d]: the *smallest* is <= 5%?
04:06 mangodev[d]: wondering what the "big" optimizations are gonna be if this is "part 1"
05:07 airlied[d]: marysaka[d]: I had another idea, I hacked kernel to delay unref in the free path and see a fault, that I don't think is right. I'll track it down more tomorrow, today is out of time
08:46 marysaka[d]: mhenning[d]: sorry was feeling sick (and still am not better atm), going to add my rb on the mesa patch, for the kernel patch I think this is not right we cannot downgrade the version like that... instead we should the next time we enable it add a new getparam query to indicate support for it (and maybe make that a bitflag so we could check for other features in the future? cc _lyude[d] and
08:46 marysaka[d]: airlied[d] on that)
08:47 marysaka[d]: airlied[d]: I see thanks a lot for looking into it again
08:47 mohamexiety[d]: could also make the version check in mesa check for 1.4.2 instead of 1.4.1 or such
08:48 marysaka[d]: tbh it's kind of silly if we have to disable it again
08:48 marysaka[d]: so just in case? idk
08:48 marysaka[d]: both work tbh
14:27 karolherbst[d]: mhhh `LD` can take 32 bit offsets 🙃
14:27 karolherbst[d]: which is.. interesting..
14:27 karolherbst[d]: because `LDG` and `LDL` only take 24 bits
14:28 karolherbst[d]: (and `LDS`)
14:28 karolherbst[d]: but only if the non uniform reg form is chosen..
14:28 karolherbst[d]: oh well.. ignoring it for no
14:28 karolherbst[d]: w
15:36 _lyude[d]: marysaka[d]: You and me both, I also got sick 🙁
15:36 _lyude[d]: Bumping the version again (though honestly I'd even be fine with downgrading the version but I don't know if there is a specific reason you said we can't do that) seems like the best bet
16:00 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1465738254715457864/rn_image_picker_lib_temp_2989cca0-674c-4347-8b5f-223f591cc14a.jpg?ex=697a32a0&is=6978e120&hm=925372ce5ce043a98a5da400baec7a81d9d1619db73a127de17eccc51ebda1f4&
16:00 mangodev[d]: was nvk tested with the commit timing mr before it was merged?
16:01 mangodev[d]: sddm just…
16:01 mangodev[d]: doesn't start
16:03 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1465739065310707762/rn_image_picker_lib_temp_3fc99090-69a1-470e-b070-4254c1f2f1cd.jpg?ex=697a3361&is=6978e1e1&hm=e6eee437070f44666ddef96e8a8434659a21a353122278a293358b7c2f98388a&
16:03 mangodev[d]: interesting
16:03 mangodev[d]: the modeset fails
16:07 mangodev[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1465739899205451947/rn_image_picker_lib_temp_400511a1-dcc7-459a-a5b0-444934eebe51.jpg?ex=697a3428&is=6978e2a8&hm=53f39a500d7cb8f71e2b6b41b2c8edd685ba24ba8e9b550c6f3a8e65139b227a&
16:07 mangodev[d]: for now, i'm downgrading to before the display timing was merged
16:07 mangodev[d]: because doing that made it work
16:07 mangodev[d]: (yes i don't have a wallpaper here, that's not nouveau's doing)
16:31 phomes_[d]: it tested it in meson devenv a while ago without problems. But not as the installed driver for the system
16:34 mangodev[d]: yeah, this is installed through mesa-tkg
16:35 mangodev[d]: tbf this is mesa-git on arch kernel
16:35 mangodev[d]: was the present timing work partly kernel-side?
16:46 phomes_[d]: not as far as I remember. Hans-Kristian wrote a test app that you can try: https://github.com/Themaister/Granite/commit/f4ae6705271c354055c6f48e1660b68ad761ad74
17:11 mohamexiety[d]: looks like for the CI failures in https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39409 it's a CTS bug. the fix is in the upstream public repo but isn't in a release yet. how does this normally get handled?
17:11 zmike[d]: you can add a manual patch to ci and bump the container to rebuild cts
17:13 mhenning[d]: @mango Could you test https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/39559 and see if SDDM works with it?
17:14 mhenning[d]: mangodev[d]: ^
17:14 anholt: mohamexiety[d]: build-deqp.sh is the file to look at, and it has instructions at the top about how to trigger the rebuild.
17:15 mangodev[d]: mhenning[d]: could definitely try later today
17:21 mohamexiety[d]: gotcha, thanks!