00:45 <alyssa----> tomeu: Ugh

00:45 <alyssa----> There's a solution, but... it's a PITA and I was hoping to push it off

00:46 <alyssa----> (Namely, we have to "convert" the buffer into AFBC first. So, create a new AFBC backing, draw a textured quad across the surfacce to load the old tiled contents, do your rendering, and when that finishes free the tiled surface and use that as the new pointer)

00:46 <alyssa----> chewitt: Awesome!

00:46 <alyssa----> Lyude: I'd push off scheduling as long as possible

00:47 <alyssa----> chewitt: As for the log, I have no idea what that's about, tomeu ^^?

04:02 <alyssa----> tomeu: FWIW, though, if you force linear (should_tile = false), gnome-shell gets a lot farther

17:57 <alyssa----> Our viewport handling is so wrong, ack

17:57 <alyssa----> That bug is single-handedly responsible for way too many deqp failures

18:21 <anarsoul|2> alyssa----: viewport transformation?

18:38 <fysa> is Ubuntu cosmic w/rk3399 a good environment for testing panfrost?

18:38 <fysa> cosmic unstable

18:40 <fysa> or should I revert back to standard Armbian Bionic due to more commonality with the community?

19:05 <alyssa----> anarsoul|2: Like, how we handle glViewport()

19:05 <alyssa----> fysa: Not sure what people are using..

19:15 <fysa> is kernel 5.0 needed?

19:26 <alyssa----> Depends on which kernel driver you want

22:06 <HdkR> How do people sanely review patches on an ML? :P

22:08 <alyssa----> HdkR: If it shows up in my inbox, I hit reply and vim pops up and I just read through / leave comments like any other mail

22:08 <alyssa----> If it doesn't show up in my inbox, I don't look at it :)

22:08 <HdkR> lol

22:09 <HdkR> I'm so used to the ability to collapse review something that this is...foreign on the review side

22:10 <alyssa----> HdkR: gitlab review is unusable for me soo

22:11 <HdkR> Yea, I just need to figure out how to sanely review in this workflow

22:12 <HdkR> I've only been on the pushing side with ML workflow :D

22:20 <Lyude> HdkR: whatcha reviewing?

22:20 <HdkR> looking at some panfrost patches

22:22 <Lyude> ah

22:23 <HdkR> Since I can now review things! :D

22:30 <alyssa----> HdkR: Huzzah :)

22:41 <HdkR> alyssa----: btw, has Midgard 1st changed behaviour yet?

22:42 <alyssa----> HdkR: Midgard 1st gen is great, because I haven't tested it :)

22:42 <alyssa----> To use robher's choice of phrase

22:42 <HdkR> Well, mostly mean has it differed since I last poked at the XU4 :P

22:43 <HdkR> I presume you're also still using the RockPro64 device?

22:44 <HdkR> need to get a feel for where things are again since I now have time

22:46 <alyssa----> HdkR: I'm using the rockpro for panwrap and the kevin for mesa

22:46 <HdkR> Ah, alright

22:46 <alyssa----> Also, a *lot* of new ALU ops are turning up. Neato squeato.

22:46 <HdkR> :D

22:47 <HdkR> I should probably keep an XU4 "untainted" for wrapping

22:48 <alyssa----> :P

22:48 <alyssa----> HdkR: You got it backwards, I think

22:48 <HdkR> :D

22:48 <alyssa----> If it's possible to wrap on the device, it's tainted

22:49 <HdkR> sure

22:50 <HdkR> alyssa----: Also has piglit running status changed?

22:50 <alyssa----> HdkR: dEQP dEQP dEQP dEQP dEQP dEQP

22:50 <alyssa----> Piglit? What's that?

22:50 <alyssa----> dEQP is what I use :)

22:50 <alyssa----> chrisf: <3

22:50 <HdkR> lol

23:19 <alyssa----> Oh, so there are a set of ball/bany ops for uvecs too (separate from ivecs)

23:19 <alyssa----> So many new ops aaa

23:41 <Lyude> Anyone have any recommendations for guides to get started with understanding nir?

23:41 <Lyude> I understand the concept of ssa and all that, just trying to figure out where to get started with understanding how to write something with it

23:44 <alyssa----> Lyude: There's no docs, silly :P

23:45 <alyssa----> I learned what I know from reading the compilers for vc4/v3d/freedreno-ir3, and the NIR codebase itself of course