05:52Sid127: test failed successfully
05:53tiredchiku[d]: cool
05:53tiredchiku[d]: migrating my znc instance, had to verify :p
05:53tiredchiku[d]: proxies are unfun
09:35notthatclippy[d]: Is there anyone already using nouveau that's willing to run a tiny little experiment (with bpftrace) for me?
09:36notthatclippy[d]: (or alternatively just reboot with a module param)
09:36tiredchiku[d]: shoot
09:37notthatclippy[d]: Which one would you prefer? What I actually want is a list of RPCs sent to GSP when you run `glxgears`. nouveau will print these by default if the debug prints are really really high. Or I can hook it up with bpftrace
09:40tiredchiku[d]: I'm already running nouveau, tell me what you want me to do c:
09:44notthatclippy[d]: Ideally, reload it with `debug=trace`, start it up. See what the last line in dmesg is (or just clear it). Run `glxgears`. Capture dmesg output.
09:44tiredchiku[d]: okay, I'll do that in a few mins and ping you with it
09:44notthatclippy[d]: Thanks a ton!
09:45tiredchiku[d]: just updating to the latest -rc kernel
09:51tiredchiku[d]: do you want me to use nvc0 as the gl driver or can I stick to zink + nvk
09:53notthatclippy[d]: Either is fine.
09:53tiredchiku[d]: okay
09:59tiredchiku[d]: https://cdn.discordapp.com/attachments/1034184951790305330/1304022293164326935/dmesg2.log.gz?ex=672de0e6&is=672c8f66&hm=7d468d6b7ebe423ba26e2b856198445abaae51f48788050135b891e4166f665b&
09:59tiredchiku[d]: notthatclippy[d]
10:01notthatclippy[d]: Amazing! Thanks so much, you saved me an hour at least! I owe you a beverage of choice.
10:04tiredchiku[d]: oh, you're most welcome :D
10:04tiredchiku[d]: least I can do
15:57phodius[d]: im getting error ./src/gallium/drivers/nouveau/nvc0/nvc0_tex.c:143: gm107_create_texture_view: Assertion `!(mt->level[0].pitch & 0x1f)' failed.
15:57phodius[d]: what pitch should I be using?
15:59mhenning[d]: I believe that assert just means the pitch needs to be a multiple of 32
16:59karolherbst[d]: today's episode in "nonsense in codegen": an TXQ with a zero immediate lod source apparently causes issues while a zero gpr source does not 🙃
16:59karolherbst[d]: I wonder if I want to dig deeper or just do a quick one line change
19:23karolherbst[d]: yeah....
19:23karolherbst[d]: TXQ can't take a RZ if it needs an actual parameter 🙃
19:23karolherbst[d]: gfxstrand[d]: not sure if you are aware, but ^^
19:24karolherbst[d]: but I think this is a general tex instruction thing 🙃
19:35gfxstrand[d]: karolherbst[d]: Yup. That's why they have an SSA source type.
19:36karolherbst[d]: I see
19:36gfxstrand[d]: Which is a horrible name. I really need to rework that stuff to make it make more sense
19:36karolherbst[d]: well.. it doesn't help that an TXQ with no actual source can take an RZ 🙃
19:36gfxstrand[d]: GPR means GPR or RZ and SSA means "no really, a GPR".
19:36karolherbst[d]: ahh
19:37gfxstrand[d]: Not great naming
19:37karolherbst[d]: I see
19:37karolherbst[d]: anyway.. I fixed a codegen bug 🙃
19:38karolherbst[d]: it only got hit on TXQ on MSAA textures
19:51gfxstrand[d]: Woo
19:52gfxstrand[d]: Oh, that makes sense. Most of them are vectors anyway and so you always need GPRs. But that one...
19:53karolherbst[d]: it was for TXQ, which just needs a zero lod 🙃
19:54karolherbst[d]: and it ended up with RZ instead of a gpr
20:27gfxstrand[d]: That'll do it
20:27gfxstrand[d]: I think there's a few other cases where you can hit it if you try real hard, like 1D TXF.
20:28gfxstrand[d]: But it's pretty rare outside TXQ