09:15glehmann: do we have a nir pass that removes single element array variables/derefs?
09:49karolherbst: and turns them into scalars?
09:55glehmann: yeah turing something: like shared uint g4[1]; into a non array: shared uint g4;
09:56glehmann: I want to write a pass that turn those simple shared variables into subgroup ops
09:57glehmann: but dxvk declares everything as an array, so it would be nice to get rid of those first
09:58glehmann: my new pass has to operate on vars/derefs to avoid allocating space for the variables that are removed
09:58karolherbst: mhhhhh
10:01karolherbst: I doubt we have that, and I can see how such a pass could also become messy real quick
10:05glehmann: I mean, we have worse opts that deal with derefs
10:06glehmann: doing anything with variables is a nightmare in nir tbh
12:16zmike: mareko: ping on https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36296
16:34mattst88: Mesa CI busted?
16:34mattst88: https://gitlab.freedesktop.org/mesa/mesa/-/jobs/82674639
16:35mattst88: [00:54] ERROR: prepare-artifacts: ret code: 2
16:35mattst88: [00:54] ERROR: unknown-section: ret code: 2
16:38mattst88: looks like it's known: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36807#note_3057093
16:44sylware: Hi, the less invasive fix of this bug is to disable all gcc optimization and everything is ok again. Namely, I force -O0 and the c++ constructor optimization bug is gone.
18:22robher: Can someone update drm-misc-next to put it in linux-next? It appears it has not switched over after rc1...
18:30anholt: Kayden: could you push whatever you're willing to, to https://gitlab.freedesktop.org/gfx-ci/private-access/shader-db-private ? Then I'll follow up with adding our stuff.
19:56Kayden: anholt: sure, I can do that. do we want to be using LFS? some of these are rather large (couple recent ones are 50MB, some older ones are 150-200M)
19:56anholt: yeah, I think LFS is the way to go.
19:56anholt: (my reasoning in our internal repo being: If we ever regenerate fossils, nobody wants to download the old ones by default, since nobody will go digging in git history)
19:58Kayden: yep, makes sense :)
20:41Kayden: anholt: pushed
20:41anholt: Kayden: \o/