11:35 <MoeIcenowy> simon-perretta-img: for hardcoded shaders, I think they might be generic enough, so okayish to exist now?

11:38 <MoeIcenowy> and for things like idfwdf, is there any way to describe it in a vendor-neutral way? add some nir intrinsics?

11:58 <simon-perretta-img> MoeIcenowy: Upstream have said from when we started contributing that they explicitly don't want hardcoded shaders/binary blobs/etc. merged, so we can't merge those

11:59 <MoeIcenowy> okay

11:59 <simon-perretta-img> I've separated out the hardcoded things into a single commit which I'll stick onto dev/bxs in the meantime though, so that can locally be cherrypicked onto main while the unhardcoding work is ongoing

12:02 <simon-perretta-img> idfwdf is pretty much just a memory store + texture reads so it shouldn't be too bad; we're aiming to have the unhardcoded bits completed and merged prior to rc4 so they can make it into 25.2

12:03 <MoeIcenowy> oh looks like clc can just use nir intrinics?

12:04 <MoeIcenowy> this sounds good

12:04 <simon-perretta-img> Yeah, it's pretty neat

12:12 <simon-perretta-img> Some shaders are still going to stay as NIR for now, as I've not yet managed to figure out if there's either a non-insane way to use load_input/store_output intrinsics with clc for fs/vs, or if there's an existing glsl precomp tool similar to clc (or something else entirely)