10:53 fdobridge: <k​arolherbst🐧🦀> @gfxstrand fyi https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25265
10:54 fdobridge: <k​arolherbst🐧🦀> I solved the static inline wrapper nonsense 🙃
10:54 fdobridge: <k​arolherbst🐧🦀> well..
10:54 fdobridge: <k​arolherbst🐧🦀> bindgen did, I just integrated it into meson+mesa
14:02 fdobridge: <g​fxstrand> 🥳
14:02 fdobridge: <g​fxstrand> I can probably simplify a bunch of my nir wrappers now
14:08 fdobridge: <k​arolherbst🐧🦀> there is other good stuff in bindgen 0.64: `Added a new set of flags --with-derive-custom, --with-derive-custom-struct, --with-derive-custom-enum and --with-derive-custom-enum to add custom derives from the CLI.`
17:45 fdobridge: <g​fxstrand> Oh, nice!
17:46 fdobridge: <g​fxstrand> I really hate that NV doesn't have an easy way to get some locations. 🤦🏻‍♀️
17:48 fdobridge: <g​fxstrand> I really want to lower it in NIR but it has to happen pretty late.
17:49 fdobridge: <g​fxstrand> Doesn't work well with the descriptor lowering model. 🫤
17:49 fdobridge: <g​fxstrand> Hrm... actually...
18:30 fdobridge: <m​henning> which locations? not sure I'm aware of this edge case
19:05 fdobridge: <g​fxstrand> For `interpolateAtSample()`
19:05 fdobridge: <g​fxstrand> Or if there are, codegen doesn't know about them.
19:05 fdobridge: <g​fxstrand> I could see what the blob does, I suppose. 🤷🏻‍♀️
19:07 fdobridge: <k​arolherbst🐧🦀> you mean sample locations?
19:07 fdobridge: <g​fxstrand> Yes, those
19:07 fdobridge: <k​arolherbst🐧🦀> there is a sysval afaik
19:08 fdobridge: <k​arolherbst🐧🦀> or uhm...
19:08 fdobridge: <k​arolherbst🐧🦀> a tex query?
19:08 fdobridge: <k​arolherbst🐧🦀> something was there..
19:08 fdobridge: <g​fxstrand> Oh... I forgot about the tex query... maybe? 🤔
19:09 fdobridge: <k​arolherbst🐧🦀> yeah..
19:09 fdobridge: <k​arolherbst🐧🦀> sample index within ms pattern
19:09 fdobridge: <k​arolherbst🐧🦀> writes to R
19:09 fdobridge: <k​arolherbst🐧🦀> 4.12/4.12 format
19:10 fdobridge: <k​arolherbst🐧🦀> as packed dx/dy
19:10 fdobridge: <g​fxstrand> There is TXQ.SAMPLER_POS
19:10 fdobridge: <k​arolherbst🐧🦀> yeah
19:10 fdobridge: <k​arolherbst🐧🦀> that's the thing I described
19:11 fdobridge: <g​fxstrand> Oh, nice! So I canjust do TXQ.SAMPLER_POS with the index as the source and get a packed offset? That's nifty.
19:11 fdobridge: <k​arolherbst🐧🦀> IPA/PIXLD use the same format
19:11 fdobridge: <k​arolherbst🐧🦀> so I guess you can just use those values
19:11 fdobridge: <k​arolherbst🐧🦀> yeah
19:11 fdobridge: <g​fxstrand> Neat!
20:16 fdobridge: <g​fxstrand> Do you know what the arguments you TXQ are supposed to look like? I'm guessing `(0, ms_idx)` but IDK.
22:41 fdobridge: <g​fxstrand> `Pass: 405083, Fail: 192, Crash: 131, Skip: 3195085, Timeout: 2, Flake: 390, Duration: 1:48:52`
22:41 fdobridge: <g​fxstrand> There's some dmesg errors, though.
23:08 fdobridge: <s​amantas5855> many skips
23:22 fdobridge: <g​fxstrand> Yeah, that's to be expected
23:22 fdobridge: <g​fxstrand> Even a full-featured driver with ESO and ray-tracing only runs like 1/2 the CTS
23:28 fdobridge: <k​arolherbst🐧🦀> uhh.. it's just the coordinates
23:28 fdobridge: <k​arolherbst🐧🦀> ehh wait
23:29 fdobridge: <k​arolherbst🐧🦀> ehh yeah, it's the sample index
23:29 fdobridge: <k​arolherbst🐧🦀> but I only know that it's one input
23:30 fdobridge: <k​arolherbst🐧🦀> ahh yeah.. x is sampler|header and y is LOD/sample position
23:30 fdobridge: <k​arolherbst🐧🦀> but
23:30 fdobridge: <k​arolherbst🐧🦀> if you don't specify/need the sampler/header, the LOD/sample moves to x