03:19 gfxstrand: dakr: In future, if you're going to CC me on patches, could you please use faith@gfxstrand.net? I prefer using that one over my Collabor e-mail for that sort of thing.
03:20 gfxstrand: It's okay if you don't remember. I can respond from either. That's just where I subscribe to most mailing lists.
11:29 dakr: gfxstrand: sure!
15:31 fdobridge: <g​fxstrand> PSA: Annoying nak rebase just landed in nak/main.
18:22 fdobridge: <g​fxstrand> And, we have XFB! That wasn't too hard.
18:25 fdobridge: <g​fxstrand> Ugh... txf broke in the rebase
18:37 fdobridge: <g​fxstrand> Yup, I messed up the rebase on top of shader cache. 🤦🏻‍♀️
18:49 fdobridge: <!​[NVK Whacker] Echo (she) 🇱🇹> EXT_map_memory_placed is now in the NVIDIA Vulkan beta driver (so :triangle_nvk: should probably support it too)
18:53 fdobridge: <g​fxstrand> Yeah. The extension hasn't actually shipped yet so we can't really pull it into meas/main
18:53 fdobridge: <g​fxstrand> I still haven't written CTS tests
21:03 fdobridge: <g​fxstrand> I just pushed all my VS patches to nak/main
21:03 fdobridge: <g​fxstrand> The only remaining VS fails there are because I haven't figured out what to do about unaligned vector ALd/ASt yet
21:29 fdobridge: <k​arolherbst🐧🦀> why would that happen anyway?
21:29 fdobridge: <k​arolherbst🐧🦀> or do you mean vectorized ones?
21:29 fdobridge: <g​fxstrand> Yeah, right now I'm just using the same vectorization as NIR gives me
21:30 fdobridge: <g​fxstrand> Sometimes that means a vec3 starting at .y
21:30 fdobridge: <k​arolherbst🐧🦀> yeah.. well.. tell nir to not do that 😛
21:30 fdobridge: <k​arolherbst🐧🦀> all load/stores are always vector size aligned
21:30 fdobridge: <g​fxstrand> Well, yeah. I just haven't bothered to write/find the code
21:31 fdobridge: <k​arolherbst🐧🦀> or just lower it with that new fancy pass we got
21:31 fdobridge: <k​arolherbst🐧🦀> for a while now I think
21:31 fdobridge: <g​fxstrand> Pretty sure that one only works on explicit I/O intrinsics
21:32 fdobridge: <k​arolherbst🐧🦀> mhhh
21:33 fdobridge: <e​sdrastarsis> vkcube is working :ferris_happy:
21:34 fdobridge: <g​fxstrand> It ought to. 99.9% of the CTS works.
21:34 fdobridge: <e​sdrastarsis> at 100 fps and on codegen it runs at 900 fps
21:35 gfxstrand: Sure
21:35 fdobridge: <g​fxstrand> We'll get to fixing that
21:36 fdobridge: <a​irlied> probably don't make it default to on until after that, so the first phoronix article isn't all negative 😛
21:36 fdobridge: <g​fxstrand> 🤷🏻‍♀️
21:40 fdobridge: <e​sdrastarsis> I know, but all this progress in such a short time is exciting 🙂
21:43 fdobridge: <g​fxstrand> If I had to take a blind guess, it's probably something with my 16 GPR minimum causing lower occupancy but IDK.