04:50 <lina> asahi/main already gets regularly rebased on top of mesa/main ^^

20:09 <alyssa> GNOME doesn't want to start on my latest rebase, gg

20:16 <alyssa> no deqp-gles2 regressions

20:29 <alyssa> works fine in a release build..

20:30 <alyssa> lod_bias assert going off. Great.

20:56 <alyssa> doesn't look like lod bias is implemented in Apple's driver but there's a chance it's still supported in (deprecated?) hw

20:56 <alyssa> time to mash random bits

20:57 <alyssa> "If the VK_KHR_portability_subset extension is enabled, and VkPhysicalDevicePortabilitySubsetFeaturesKHR::samplerMipLodBias is VK_FALSE, mipLodBias must be zero"

20:57 <alyssa> ah yes I forgot that MoltenVK gets to cheat.

21:08 <alyssa> mashing bits isn't finding anything yet..

21:18 <TellowKrinkle> Isn't lod bias part of GL 4.1? Do they just ignore it?

21:19 <alyssa> TellowKrinkle: Apple's GL driver implements every single texture op as textureBias with the bias from a uniform register

21:19 <alyssa> now I know why :|

21:21 <TellowKrinkle> Well that's one way to do it

21:21 <TellowKrinkle> (That's what Dolphin does in the MVK and Metal renderers)

21:22 <alyssa> Delight

21:22 <alyssa> the perf impact is probably not *terrible* but it would still suck

21:23 <alyssa> although... flipping sampler descriptor bits is definitely causing tests to regress related to LOD so

21:23 <alyssa> hm

22:26 <alyssa> ----

22:26 <alyssa> Pushed the latest asahi/main

22:26 <alyssa> You know the drill