#asahi-gpu · 2023-06-27 18:35
alyssajannau: yes, as far as I can see all Mesa drivers#asahi-gpu · 2023-06-27 18:32
alyssawe can.. probably submit CTS with X11. although it feels wrong to do so if we intend to ship wayland..#asahi-gpu · 2023-06-27 18:32
alyssanot Asahi's bug but probably Mesa's#asahi-gpu · 2023-06-27 18:32
alyssaok. one regression (fixed) and one pair of tests that are failing on wayland but not x11#asahi-gpu · 2023-06-27 16:59
alyssathanks dEQP-EGL#asahi-gpu · 2023-06-27 16:59
alyssayay, more fails to work through#asahi-gpu · 2023-06-26 23:09
alyssaunsure about deqp-egl#asahi-gpu · 2023-06-26 23:09
alyssaat any rate my branch should pass {deqp,khr}-gles{2,3,31}#asahi-gpu · 2023-06-26 21:25
alyssaand render target spilling (which I choose to implement with images.. apple does something a lot fancier)#asahi-gpu · 2023-06-26 21:25
alyssashader-writeable images#asahi-gpu · 2023-06-26 21:25
alyssamm, both of those are already upstream#asahi-gpu · 2023-06-26 21:22
alyssaa huge chunk of gles3.1 already landed over the past 6 months#asahi-gpu · 2023-06-26 21:22
alyssawell, a collection of gles3.1 features#asahi-gpu · 2023-06-26 21:21
alyssagles3.1#asahi-gpu · 2023-06-26 21:21
alyssanah, gles.31#asahi-gpu · 2023-06-26 21:20
alyssawhistles#asahi-gpu · 2023-06-26 21:20
alyssa78#asahi-gpu · 2023-06-26 21:20
alyssa$ git log asahi/main...HEAD --oneline | wc -l#asahi-gpu · 2023-06-16 17:41
alyssashould be straightforward to lower with a similar technique to texture atomics, but ugh#asahi-gpu · 2023-06-16 17:09
alyssais annoyed#asahi-gpu · 2023-06-16 17:03
alyssacause if not i'm going to be annoyed#asahi-gpu · 2023-06-16 17:03
alyssaSo does the hardware support multisampled images? >:#asahi-gpu · 2023-06-16 17:03
alyssaUghhhhh#asahi-gpu · 2023-06-16 17:02
alyssaOooooooof#asahi-gpu · 2023-06-16 17:02
alyssa/System/Library/PrivateFrameworks/GPUCompiler.framework/Versions/31001/Libraries/lib/clang/31001.523/include/metal/metal_texture:7168:3: error: static_assert failed due to requirement '_is_any_of((metal::access)2, access::read)' "Multisample textures must have access qualifier access::read"#asahi-gpu · 2023-06-16 17:01
alyssamaybe a metal bug, who knows#asahi-gpu · 2023-06-16 16:58
alyssanot sure what to make of this#asahi-gpu · 2023-06-16 16:56
alyssaisn't it running tets twice?#asahi-gpu · 2023-06-16 16:55
alyssabut how does the store get masked?#asahi-gpu · 2023-06-16 16:55
alyssain principle, this is saying "run depth/stencil tests, then discard" as required by the spec#asahi-gpu · 2023-06-16 16:51
alyssayeah no this seems wrong#asahi-gpu · 2023-06-16 16:51
alyssasample_mask (1<<sample_id, ~0)#asahi-gpu · 2023-06-16 16:51
alyssastore#asahi-gpu · 2023-06-16 16:51
alyssa}#asahi-gpu · 2023-06-16 16:51
alyssa sample_mask(1 << sample_id, 0)#asahi-gpu · 2023-06-16 16:50
alyssaif (discard condition) {#asahi-gpu · 2023-06-16 16:50
alyssasample_mask(1 << sample_id, ~0)#asahi-gpu · 2023-06-16 16:50
alyssaInside that if, it does#asahi-gpu · 2023-06-16 16:50
alyssano this is ok#asahi-gpu · 2023-06-16 16:48
alyssaum this is probably invoking undefined behaviour isn't it lol#asahi-gpu · 2023-06-16 16:47
alyssaEach iteration of the sample loop is wrapped in `if ((1 << sample_id) & sr60l != 0)`#asahi-gpu · 2023-06-16 16:47
alyssa+ early frag#asahi-gpu · 2023-06-16 16:46
alyssafor sample shading + fs side effects + discard..#asahi-gpu · 2023-06-16 16:45
alyssafor sample shading + fs side effects, Apple wraps device_store instructions in `if ((1 << sample_id) & sr60l != 0)`#asahi-gpu · 2023-06-16 16:32
alyssaand with a zero extension has flags 3, second source#asahi-gpu · 2023-06-16 14:51
alyssanot by a lot, but a little#asahi-gpu · 2023-06-16 14:51
alyssaIt's nice, retesting Dolphin and seeing perf has creeped up since last time#asahi-gpu · 2023-06-16 14:51
alyssa--#asahi-gpu · 2023-06-16 14:12
alyssaplease test with Darwinia since clearly I'm no good at testing discard, lolz#asahi-gpu · 2023-06-16 14:11
alyssabecause I decided I'd had enough of CTS crunch time and wanted to write a silly optimization pass just for fun ;)