00:45 cwabbott: dj-death: whenever konstantin gets around to rebasing and merging the radv thing
00:48 cwabbott: dj-death: but what about something like, a closest hit shader A calls TraceRay, it calls another closest hit shader B, then it reads the ray origin
00:48 cwabbott: from what I can tell A's resume shader will read the same memory as B does
00:49 cwabbott: to get it to work you'd need an arbitrarily deep stack of hit info, or you'd have to move the intrinsics up to the top (so that they get spilled onto the user's stack)
00:49 cwabbott: separate structures for setup and hit info doesn't cut it
02:28 DavidHeidelberg: cleanup: would anyone object dropping i915g label and put everything under i915 as we don't have classic X gallium?
04:40 mareko: jenatali: can WSL2 use WARP? (the software rasterizer)
04:40 jenatali: Unfortunately no
04:41 jenatali: I'd started playing around with getting it compiling at one point, WARP used to work on Android back in the day, but the platform abstraction got leaky over time
04:46 mareko: jenatali: I was wondering whether I could make WARP work as a normal DX12 driver, then WSL2 could use it
04:47 jenatali: mareko: the problem is that the UMD needs to be projected into the WSL2 VM as a Linux .so
04:47 mareko: I see
04:47 jenatali: For WARP, that .so also needs custom sync so that it can interact with the OS scheduler across that virtualization boundary
04:48 jenatali: Normally there's Win32 events in the mix which would need to be converted
04:49 jenatali: Doable for sure, just didn't seem worth it when llvmpipe/lavapipe/swiftshader were already there
04:49 jenatali: mareko: is this just for testing the Mesa d3d12 driver? Or are you thinking a productization scenario?
04:52 mareko: jenatali: just testing to fix d3d12 for https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/31942
04:53 jenatali: Got it, yeah
12:54 pandoracast: now for the first algorithm that is bit is inferior, i did not upload the single element extraction procedure. 883+883(−213−210-141 aka 141-141-141)-1024-115=204 and yet more 141=63 and the 3*3 we got added is 72, works generically again, such formula for this. But the maximum only optimization is faster. And your trash i do not want to see along my hotels. We slaughter yet more couple
12:54 pandoracast: people for demonstration, for my fans i say they tied an anchor to me, assaulted and humiliated me for 2.5 years, some of them we already killed some we kill more.
14:20 alyssa: jenatali: ~~Vkd3d~~ delet
14:43 HdkR: Fix all your problems in one easy step!
14:50 ccr: "one weird trick your fellow developers are trying to hide from you"
16:39 FireBurn: I noticed VAAPI accellerated vidoes are just playing green on the main branch I reverted back to 2089bf7b571f712500c470208506a183fcaf39da to fix it, I'll bisect later when I have time
17:19 nowrep: 152b06acd8e broke vulkan video, but vaapi should work
17:29 FireBurn: I use QMplay2 it defo has support for vulkan video, can't remember which is set, will check later, is there a bug open for that breakage already?
17:31 nowrep: no, but fix is here https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32414