00:49 zmike: mareko: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/36634
04:49 sima: airlied, ah
04:49 sima: well I'm already in party-w/e mode, so defo won't do the others this Friday now :-P
10:38 emersion: tzimmermann: what is the difference between boot_vga and the new boot_display sysfs attr?
11:13 tzimmermann: emersion, boot_vga only works with VGA compatibles. boot_display works with any display.
11:13 emersion: nice
11:13 tzimmermann: this was discussed at https://gitlab.freedesktop.org/xorg/lib/libpciaccess/-/merge_requests/37
11:15 emersion: thanks for the link
11:16 emersion: do you think it would make sense to add to libdrm?
11:16 tzimmermann: there was some back and forth about the actual policy. we ended up with exporting the kernel's logic to the userspace. all pieces of the stack should now have the same information
11:18 tzimmermann: emersion, that's a question for dave or mario
11:18 tzimmermann: airlied ^
11:18 emersion: or, udev?
11:18 emersion: i'd like to add support for this in wlroots, but not sure what the best way would be
11:19 tzimmermann: it's in libpciaccess apparently: https://gitlab.freedesktop.org/xorg/lib/libpciaccess/-/merge_requests/39
11:20 tzimmermann: i really don't feel qualified to answer that question
11:20 emersion: libpciaccess is not in my dependency tree
11:29 tzimmermann: i think it would make sense to provide the boot_display flag via libdrm. but IDK how this would best be implemented
11:36 glehmann: jenatali: can I get a Rb for the MS commit? https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37258
14:00 zmike: can I get an ack on https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/37081
14:22 gfxstrand: dcbaker: Nice!
15:22 glehmann: zmike: how is that extension used? If there are no fragment outputs, we we could reduce the fragment shader invocations in each wave to 1 per primitive on amd. Ofc that still not as good as just one globally, so not sure if it's good enough
15:22 zmike: glehmann: I don't really know? it's for minecraft
15:23 zmike: the result is always implementation-defined, so in theory you can do anything
15:27 glehmann: I'm actually surprised that nvidia needs an extension for this, this seems like something the driver can figure out on its own
15:28 zmike: 1 wave per primitive sounds like it could be legit? but also it needs dynamic state support which would be annoying
16:58 karolherbst: Vulkan question: if I call into vkGetPipelineExecutablePropertiesKHR with a compute shader, could I practicalaly get multiple VkPipelineExecutablePropertiesKHR objects with different subgroup sizes?
17:03 alyssa: karolherbst: intel? :D
17:03 karolherbst: no, intel gives me one
17:03 karolherbst: which is neat
17:04 karolherbst: workgroups are part of the shader in vulkan, and so for a given workgroup it gives me a single entry with a single subgroup size, so that's useful for my use case
17:06 glehmann: I think it wouldn't be invalid for the driver to return more than one properties, but in practice I doubt it will happen
17:08 alyssa: yeah..
17:08 alyssa: most of the use case for multiple binaries is only for graphics
17:08 alyssa: I think
17:12 karolherbst: okay, that's good enough for me, because then I won't have to pin the subgroup size and just report whatever the driver is using
17:12 karolherbst: and if a driver reports multiple ones, then I still have a fallback for that
18:34 sghuge: any idea how to launch steam in offline mode from terminal?
19:14 HdkR: sghuge: Change the option once in the GUI to always start in offline mode
19:14 sghuge: HdkR: ACK!
19:15 HdkR: Settings-> Friends & Chat -> "Sign in to friends when Steam starts"