04:17 mareko: do you know who can assign this to Marge? https://gitlab.freedesktop.org/xorg/xserver/-/merge_requests/1612
12:59 Lynne: is there any way or are there any hacks I can do to get the address of a local array in glsl?
13:00 Lynne: I have a function that takes a uint64_t address, and I just need a scratch 32-byte array for a single instance
13:02 Lynne: allocating and assigning a descriptor or push addr for 32 bytes is too much
13:09 pendingchaos: Lynne: no, I don't think there is
13:16 Lynne: sad, guess I'll stuff it into an array and waste a descriptor
13:17 Lynne: whoever designed the buffer reference extension did it just before they left on a friday night with all their friends yelling "come on I won the lottery we're going pub crawling, drinks on me"
13:19 glehmann: the thing is, local arrays might not map to the same hw mechanism as buffer device address
13:20 glehmann: so it's not like this is just an oversight, supporting this case is a ton of work and it might result in sub optimal performance
13:21 karolherbst: global address also means it's accesible by every thread which local arrays defiently are not
13:24 Lynne: I need it accessible by a single thread
13:31 karolherbst: sure, but glsl doesn't have pointer semantics at all, so you might want to rework the function to allow an alternative interface and wrap the one over the other. But anyway, this isn't something which can easily be fixed in glsl.
13:32 karolherbst: CL C has it, and it has 4 distinct memory spaces as a result and an optional tagged pointer feature. it's a lot of work
13:46 glehmann: I think glsl allows arrays as function params, so you can pass local arrays to it and if you deference the bda pointer that can also be passed as an array?
13:47 glehmann: obviously that means you can't do pointer arithmetic in the function tho
13:48 karolherbst: arrays with an unspecified size as well? Because I suspevt that's what's needed here, but if not, yeah could just wrap the bda interface over an array one
14:01 Lynne: nah, you can't pass unsized arrays as arguments
19:28 LL4: ?
21:21 DavidHeidelberg: eric_engestrom: any chance you could get 0f6fa4679d6705660b83579a20fb50cfad2c0184 (clean apply) to 24.1? Just tried to get older build running and clc stuff got me sadomasochistic experience
22:00 eric_engestrom: DavidHeidelberg: it's already there: https://gitlab.freedesktop.org/mesa/mesa/-/commit/db3da0d302a33632a502cd271dfd9487f973bd2d :)
22:00 eric_engestrom: present in the latest release
22:00 eric_engestrom: oh wait no, I jsut re-read your message and you said 24.1
22:00 eric_engestrom: that branch is EOL, there's no new release planned
22:01 eric_engestrom: sorry :]
22:05 Ermine: How should I pull changes for drm-misc tree? Should I use git remote update or will git pull suffice?