00:09fdobridge: <gfxstrand> Yeah, probably. We probably need to restructure nvk_shader.c a bit as part of it. I've been avoiding it until we have NAK but it can probably be done now if we work in the right direction.
00:10fdobridge: <gfxstrand> There's a lot that's automatic if we use vk_pipeline_cache.h
05:51symbion: Ermine kind of guys who put cones in their ass behind minix shell, don't realize the truth much, as they do not realize that this is not quite the same as beating armed forces of Russian federation, but it's not the first thing you misunderstood, you understand nothing in general, but for abortion leftovers based of your interactions in life, that is just inconvenience to others, it's not a dream experience to deal with such, when Russians
05:51symbion: decide to kill you , this is one way to end such miserable situations.
05:52symbion: especially since you were not very friendly with your domededov bombing either.
05:54symbion: the indirection thought is to stretch a buffer, where the mapping of 3+5=8 becomes unique amongst 4+4 6+2 1+7 3+5 due to shuffling them
05:55symbion: so you'd store the result of alu to that index location and access it with elimination technique
06:06symbion: you can do it in whatever fashion cause after shortcut of the magic value gotten a decimal number, it is no longer performance critical, you just append that alus magic value and that is it
06:14symbion: otherwise yes, you probably find quite performant way to do it with nested loop in the compiler too, it's just for the say DIV alu unit, the events of operand loads do happen in verbatim indexes
06:18symbion: i have not tested but likely 1 2 2 3 3 4 4 5 5 6 6 works too
07:07castrator: value 1 2 3 1 2 3 4 5 6 4 5 6 7 8 9 7 8 9
07:07castrator: index 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
07:07castrator: needed 1+7=1+13=14
07:07castrator: shifts/ 2+6=2+9=11
07:07castrator: transforms 4+4=7+10=17
07:08castrator: such a format suites
08:02Ermine: Lol, I've made this guy angry
10:10fdobridge: <youmukonpaku1337> fuck is up with this guy
10:11fdobridge: <youmukonpaku1337> ~~this is confirming my suspicion that graphics programmers are insane~~ /j
10:50fdobridge: <maba_kalox> Hi, stupid question, how do you set up your working machine to keep nvidia proprietary drivers for normal usage, but for dev use nouveau and NVK? My current point of struggle is nvidia user space driver, which some applications continue trying to use, despite the fact that nvidia kernel space driver is not loaded, e.g if such app is hyprland wm, which tries to use nvidia drm
10:50Ermine: you can try to blacklist nvidia module
10:50fdobridge: <maba_kalox> Hi, stupid question, how do you set up your working machine to keep nvidia proprietary drivers for normal usage, but for dev use nouveau and NVK? My current point of struggle is nvidia user space driver, which some applications continue trying to use, despite the fact that nvidia kernel space driver is not loaded, e.g if such app is hyprland wm, which tries to use nvidia drm. The only solution I found so far - remove nvidia-utils package o
10:51fdobridge: <maba_kalox> Kernel drivers are already not loaded, such app detects nvidia user space line and tries to use it😅
10:52fdobridge: <maba_kalox> Kernel driver is not loaded, some applications detects nvidia user space line and tries to use it😅 (edited)
10:52fdobridge: <maba_kalox> Kernel driver is not loaded, some applications detects nvidia user space libs and tries to use it😅 (edited)
10:53fdobridge: <maba_kalox> Of course that’s such apps fault, but maybe someone faced the same, and found a workaround
10:55fdobridge: <![NVK Whacker] Echo (she) 🇱🇹> I'll try Hyprland later
11:18fdobridge: <georgeouzou> I do not think Nvidia user space libs will work without the kernel modules loaded
11:29fdobridge: <georgeouzou> Maybe this will help:
11:29fdobridge: <georgeouzou> https://wiki.archlinux.org/title/nouveau
11:33fdobridge: <georgeouzou> https://download.nvidia.com/XFree86/Linux-x86_64/319.12/README/commonproblems.html#nouveau
11:43fdobridge: <maba_kalox> That’s the problem, they sugfault😂
11:43fdobridge: <maba_kalox> That’s the problem, they coredump😂 (edited)
12:36Ermine: You can also try installing another small linux distro nearby, where you'll have only nouveau
15:35fdobridge: <gfxstrand> I don't. 😂 I have the blob driver installed on a different box for the odd bit of testing. I play games on my PlayStation or Switch.
15:38fdobridge: <gfxstrand> Most of my daily work is done on an Intel laptop.
15:38fdobridge: <gfxstrand> I hope to switch to an NVIDIA laptop one of these days but they're still huge. 😢
15:38fdobridge: <gfxstrand> I hope to switch to an NVIDIA laptop one of these days but they're all huge. 😢 (edited)
15:40fdobridge: <maba_kalox> Okay, stupid question, how is task distribution happens in NVK specifically and open source in general? A bit to used to have Jira tickets and having project manager, I need hand of help with it😂
15:40fdobridge: <maba_kalox> Okay, stupid question, how is task distribution happens in NVK specifically and open source in general? A bit too used to have Jira tickets and having project manager, I need hand of help with it😂 (edited)
15:44RSpliet: maba_kalox: there's historically been a reluctance to tell community members what to do; it's their time, can't claim their time if you're not paying for it
15:45RSpliet: best advice I can give you is to find an itch to scratch. Find something that doesn't work that you want to work and start chipping away at it.
15:45RSpliet: Which can be a frustratingly slow process at first, but that's unfortunately how you learn these sorts of things
15:46RSpliet: Back in the old old days there used to be a public Trello board with inspiration for the kind of itches that you might experience. Think that went out of vogue a while ago
15:47RSpliet: by which I mean: whatever might still be on there is outdated and possibly no longer relevant
15:54fdobridge: <maba_kalox> Eh, that’s sounded like a good starting point 😅
16:43fdobridge: <gfxstrand> At one point, I was filing issues for missing features. There might still be some on the nouveau/mesa issue tracker but that's decidedly less convenient now that we're upstream.
16:44fdobridge: <gfxstrand> The reality is that most of the roadmap is in my brain.
16:44fdobridge: <gfxstrand> The reality is that most of the roadmap for NVK is in my brain. (edited)
16:45fdobridge: <gfxstrand> That's not a particularly accessible place to store a roadmap but it's pretty well-organized in there. 😅
16:46fdobridge: <gfxstrand> Really, the best thing to do is probably to look at docs/features.txt, find something that's not done yet and looks interesting and then ask.
16:47fdobridge: <gfxstrand> Ideally, I'd have spent the time to JIRA it all in GitLab issues but I haven't. 🙃
16:48fdobridge: <pixelcluster> yeah IME the project management usually consists of "ask a maintainer if someone else is already doing the thing else do it"
16:48fdobridge: <gfxstrand> Finding and fixing bugs is always good, too.
16:49fdobridge: <gfxstrand> Unfortunately, it's often not at all clear which bugs are reasonable starter tasks.
16:53fdobridge: <gfxstrand> Unfortunately, GitLab doesn't really have a concept of a dependency graph of issues. It cross-links issues when one mentions another but it's kinda crappy. 🫤
16:53fdobridge: <gfxstrand> https://gitlab.com/gitlab-org/gitlab/-/issues/212329
17:08fdobridge: <maba_kalox> Okay, then need to develop brain query language 😂
17:08fdobridge: <maba_kalox> Sure, thanks
17:10fdobridge: <gfxstrand> I would like to do *something* to PM NVK properly. I just need to figure out what and go do it.
17:10fdobridge: <gfxstrand> We do have tracker issues for Zink and DXVK and those help a bit.
19:52fdobridge: <airlied> We trelloed radv for a while pre gitlab, I won't say it was a win but it wasn't a total loss either
20:31fdobridge: <gfxstrand> I think I'd rather use GitLab than some separate thing. I just need to decide how I want to organize it. I don't really like the checkmark thing as you have to manually check stuff off.
20:35fdobridge: <gfxstrand> I should play around with the nouveau issue tracker a bit to figure it what I want and then file stuff
21:02fdobridge: <gfxstrand> I think Ken used Trello for iris for a while.
22:53fdobridge: <gfxstrand> https://gitlab.freedesktop.org/mesa/mesa/-/issues/9610
22:53fdobridge: <gfxstrand> I've broken it down a bit for you. Feel free to ask questions if you have any.
22:59fdobridge: <phomes> that is very helpful. Thank you
23:00fdobridge: <phomes> I have one question now. Do we want to cache both the shader and nir?
23:01fdobridge: <phomes> I will probably have more questions when I start to serialize
23:45fdobridge: <gfxstrand> If we link, yes. We probably want to cache before we apply any pipeline state.
23:45fdobridge: <gfxstrand> PSA: NVK has been project managed:
23:45fdobridge: <gfxstrand> https://gitlab.freedesktop.org/mesa/mesa/-/issues/?sort=updated_desc&state=opened&label_name%5B%5D=NVK&label_name%5B%5D=feature&first_page_size=20
23:46fdobridge: <gfxstrand> Trackers can be found here: https://gitlab.freedesktop.org/mesa/mesa/-/issues/?sort=updated_desc&state=opened&label_name%5B%5D=NVK&label_name%5B%5D=tracker&first_page_size=20
23:46fdobridge: <gfxstrand> If anyone wants to claim a feature, assign it to yourself. If you can't do that, comment on it and ask me to assign and I'll make you a Reporter on the project and assign it to you.
23:49fdobridge: <gfxstrand> @asdqueerfromeu Feel free to update your tracker issues and make sure all the feature issues are properly linked.