01:01 kode54: oh, I didn't realize the descriptor and other such things would result in such heavy stalling
09:15 x512[m]: Do new Nvidia GPUs support legacy 2 bit cursors with and-or masks?
09:48 kode54: let me check drm_info for my cursor planes
09:49 kode54: drm_info says only ARGB8888 for cursor planes
09:49 kode54: but I don't know what VGA level features they support, if any
09:51 kode54: oh
09:51 kode54: crap, I forgot this was #nouveau, I thought I was looking at #wlroots
09:52 kode54: I can report what nouveau supports on this machine later, but I'm kind of compiling Chromium right now
13:31 tuur: Heya, I am generally looking to contribute to important Linux projects. I'm a C++ programmer but haven't yet done any driver development. Forgive the perhaps dumb question, but does development on Nouveau put my GPU at risk of bricking if I mess something up?
13:31 ahuillet[d]: It doesn't
13:36 tuur: Very good to know, thank you.
17:05 hatfield: anyone have a TU102 (Titan or 2080TI) to test with? I am getting 10fps in CS2 on Titan when i switch to use NVK/nouveau, saw someone with RTX 2060 getting 60fps so im not sure if its just a problem with my installation
17:06 mhenning[d]: hatfield: what distro and kernel version are you on?
17:07 karolherbst[d]: random thought: maybe some GPUs boot with a different perf profile as the default one depending on the vbios...
17:07 hatfield: Ubuntu 24.04.3 LTS 6.14.0-36-generic
17:09 karolherbst[d]: is that before GSP was made default or somehting?
17:09 mhenning[d]: hatfield: Okay, you probably don't have GSP enabled by default then. Try booting with nouveau.config=NvGspRm=1 on the kernel command line
17:09 pjb[d]: Hi, sorry if I'm barging in, I have a question and this is the only place I suspect I might get a straightforward answer: Nvidia advertises anti-aliasing counts of up to 32 on OpenGL, but not Vulkan. Is it true, as I suspect based on what I've been able to find, that these are "fake" and not just MSAA, but some FXAA blurring or similar too?
17:10 karolherbst[d]: pjb[d]: yeah
17:10 pjb[d]: Knew it
17:10 karolherbst[d]: they mix a few different AA methods
17:10 karolherbst[d]: but not FXAA
17:10 karolherbst[d]: their GUI has a thing somewhere for it
17:10 karolherbst[d]: application profiles has some settings where it's more obvious what's going on
17:11 pjb[d]: Yeah that makes a lot of sense. Thanks
17:11 pjb[d]: Specifically, it seems Starsector relies on this nonstandard behavior to make their UI scaling work properly, and I guess the dev isn't aware that it's nonstandard and nvidia specific. At least that's my guess until they get back to me.
17:14 karolherbst[d]: in what way?
17:14 pjb[d]: https://fractalsoftworks.com/forum/index.php?topic=34575.0 I'll just link the bug report if you're interested
17:15 pjb[d]: (also I hope you don't mind being quoted as another reply)
17:20 karolherbst[d]: well I never checked what nvidia specifically is doing, but I think technically x16 is the biggest the hardware can do, but there are other reasons why x16 isn't advertised or so?
17:20 karolherbst[d]: but they have special antialising modes on x8 and x4 msaa
17:20 pjb[d]: hm, I see
17:33 hatfield: mhenning[d]: booting with that config option definitely improved things
17:40 karolherbst[d]: I wonder if we can convince ubuntu to enable it by default....
17:40 karolherbst[d]: but the next LTS release isn't _that_ far awaay
17:57 jannau: change the kernel default with Cc: stable. should be enough time if it's merged in the next merge window. no convincing necessary
19:10 mhenning[d]: The kernel default has already been changed in 6.18. I don't think it's appropriate to backport to stable
21:38 airlied[d]: karolherbst[d]: do we leak the type remapping hash table in the cmat lowering?
21:39 karolherbst[d]: maybe?
21:39 karolherbst[d]: looks like we doo...
21:40 karolherbst[d]: oopsie
21:40 karolherbst[d]: airlied[d]: do you want to fix it?
21:41 airlied[d]: I'll throw an MR up now
21:42 karolherbst[d]: maybe just throw it onto the stack and use `_mesa_pointer_hash_table_init` instead
21:43 airlied[d]: I'm not near anywhere I can test it, so I'll just clean it up
21:53 airlied[d]: karolherbst[d]: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38843