03:23 fdobridge_: <a​irlied> @gfxstrand needed https://paste.centos.org/view/raw/0ae81ceb to get past that, now libdrm links wrong
03:23 fdobridge_: <g​fxstrand> It builds for me on f35 with no patches
03:23 fdobridge_: <g​fxstrand> Including 32-bit
03:24 fdobridge_: <a​irlied> maybe I have an old meson or someat
05:49 i509vcb: before I can even try to do stuff with nouveau on T234 it seems like the simpledrm platform buffer that was being used just decided to stop working somewhere between 6.5.6 and 6.8-rc2...
14:55 fdobridge_: <t​om3026> @asdqueerfromeu the packages need an update "nvk-compressed-image.patch" doesnt apply anymore
14:58 fdobridge_: <!​DodoNVK (she) 🇱🇹> I see that MR got merged (so I have to delete the patch)
16:26 fdobridge_: <g​fxstrand> What all patches are you carrying right now?
16:28 fdobridge_: <S​id> ```nak-iadd3-imad.patch
16:28 fdobridge_: <S​id> nvk-memory-budget.patch
16:28 fdobridge_: <S​id> ```
16:28 fdobridge_: <S​id> ```
16:28 fdobridge_: <S​id> nak-iadd3-imad.patch
16:28 fdobridge_: <S​id> nvk-memory-budget.patch
16:28 fdobridge_: <S​id> ``` (edited)
16:31 fdobridge_: <g​fxstrand> Okay, that's not bad at all. I doubt iadd3/imad will make much perf difference in most applications but it's harmless to carry.
16:32 fdobridge_: <g​fxstrand> Memory budget I'm hoping to finish up and land in the next couple of weeks. It'll probably be gated behind 6.9, unfortunately, but at least it'll be in.
16:32 fdobridge_: <g​fxstrand> And we can probably do something crappy on old kernels.
17:10 fdobridge_: <!​DodoNVK (she) 🇱🇹> Does external_memory_host require kernel work too?
17:32 fdobridge_: <g​fxstrand> Yes
17:32 fdobridge_: <g​fxstrand> That one's actually non-trivial but I think dakr was going to look into it
17:35 fdobridge_: <g​fxstrand> I think I'm going to invoke the EXT fast-track process and go ahead and ship VK_EXT_memory_map_placed without CTS tests. Hopefully in the next few weeks.
17:36 fdobridge_: <g​fxstrand> It's been blocking for like a year on CTS tests when there are multiple working implementations. It's time we just ship the damn thing.
17:38 fdobridge_: <!​DodoNVK (she) 🇱🇹> RADV/NVK doesn't support it though (which are the 2 drivers I'm using)
17:39 fdobridge_: <d​adschoorse> isn't that every second vulkan extension in a nutshell?
17:43 fdobridge_: <g​fxstrand> It will take me all of 20 minutes to type and test an NVK implementation.
17:43 fdobridge_: <g​fxstrand> An ANV implementation exists as does one for the nvidia blob.
17:43 fdobridge_: <g​fxstrand> I'm sure the RADV folks can implement it. In fact, I about did. I just got fed up with the RADV winsys getting in the way.
17:44 fdobridge_: <d​adschoorse> yeah, winsys and libdrm
17:44 fdobridge_: <!​DodoNVK (she) 🇱🇹> I think I can even implement it by referencing the ANV implementation
17:45 fdobridge_: <d​adschoorse> I think part of the problem is that it's just really hard to write cts tests if you are not already familiar with the code base.
17:46 fdobridge_: <d​adschoorse> if it's easy for me to write a standalone test than develop one using the cts framework, there's something wrong imo 🐸
17:46 fdobridge_: <g​fxstrand> go for it
17:46 fdobridge_: <d​adschoorse> if it's easier for me to write a standalone test than develop one using the cts framework, there's something wrong imo 🐸 (edited)
17:46 fdobridge_: <g​fxstrand> there are crucible tests
17:50 fdobridge_: <r​inlovesyou> oh shit?? 6.7.3 fixes GSP for turing cards that had issues 🔥
17:50 fdobridge_: <r​inlovesyou> https://cdn.discordapp.com/attachments/1034184951790305330/1203034656413323284/image.png?ex=65cfa055&is=65bd2b55&hm=29ab62d33784d003bdbaafd361de215b9e94196bc8963f216d17efb792211488&
17:51 fdobridge_: <k​arolherbst🐧🦀> let's goooooo :ferrisBongo:
17:51 fdobridge_: <S​id> yippee
17:55 fdobridge_: <!​DodoNVK (she) 🇱🇹> It also re-introduces the malicious bug on PRIME so it's a neutral release 🍩
17:58 fdobridge_: <r​inlovesyou> ahh luckily i don't use that
17:58 fdobridge_: <r​inlovesyou> :Hehe:
17:59 fdobridge_: <S​id> I'm a dumbass and stick to -rc releases
17:59 fdobridge_: <S​id> and apply any patches on top
17:59 fdobridge_: <S​id> ...maybe I should just start using drm-misc-next
18:01 fdobridge_: <r​inlovesyou> i could not use the 6.8 series because my screen kept freezing
18:02 fdobridge_: <r​inlovesyou> Linus said that would be fixed in rc2, and to trust him. My trust in him is forever broken 🥺
18:02 fdobridge_: <S​id> even more reason to use it, so you can get a log and get dave to fix that issue :D
18:02 fdobridge_: <S​id> /j
18:07 fdobridge_: <m​ohamexiety> bleeding edge gang
18:08 fdobridge_: <k​arolherbst🐧🦀> I ran rc kernels on my gentoo install until an rc kernel corrupted my fs, since then I.. uhh... tried to avoid it
18:08 fdobridge_: <r​inlovesyou> the only bleeding edge i'm on is mesa
18:08 fdobridge_: <r​inlovesyou> :Hehe:
19:59 fdobridge_: <a​irlied> I've sent the new prime + sync fix for 6.8-rc3 just now
20:01 aswithpaulig: Someone generated a 16-bit processor in excel, however that was not what i meant , this can be generated with random calculator using only adder there is no logic gates needed so with this 512 magic i did such a line 1255−112−212−419, that is one nice example how my algorithms work, so everything above 512 outputs as input that gets 512 removed, below just inverts, except for the values seen as 212 or 724 12 or 524 419 o
20:01 aswithpaulig: r 931 those were used to generate this little processor for an example lets append -724 1255-724-112−212−419 that gives -212 and -524 instead gives -12, so let's remove 419 from the equation/formula, now inverter clamps to 931 if above 512, it's because messing with such procedures hits some Mersenne base numbers likely, you can do the same with pocket calculator, its so weird.
20:03 fdobridge_: <m​ohamexiety> btw speaking of bleeding edge @airlied would the new super GPUs work with nouveau? or would they need newer firmware
20:06 fdobridge_: <!​DodoNVK (she) 🇱🇹> Where can I find it?
20:08 fdobridge_: <S​id> @asdqueerfromeu
20:08 fdobridge_: <!​DodoNVK (she) 🇱🇹> This was older than I thought
20:08 fdobridge_: <a​irlied> @mohamexiety don't know, cross fingers they just need nouveau enablement, but until someone gets one and tries I've no real idea (I think phoronix had one and it didn't bind)
20:09 fdobridge_: <m​ohamexiety> got it. thanks!
20:09 fdobridge_: <!​DodoNVK (she) 🇱🇹> Could it be backported to 6.7? 🤔
20:09 fdobridge_: <S​id> it can be backported, yeah
20:09 fdobridge_: <S​id> I think so
20:14 aswithpaulig: it turned out that this was a woman who researched such numbers in 17th century
22:05 fdobridge_: <t​om3026> yeah they apply cleanly ontop of 6.7.3 which has reverted the earlier fence commit
22:05 fdobridge_: <t​om3026> building it myself now :p
22:06 fdobridge_: <S​id> nice :D
22:07 fdobridge_: <S​id> I'm a bigger number betterer dumbass, so I stick to the rcs
22:10 fdobridge_: <t​om3026> yeah im like karol i have suffered on such bleeding edge kernels before so i rather keep it somewhat sane unless its for testing
22:13 fdobridge_: <S​id> I keep a stable kernel installed for backup purposes
22:13 fdobridge_: <S​id> but I main rc
22:14 fdobridge_: <S​id> because it's fun
22:14 fdobridge_: <r​inlovesyou> I don't really care because i can always just chroot in and revert
22:15 fdobridge_: <r​inlovesyou> I don't really care because worst case scenario i can always just chroot in and revert (edited)
22:20 fdobridge_: <t​om3026> not if your FS blows up in a spectacular boom
22:21 fdobridge_: <S​id> funny part is
22:21 fdobridge_: <S​id> I've been using a bcachefs pool as my root drive since 6.6rc
22:21 fdobridge_: <S​id> I love living on the edge :D
22:30 fdobridge_: <t​om3026> if it werent for shader compiling stutter cs2 is running at like 100-200 fps O_o things have surely improved the last week
22:30 fdobridge_: <t​om3026> or two
22:31 fdobridge_: <t​om3026> would be neat if mangohud actually worked too so a bit more reliable metrics could be provided heh
23:09 pasapeadra: You might be ill or sick i assume, much worse than me taken over how you behave, but you deal with complex issues though albeit from slightly pointless angle, but work is work, if someone pays for it that is not so big problem, someone has to deal with that stuff too, but if you bother me again in violent ways overseas, you gonna get very hurt, there is a queue of earlier precedent cases to carry out slaughters at. I think i am going
23:09 pasapeadra: for it, and that is first time in my career, when someone asked have i killed before, i said no, cause i haven't but seeing such questions and gangster wannabes i definitely would if needed and i will.
23:23 Lyude: sometimes it feels like he outsourced his spam to an AI
23:51 fdobridge_: <r​inlovesyou> I really don't get why they do it here
23:52 dwfreed: I don't recommend traveling down that rabbit hole, it may cost you your sanity