09:36 fdobridge: <j​oobei> Does anybody remember this error when installing python dependencies for tizonia?
09:36 fdobridge: <j​oobei> https://cdn.discordapp.com/attachments/1034184951790305330/1213781988461514782/image.png?ex=65f6b991&is=65e44491&hm=0fbfc26636867805a7ec1dab8fdb79900789a1141d6272def79f825029fcd00a&
09:36 fdobridge: <j​oobei> interwebs say that setuptools need to be downgraded to <58 but that didn't solve it for me.
09:45 fdobridge: <j​oobei> "use_2to3 is invalid" it says
10:51 fdobridge: <S​id> hmm
10:52 fdobridge: <S​id> I just noticed how you still need something other than nouveau/zink for gallium-va
11:30 fdobridge: <!​DodoNVK (she) 🇱🇹> Do you need VA-API? The VA-API support on nouveau is very limited
11:39 fdobridge: <S​id> don't need, was just curious about the current state and what plans we have for it
11:39 fdobridge: <S​id> my iGPU does all the va-api lifting anyway
11:42 fdobridge: <!​DodoNVK (she) 🇱🇹> Right now the focus is on Vulkan video instead
11:44 fdobridge: <S​id> right, fair
12:22 fdobridge: <e​sdrastarsis> Btw, its possible to translate vk video to vaapi in Zink
12:23 fdobridge: <S​id> oh
12:23 fdobridge: <S​id> huh
12:23 fdobridge: <S​id> neat
12:23 fdobridge: <e​sdrastarsis> Btw, its possible to translate vaapi to vk video in Zink (edited)
12:24 fdobridge: <e​sdrastarsis> https://www.phoronix.com/news/VA-API-On-Vulkan-Video-Hacks
12:24 fdobridge: <S​id> nice
12:25 fdobridge: <!​DodoNVK (she) 🇱🇹> @ airlied Has there been any progress after over 2 years? 🤔
13:13 sin_ack: good day to everyone. could anyone help me with what i posted yesterday? i can re-post if needed.
13:18 fdobridge: <!​DodoNVK (she) 🇱🇹> Discord users can see your older messages thanks to a bridge
13:23 sin_ack: got it, thanks
13:27 karolherbst: sin_ack: yeah sorry, nobody got to your message yet, maybe also because it's like weekend 🙃 anyway... it doesn't seem like anything is wrong looking at those logs, so that's kinda weird
13:28 sin_ack: no worries at all, i understand completely :) i was just asking in case someone's interested
13:28 karolherbst: sin_ack: you are on mesa 24.0?
13:28 sin_ack: 24.0.2
13:28 karolherbst: mhhhh...
13:28 sin_ack: that shouldn't matter though, right? since not even framebuffer comes up
13:28 karolherbst: yeah.. something is really weird here
13:29 karolherbst: sin_ack: it's a hybrig GPU laptop right?
13:29 karolherbst: did you put it into discrete mode or something?
13:29 sin_ack: yes, it's in discrete mode right now
13:29 sin_ack: normally i would just use intel, but the intel driver in the kernel causes significant flicker and i haven't figured it out
13:29 karolherbst: huh...
13:29 karolherbst: that's also weird...
13:30 sin_ack: like i said, this is the laptop from hell. so many issues on linux
13:30 sin_ack: even with the official nvidia drivers and xorg, it can't resume from suspend
13:31 karolherbst: I wonder if simplefb interfers in weird ways?
13:31 sin_ack: ah, that's possible. should i disable it in my kernel?
13:31 karolherbst: I dunno...
13:31 karolherbst: I have an idea...
13:32 sin_ack: i'm all ears ;)
13:33 karolherbst: sin_ack: mind using my config from fedora with `make olddefconfig` and see if booting that kernel makes it all work? https://gist.githubusercontent.com/karolherbst/b42d0624747c35edc86d953eddb6e4db/raw/72077943d2f802120d7e58d396e24088a6112fc5/gistfile1.txt
13:33 karolherbst: though mhh
13:33 karolherbst: depending on your initramfs setup, you might have to change a few things
13:33 karolherbst: maybe just puck the drm configs
13:33 karolherbst: (and fb/vt/whatever)
13:33 karolherbst: or use it as a reference
13:34 sin_ack: alright, let me try that
13:34 fdobridge: <S​id> oh boy I got a long email :D
13:38 sin_ack: huh, "CONFIG_DRM_ACCEL is not set" in my kernel... that's probably not good
13:39 sin_ack: i saw that 2D support is done with DDX on the ada GPUs
13:39 karolherbst: the nouveau ddx isn't used on newer gpus
13:39 karolherbst: it's all the modesetting
13:39 karolherbst: sin_ack: CONFIG_DRM_ACCEL is just for accelerator devices
13:40 karolherbst: `drivers/accel`
13:40 sin_ack: ah, alright then
13:40 sin_ack: i isolated just the "Graphics support" part from the config, will try applying whatever's in there now
13:44 fdobridge: <t​om3026> redhat want to hire you as a kernel maintainer?
13:44 fdobridge: <S​id> heh, no
13:44 fdobridge: <S​id> something about regression tracking
13:44 fdobridge: <k​arolherbst🐧🦀> ahh thorsten
13:46 sin_ack: hmm, building a lot of nouveau objects now :) looks promising
13:47 karolherbst: it should have done that before tho
13:47 sin_ack: probably, yeah
13:48 sin_ack: alright, testing now
13:48 karolherbst: sin_ack: also picked the vtcon stuff?
13:49 karolherbst: or console stuff in general
13:49 sin_ack: err, i don't recall. i just grafted the Graphics support part, which i think includes that bit, no? i'll check in a moment
13:49 sin_ack: no display atm, for the record
13:50 karolherbst: yeah.. no idea what specifically would be important here
13:50 sin_ack: let me see if nouveau got loaded right, one sec...
13:52 sin_ack: https://paste.debian.net/1309308/ hmm, seems alright...
13:52 sin_ack: at first i thought it might have been something to do with the nvidia drivers trying to load themselves in despite being blacklisted, but even with module_blacklist, it's not working
13:53 karolherbst: mhhh
13:53 karolherbst: but it does work with intel, right?
13:53 karolherbst: ohhhh wait...
13:53 karolherbst: mhhh
13:53 fdobridge: <S​id> I'm not entirely sure how to respond, do I just answer their questions :D
13:53 karolherbst: sin_ack: I think I know what this could be...
13:53 sin_ack: the display works with the open intel drivers and the proprietary nvidia drivers
13:53 karolherbst: yeah....
13:54 karolherbst: sin_ack: soo.. if you boot with intel enabled in the firmware, could you check if you have two `eDP` entries inside `/sys/class/drm`?
13:54 karolherbst: it could be that nouveau needs to trigger switching the display to the other GPU
13:54 sin_ack: ah, sure. i'll do that now
13:55 karolherbst: sin_ack: but you said you see stuff before nouveau loads?
13:55 sin_ack: yep, the VGA console and GRUB both work
13:55 karolherbst: mhhh
13:55 karolherbst: let's see...
13:56 sin_ack: incidentally, the flickering on intel only starts when intel takes over the framebuffer
13:56 sin_ack: alright, got display with hybrid graphics now, reporting what i see...
13:57 karolherbst: sin_ack: does disabling simpledrm do anything?
13:57 sin_ack: can i disable that with a kernel parameter? otherwise i'll try next boot
13:57 karolherbst: no idea
13:58 sin_ack: ok, then i'll try removing it from the kernel
13:58 fdobridge: <t​om3026> initcall_blacklist=simpledrm_platform_driver_init
13:58 sin_ack: https://paste.debian.net/1309311/ /sys/class/drm/ contents
13:58 fdobridge: <t​om3026> should do it i think
13:58 sin_ack: tom3026: thanks! i'll try that now
13:59 karolherbst: yeah so that's two eDP.. but mhh
13:59 sin_ack: i can also post the /sys/class/drm/ contents with the proprietary nvidia driver
13:59 sin_ack: let me try disabling simpledrm with discrete only first
14:02 sin_ack: still no display with that, and i didn't get the initial kernel output before nouveau took over this time
14:03 karolherbst: mhhh okay
14:04 sin_ack: https://paste.debian.net/1309312/ dmesg, discrete only + nouveau + simpledrm disabled
14:04 sin_ack: oh, i only have one eDP when discrete only is selected...
14:04 karolherbst: yeah..
14:04 karolherbst: because the intel one is gone now
14:05 sin_ack: https://paste.debian.net/1309313/ ls /sys/class/drm/, discrete only + nouveau
14:05 karolherbst: I wonder if something silly is going on.. but I guess it's more of a nouveau bug at this point
14:05 karolherbst: it's just strange that.. uhh.. no errors should up
14:06 sin_ack: interestingly, errors do show up once i try to start up xorg
14:06 karolherbst: sin_ack: does anythign show up as connected with this `grep . /sys/class/drm/*/status`?
14:06 sin_ack: /sys/class/drm/card0-eDP-1/status:connected
14:06 karolherbst: sin_ack: mind trying to connect an external display and see if that shows something?
14:06 sin_ack: ah, sure. i only have my desktop monitor, but i'll try that now
14:07 sin_ack: brb
14:08 sin_ack: yes, that works perfectly!
14:08 karolherbst: mhhh
14:08 karolherbst: so I guess something is broken with eDP specifically
14:09 karolherbst: so yeah.. either something special needs to be done with your internal display, or nouveau doesn't handle it correctly
14:09 sin_ack: wonder why the intel driver figures it out and nouveau doesn't :S
14:09 sin_ack: and nvidia for that matter
14:10 karolherbst: well.. nouveau probably has still a few bugs here and there
14:10 sin_ack: fair :P
14:11 sin_ack: what would you recommend now? should i open a bug report somewhere?
14:13 karolherbst: you could try to compile from git https://cgit.freedesktop.org/drm/drm-misc/ branch: drm-misc-next or drm-misc-fixes
14:14 sin_ack: oh sure, i can try that first. i'll be away for a couple hours, will post my findings.
14:23 sin_ack: one more thing i noticed with nouveau before i go: for some reason, glxinfo/glxgears don't work when running under wayland with some X opcode, and i get "module nvidia is blacklisted" in my dmesg when i try. they work fine when i launch xorg, though
14:23 karolherbst: mhhhhhh
14:23 karolherbst: shouldn't impact the console though
14:24 sin_ack: yeah, should be unrelated
14:24 karolherbst: but maybe it makes sense to remove everything nvidia related
14:24 sin_ack: i'd have to do that next weekend, work requires 3D acceleration ^^;
19:57 sakuska: So i figured that the transition graph is compiler thing,example 334 to 76 reductive transition, is managed 254+334is 588 then it adds the mask of 254 to all elements which if all elements is same n times 258, so now 76+258 is 334 and other elements are 258, so 334-334is0 and 334-258 is 76, if you subtract that from all elements transitioned it gives you the correct element, so only that 334 is generated at runtime and that comes from operands,
19:57 sakuska: 258 doubled adds is done by compiler, compiler also adds odd and even operand offsets, and those transitions they come per arithmetic so 3+4 is transitioned into some result dict. compiler also analyses if it needs to add 512's since predicates can only reduce the way i see. I have not been able to make a bigger value from smaller that is.