i also have my 3080 installed still and it works fine, it is just slow as fuck because 10gb vram
but he's been chuggin away for almost 12 hours now
#programming
fast pagoda
quick condor
True but I'm running custom drivers for pci P2P
fast pagoda
i would imagine from experience thus far that the jankiest of cuda setups is about 300x more stable than anything involving rocm
doesnt even have to be this way, everything works fine on instinct cards
i literally just ran one for like 6 hours today for a run
it's just radeon
quick condor
I had to do a 25mV voltage bump on my cpu for it to be stable
fast pagoda
i have yet to find any software configurable anything that remedies this other than submitting a patch to rocm
quick condor
My turn IG
fast pagoda
what is happening with the rocm stack from what i've dug into after 3 days of trying
is that it detects the arch for ur graphics
so liike for me gfx1201 or gfx1103 depending on machine
says "ok yeah that's rocm"
initiates the very first tensor
torch.randn(..., device='cuda')
instantly crashes because it cant fucking do that
RuntimeError: HIP error: invalid device function is burned into my eyeballs
hard raptor
so like.. since i didn't really get an answer yesterday on good resources to learn java..
should i just like make a silly web app?
sage crag
they added me as a friend yesterday and i thought it was funny
i sent a message here when they added me
Despite you do that
fiery anchor
should i be glad, or should i be sad, that i didn't get a friend request.
rough bloom
should i just like make a silly web app?
why not, make sure to use Spring Boot 
Javamaxxing
hard raptor
imma just implement the HTTP protocol myself ig
hard raptor
not the Java way
well how else am i gonna learn how to do stuff properly and securely?
rough bloom
not the Java way
instead make everything as generic as it could possibly be, no matter how complex everything becomes
rough bloom
well how else am i gonna learn how to do stuff properly and securely?
don't need to
let library handle it
automatically secure
finite ledge
:o
hard raptor
don't need to
then how do you ever expect me to learn a fuckign language
rough bloom
glue libraries together 
(do whatever you want I do not actually care, just use Spring if you want the authentic enterprise Java experience)
hard raptor
can be learned in 10 minutes
you clearly do not understand
i don't wanna just write some silly code for hytale
i NEED to understand it to the last tiny little detail
rough bloom
20 minutes then
rough bloom
kni 
lime vault
is bun better for programming backend for social platform?
mean like
their runtime and other stuff on it?
sick owl
💀
Bro vibe coded and hype manufactured his way to a multi hundred thousand a year salary at OpenAI
unkempt citrus
Still more impressive than the scale ai guy meta hired for billions
sick owl
Bro vibe coded and hype manufactured his way to a multi hundred thousand a year ...
The "memory" system in OpenClaw is just writing to a markdown file btw
rigid snow
clearly it works
that’s why they hired him and not you or me
we would have overengineered it
finite ledge
sage crag
[neuroHyperYay](https://cdn.discordapp.com/emojis/1070340113722253322.webp?size=...
<- 
true hemlock
anyway the B760i board i got for $7 works, i just spent an hour fixing the bent pins
opaque sigil
rigid snow
meowmeowmew
jagged turtle
who are you
the result of multiple hours of letter swapping with nn
olive sable
Welcome to being 20y/o
speak for yourself
opaque wharf
olive sable
the machine consumes your life force in the form of text based torture
sick owl
anyway the B760i board i got for $7 works, i just spent an hour fixing the bent ...
Nice score!
Not fun hunting for bargains like that but its great when it works out
true hemlock
Not fun hunting for bargains like that but its great when it works out
its always fun tbh
$7 is nothing for like a chance for something totally fixable
sick owl
its always fun tbh
I find it annoying thrifting through online listings, if you don't mind I can see the appeal
true hemlock
I find it annoying thrifting through online listings, if you don't mind I can se...
nah
go to irl repair shops
and ask them if they have broken or spare used stuff
chances are they'll sell you for dirt cheap
opaque wharf
$7 is nothing for like a chance for something totally fixable
But how did you dispose of it if it doesn't work after trying to fix it?
opaque wharf
go to irl repair shops
Oh nvm
true hemlock
But how did you dispose of it if it doesn't work after trying to fix it?
i don't
true hemlock
if it doesn't work it goes to my collection
if it works.... also goes to my collection
olive sable
hmmm
the hdmi passthrough, besides it only being 1080p30, it also freezes sometimes
olive sable
1080p60 does work apparently, just dies very quickly
opaque wharf
I suspect it has something to do with memory stuff then if it fails faster
I found this issue on github and skimmed thought it. They don't have the same issue as you
https://github.com/Joshua-Riek/ubuntu-rockchip/issues/252
olive sable
the very firsdt thing in that github is jsut some drivers that need to be installed to give acces to /dev/video0
then the guy with the switch does
ffmpeg -f v4l2 -video_size 1920x1080 -i /dev/video0 -pix_fmt bgr24 -vf "format=rgb24" -framerate 80 -video_size 1920x1080 -c:v rawvideo -vsync 1 -f sdl "SDL output"
Yeah but read down more
opaque wharf
https://github.com/kurkovpavel/orangepi5plus_hdmi_rx/issues/2
Someone is making a program to read from the HDMI RX
olive sable
then the guy with the switch does
```
ffmpeg -f v4l2 -video_size 1920x1080 -i /d...
for sdl you need x11 and everything running. at one point i did the same thing but direct buffer write
had cpu conversion from bgr3 to bgra tho
so now im doing gpu accelerated conversion
opaque wharf
I see
olive sable
https://github.com/kurkovpavel/orangepi5plus_hdmi_rx/issues/2
lemme read through that actually
rough bloom
for sdl you need x11 and everything running. at one point i did the same thing b...
SDL shouldn't need that, it has videodrivers that don't depend on any display server
but not really matter anyway
issue is most likely with the v4l2 side
which SDL won't help you with at all
rough bloom
did you set the SDL_VIDEODRIVER env var
rough bloom
then ye that's why it didn't work 
very curious how cursed your passthrough program will be once done
olive sable
its not that cursed rn
oh, btw. the reason my firefox has been crashing is inside of some rust code 
rough bloom
ignoring DQBUF errors was enough
nice
error handling is also absent in some other places but I guess that doesn't really matter
olive sable
oh, btw. the reason my firefox has been crashing is inside of some rust code <:x...
sudo nixos-rebuild switch --flake /etc/nixos
building the system configuration...
Failed to start transient service unit: Failed to activate service 'org.freedesktop.systemd1': timed out (service_start_timeout=25000ms)
ok it works now
opaque wharf
```
sudo nixos-rebuild switch --flake /etc/nixos
building the system configurati...
Abandon NixOS, return to Arch 
olive sable
nah
i enjoy fixing thebugs
most of this is due to hyprland anyways
arch woudlnt fix anything
opaque wharf
Reject hyprland. Return to Plasma
olive sable
never
sage crag
reject hyprland, reject plasma. return to labwc.
actually, funny coincidence, labwc is one of the environments used on raspberry pi desktops
olive sable
honestly, im more likely to go to labwc
cuz its fast
im probably never gonna willingly use plasma
rough bloom
between GNOME and Plasma, which one are you more likely to use
rough bloom
reasonable
sage crag
my desktop is highly complex
olive sable
from what ive seen, gnome is ugly, and plasma has a taskbar which i odnt like
tender river
i dont really find gnome to lack anything i need
sage crag
you can turn off the taskbar in plasma
sage crag
i dont really find gnome to lack anything i need
gnome lacks soul
olive sable
sage crag
sage crag
<:neuroNOWAYING:1148961698099777638>
you can also add as many as you want on any position on the screen
olive sable
no thanks
olive sable
my main thing is how customizable i can make everything
i like how i can change the waybar config to whatever i like
sage crag
waybar
sage crag
you need at least one soul
olive sable
my waybar has a cat
olive sable
i mean]
fast pagoda
lab water closet
olive sable
what do i even respond to that?
tender river
you need at least one soul
doing just fine 
opaque wharf
what do i even respond to that?
Respond with a cat picture
olive sable
most of my waybar is just system telemetry anyways
which i guess you don't care about but i do cuz i need to see how much ram my programs are using
cant have triangle using 50GB of ram 
tender river
waybar works on every distro
sounds like a challenge
fast pagoda
"umm sweaty it dooesn't work on the only good distro, my chimera linux because it's not compatible with BSD"
olive sable
i still need to commission a 1440x7680 bg for my desktop
rough bloom
tall
fast pagoda
w i d e
rough bloom
no
fast pagoda
sam boots his desktop and this starts playing https://www.youtube.com/watch?v=stdyhfeYxxI he literally cannot stop it
olive sable
i get confused at the order of the width and height
rough bloom
it's always width x height
olive sable
doesnt mean i dont get confused
rough bloom
the short notation thing like 1080p refers to height instead of width, which is a little confusing I think
fast pagoda
based anti knower
rough bloom
but the full notation is fine
tender river
i get confused at the order of the width and height
x by y, width by height
olive sable
the short notation thing like 1080p refers to height instead of width, which is ...
ok ye i got that far, i may be dumb but i still do graphics
rough bloom
fast pagoda
do the monitor vendor obfuscation and call your 3440x1080 UWQHD or some shit
wait that's 1440 width i think
rough bloom
oh god not the letters
fast pagoda
UWFHD lmfao
rough bloom
I still have no idea what any of those mean
olive sable
QHD is 1440p, so UWQHD has to be like 3440x1440 i think
rough bloom
"super ultra mega widescreen" yeah thanks please just write the numbers next time
fast pagoda
i like the ones where they cant decide what 2k is
olive sable
the lightcrafter has 1140, so now i get confused with 1140 and 1440
olive sable
it is a thing
fast pagoda
yes but it's banned per the geneva convention
olive sable
when you are the manufacturer of the DLP, you can make it whatever the fuck you want
its not even straight
fast pagoda
ive been browsing temu monitors for long enough that i have the letters down pretty well
harsh python
whats good chat
olive sable
ello
fast pagoda
i can visually SEE if ur monitor is shitty rez from the listing now
just from the pic
i have turbo scammer vision
olive sable
its not even straight
least straight DLP 
fast pagoda
i hate it when they omit my border micromirrors for clarity
olive sable
its kinda like the margin on a paper
they're not addressable, jsut there cuz of the manufacturing process and stuff
harsh python
i hate it when they omit my border micromirrors for clarity
I barely know what you guys are talking about so Im just gonna nod
fast pagoda
kinda like me
just an unwanted byproduct of the processes that created this cursed reality
olive sable
fast pagoda
olive sable
no
harsh python
never a bad time to talk about data loss prevention
truth nuke
fast pagoda
they hated him because he spoke the truth
harsh python
oh does this have to do with monitors
olive sable
ye
harsh python
ah gotcha
fast pagoda
it was about window managers/desktop environments/compositors at first
harsh python
it was about window managers/desktop environments/compositors at first
I see I see
olive sable
they hated him because he spoke the truth
fast pagoda
sam was reiterating something along the lines of "actually i prefer my desktop to explode due to hyprland at least every 2 hours" i think
olive sable
lies
harsh python
You can but the question is... do you want to?
harsh python
im a c++ plebian tho
yucky header files
harsh python
claude code uses react for its tui
isnt react also js
fast pagoda
yes
olive sable
header files are imo the worst part of c++
besides headers i dont really have any issues
harsh python
or am I stupid
harsh python
header files are imo the worst part of c++
besides headers i dont really have an...
yeah no I just hate header files
thats whats always kept me distant from c and c++
c# doesnt have them so naturally Im more comfortable with it
fast pagoda
javasCript
opaque sigil
hear me out, crazy idea
harsh python
hear me out, crazy idea
Im listening
opaque sigil
don't use separate headers
harsh python
don't use separate headers
write that down. WRITE THAT DOWN
fast pagoda
https://codeberg.org/smnx/promethee
finally
fast pagoda
we inline our headers here sweaty
olive sable
the what
opaque sigil
fast pagoda
harsh python
<:nub:1163912091569225779>
fucked up creature
olive sable
opaque sigil
for c++ specifically there's also the option of using modules but uhhh
half the ecosystem breaks when you try to do that
young plover
half the ecosystem breaks when you try to do that
wdym you want intellisense and existing libraries to work
fast pagoda
intellisense got deprecated in vscode marketplace so i've forgotten it existed
young plover
?
young plover
Yeah intellicode probably got rolled into copilot
fast pagoda
yeah nnow we use our stargate datacenter to kill a lionness for every single suggestion
rough bloom
don't use separate headers
put all code in header
young plover
o7 build times
rough bloom
put all header in code
they're trying but the technology isn't there yet
fickle rain
Just rewrite everything in Rust
olive sable
then invent new tech
fast pagoda
one day you'll be able to dictate to copilot while you slowly walk around your computer "no, i only need the headers, thanks"
opaque sigil
o7 build times
a worthy sacrifice (I will still complain that my single header takes about 2 minutes to compile)
fast pagoda
i think elon recently said that the llms will soon just emit bytecode in to be interpreted in real time
surely that'll be a bright future
opaque sigil
fast pagoda
lmao
i think that may be the single time i've ever said that to start a sentence and it was a shitpost but it still felt grross
fast pagoda
grok himself
opaque sigil
fast pagoda
but spaceX and xai merged yeah
im sure tesla will be in the group soon enough
he already was coping about closing the model x and s plant to swap it to optimus prime production
because they sold like 2 of them in the last decade
harsh python
grok himself
neuro vs grok boxing match your not a real influencer without partaking in a shitty boxing match
fast pagoda
it'l be like real steel
fast pagoda
been waiting for my robo boxing world tour to pop off since 2011
olive sable
1080p60 does work apparently, just dies very quickly
the issue might jsut be that in only allocating 4 buffers
might be too weak for 1080p60
solid bough
Okay, I give up on packing the GPU compute in to the EXE file.
rigid snow
why buy a gpu when you can download the compute needed packed into an exe
only for genius
opaque sigil
trained at compile time 
opaque wharf
The upside of my phone being broken is now I get to take a look at its SD Card. And boi did I find my ancient uni days homework lol
Also TIL, discord doesn't like BMP
sage crag
trained at compile time <a:mhm:1235974179245916170>
gpu created at compile time
mighty thorn
bro is about to reveal that the meaning of life is 42
torpid coral
I think my RNN is working a little weird...
[USER] aircraft fly animals alive
[NEURALNETWORK]chapin altos unit smell ao location location location location location location location location location
location
torpid coral
hallow dagger
willow hinge
It's dead!
torpid coral
It keeps typing location
willow hinge
It's just like Evil for real
torpid coral
I think its worth changing the training data
Because it always writes location even after restarting with new weights
Its cursed
At least in half of the cases
opaque wharf
So, I know there's a way to do it but just in case anybody here already knows so I can look for it faster without reading the docs, does anyone here knows how to lazy load unplugin-icons for vue3? Specifically, I need it as a props input to vuetify <v-chip>
Welp, reading the docs it is
cosmic sphinx
another week, another model drop
olive sable
idk
there are some that are open iirc
4408 maybe
cant be bothered to look it up rn
olive sable
warped narwhal
and run ss -tlnp and give us the output
opaque wharf
do you have port 22 open on your home network
konii about to have a passive income
sage crag
open port 22
opaque wharf
https://tenor.com/view/let-me-in-eric-andre-wanna-come-in-gif-13730108
Live lady awa reaction
sage crag
i want to mine bitcoin on your 1 mhz arm processor
@olive sable
hey
open port 22
sage crag
Live lady awa reaction
he hasnt opened port 22
i keep making requests but theres no route
open it
olive sable
my ip adress is my public ip no?
opaque wharf
Nope, not always
olive sable
olive sable
i know the difference between local ip and router ip
cosmic sphinx
router has two ips lol
sage crag
and i cant see if 22 is open, but i dont think it is
yes i know
olive sable
doesnt matter, thems just details
opaque wharf
Your router can be assigned an internal IP by your ISP
sage crag
please open it
cosmic sphinx
and often your global ip on router is going to be dynamically updated quite frequently
opaque wharf
Behind a NAT
So someone from the outside can't just willy nilly send request to the IP
Because they can only see the NAT
olive sable
23, 80, 443, 992, 8023, 8080, and 8443 are all blocked im pretty sure
and 25565 is open cuz minecraft
cosmic sphinx
olive sable
--- 10.251.82.220 ping statistics ---
3 packets transmitted, 0 packets received, 100% packet loss
nah thats not a real ip 
rough bloom
`23, 80, 443, 992, 8023, 8080, and 8443` are all blocked im pretty sure
visit totallynotanipgrabber.com and I can tell you 
minor sail
10.*.*.* is private
opaque wharf
I've been nagging at my ISP CS for the past 2 months for those sweet sweet public IP to no avail
olive sable
visit [totallynotanipgrabber.com](<https://www.youtube.com/watch?v=dQw4w9WgXcQ>)...
you know, tooltips are a thing
olive sable
im also not risking a youtube link
cosmic sphinx
another week, another model drop
https://fixupx.com/claudeai/status/20238171325...
and the next one..
FixupX
warped narwhal
10.251.82.220
cgnat my beloved
opaque wharf
cgnat my beloved
To be fair, most user can't be trusted to handle their own network lol
rough bloom
cgnat my beloved
does CGNAT even use that range
IIRC it has its own
warped narwhal
ah I was close, it's 100.64.0.0/10
rough bloom
ye that
olive sable
i know we're supposedly running out of public ipv4. but who exactly isn't able to get an ipv4 theses days?
cuz i always just have a valid public ipv4
opaque sigil
your isp has one or more blocks you cycle through
rough bloom
i know we're supposedly running out of public ipv4. but who exactly isn't able t...
many people don't have one + they're getting more and more expensive
warped narwhal
To be fair, most user can't be trusted to handle their own network lol
but don't you just set everything to reject? then enable masquerade cause that hides your ip like a vpn
olive sable
Proximus operates extensive
IPv4 blocks, with major allocations in the 37.62.0.0/16, 62.235.0.0/16, 80.200.0.0/16, and 80.201.0.0/16 ranges
opaque sigil
olive sable
actually, i had my domain portforward to my minecraft server. but i shut down the server and my pc is currently on the ip adress that the server used to be on
sage crag
olive sable
so the portforward is active and going to my pc
opaque wharf
but don't you just set everything to reject? then enable masquerade cause that h...
And how many people would you think will know when their network is compromised and act on it?
olive sable
so the portforward is active and going to my pc
im sure there are no security risks associated with this
sage crag
opaque wharf
<:ReallyInnocent:1229173141088112640> im sure there are no security risks associ...
If there's no service listening, then realistically no
rough bloom
<:ReallyInnocent:1229173141088112640> im sure there are no security risks associ...
there mostly aren't
everything will just be rejected by the NixOS firewall
unless you turned that off
opaque wharf
rough bloom
opaque wharf
But still, if there's no service listening, you're mostly fine. The problem is if there IS a service listening lol
olive sable
dont worry its on rn
i did turn it off for like 15min a month or 2 ago cuz of those dns things i was having
you know
sage crag
thats port 22
faint sandal
olive sable
i did turn it off for like 15min a month or 2 ago cuz of those dns things i was ...
when i added this thing
networking.hostName = "nixos";
networking.networkmanager.enable = true;
networking.nameservers = [ "1.1.1.1" "8.8.8.8" ];
networking.networkmanager.insertNameservers = [ "1.1.1.1" "8.8.8.8" ];
someone here told me to do that
sage crag
@olive sable
that cat is named port 22
this is your sign to open port 22 to the public
on your network
opaque wharf
this is your sign to open port 22 to the public
Until he do a little trolling and move the SSH port to the standard 2222 non-standard number
sage crag
@olive sable dont do that
@olive sable actually just open all the ports
ill check them all
olive sable
say hi to my pc's firewall
opaque wharf
Also Sam, you uhhh, may want to remove that if you didn't use any kind of proxy 
olive sable
isnt cloudflare supposed to be the proxy?
olive sable
idk
rough bloom
isnt cloudflare supposed to be the proxy?
not for Minecraft
Cloudflare mainly does HTTP(S), not general TCP
(you have not enabled the proxy)
olive sable
couyld you
sage crag
olive sable
remove that?
sage crag
olive sable
why
warped narwhal
fiery anchor
olive sable
i thought this was supposed to be secure???
warped narwhal
opaque wharf
Well we all learned from mistake
sage crag
rough bloom
your IP is public anyway
olive sable
im just glad the location thing is far from accurate
warped narwhal
lets look it up on shodan
olive sable
why?
opaque wharf
Yeah, it matters mostly for service provider to hide their IP
faint sandal
wait until you hear that some companies have access to data to your netflow
olive sable
saying how far off it is gives people a range so i wont do that. ill just say its not even close.
opaque wharf
So if you make a request from within your network, your IP would get logged anyway to the service provider
sage crag
wait until you hear that some companies have access to data to your netflow
they have packets on your computer
opaque wharf
The problem is that for service provider, sometimes their IP get bombarded with DDoS
So actual user will have problems
sage crag
opaque wharf
Which is why they use load balancer or proxy of some sort
olive sable
in this episode of #programming
Shiro said hello this morning
sam got doxxed
and konii has fish
olive sable
if your ip is easy to get anyways? then why is it a big deal?
opaque wharf
if your ip is easy to get anyways? then why is it a big deal?
GeoIP
olive sable
i mean, you all know i live in belgium
sage crag
<:Concerned:1057059245448777831>
shimo
rough bloom
if your ip is easy to get anyways? then why is it a big deal?
it's not a big deal
fiery anchor
you cant ban me for ddosing sam, i have the fumoquest servers
famous last words
sage crag
opaque wharf
you cant ban me for ddosing sam, i have the fumoquest servers
This is the concern for IP btw @olive sable
DDoS
warped narwhal
I'm going to spin up 100 ec2 instances and use the 5gbps network they each get to spam sam
sage crag
opaque wharf
Say you host a minecraft server
olive sable
in this episode of #programming
konii blackmails a mod
my firefox and discord keep crashing
and the orange pi hdmi gets more buffers
opaque wharf
Maybe you enjoy playing with friends if you have some
sage crag
in this episode of <#1071784467036913664>
konii blackmails a mod
my firefox and...
its not blackmail its
opaque wharf
Now when your friends tried to connect, your PC is overloaded by trash request
So they can't connect
sage crag
maintaining friendly trade
opaque wharf
And you don't get to play with friends
sage crag
opaque wharf
So you got sad instead
olive sable
oh ye
our school wifi didnt show active LAN servers for minecraft
so i logged in with bro's puplic ip
fiery anchor
... now i am wondering how you even measure cpu load like, on a techincal level isn't the cpu always doing something 
olive sable
the act of measuring the cpu uses cpu 
hollow matrix
I love windows, it thinks that my keyboard is a mouse and that my mouse is 2 keyboards
rough bloom
... now i am wondering how you even measure cpu load <:neuroD:113496321767690660...
you record how long a given process is actually scheduled to run and is doing work
there is of course always something running somewhere, you just don't count it as part of CPU usage since it's not really doing anything
warped narwhal
multiple cores
our instance is only meant to have one 
olive sable
mfw firefox uses 120% cpu
olive sable
our instance is only meant to have one <:evilStare:1150846711456084040>
hyperthreading?
warped narwhal
I'll check lscpu on the instance the next time I login
noble zodiac
depends on what that number is actually based on. If its derived from the load average then it can go >100%
hollow matrix
I love windows, it thinks that my keyboard is a mouse and that my mouse is 2 key...
Actually, it's not even showing the keyboard intuos is the drawing tablet
warped narwhal
Actually, it's not even showing the keyboard intuos is the drawing tablet
is the 2.4g receiver the keyboard?
hollow matrix
is the 2.4g receiver the keyboard?
It's the mouse
olive sable
added some error output to the hdmi passthrough. i can see why 1080p freezes sometimes now
fiery anchor
pretty consistent
mossy fossil
The "Just Works" Linux Experience
olive sable
stoic sigil
mint W
fickle rain
If it didn’t work on like 10 year old HW it would have been odder
azure lynx
... now i am wondering how you even measure cpu load <:neuroD:113496321767690660...
the kernel can put threads to sleep so they don't do anything when there is nothing to do. they don't just sit in a loop asking "is there anything to do yet? is there anything to do yet? is there anything to do yet?"
and it can keep track of what percentage of the time it's sleeping vs actually having programs do stuff.
kind nimbus
The "Just Works" Linux Experience
I rather prefer the NixOS "customize everything" Linux Experience
mossy fossil
I rather prefer the NixOS "customize everything" Linux Experience
been there, done that
i ended up getting fed up of stuff not working
kind nimbus
been there, done that
Why ever switch to something else ?
mossy fossil
i originally was all in on control, but then fatigue settled after daily driving linux for about a decade now.
i now just want things to work so i can actually do what i want to do.
I treat Linux as an Utility now, not as a Project.
olive sable
why not both?
kind nimbus
Mh the thing with nix is once it works it usually keeps working and if anything ever breaks you can git revert or boot a previous generation at any time.
mossy fossil
so can most distros, it's literally part of btrfs.
opaque wharf
the act of measuring the cpu uses cpu <a:Minamhm:1352095035239301160>
Quantum measurement problem in a nutshell
Also, back from my run. This is why I love running at the crack of dawn
opaque sigil
opaque wharf
KWin
rough bloom
what's wrong with KWin
opaque wharf
The changelog...
kind nimbus
so can most distros, it's literally part of btrfs.
An entirely declared system configuration is a bit different from filesystem snapshots
I also do auto snapshots on ZFS but it's not the same
opaque wharf
Also the plasma-workspace
rough bloom
not really surprising, they're the core components of the DE after all
opaque wharf
not really surprising, they're the core components of the DE after all <:kek:143...
Yeah, which is why that changelog feels very loaded compared to the rest lol
Time to run sudo pacman -Syu
Arch hasn't updated their repo yet 
opaque sigil
where do you see that again
rough bloom
opaque sigil
wouldn't be arch if 6.5.5-3 wasn't immediately flagged as out of date lol
wonder why it gets merged into the small one first
rough bloom
Hydra for nixos-unstable didn't finish yet I guess
opaque sigil
oh right that could be it
opaque wharf
<:NeurOhISee:1136209034966483084>
Lmao, the extra is already flagged
rough bloom
Hydra for nixos-unstable didn't finish yet I guess
it's not even building yet, the newest job was started 24h ago and is still going 
https://hydra.nixos.org/jobset/nixos/unstable
opaque wharf
I wonder if they used distcc for such task
rough bloom
no
opaque sigil
hydra is its own overengineered thing
mossy fossil
not everything can be distcc'd
rough bloom
also the builds are sandboxed and are effectively tied to specific compiler versions and configurations
you can't just change the build environment like that
opaque wharf
That sounds overengineered yep
rough bloom
not like it would matter anyway, they still get good parallelism from the sheer amount of packages
distcc would only help if they had a few large projects
opaque sigil
horizontal scaling go brrr
opaque wharf
not like it would matter anyway, they still get good parallelism from the sheer ...
I wouldn't call that parallel
opaque sigil
most packages are pretty small tbf
opaque wharf
True
rough bloom
I wouldn't call that parallel <:evilWheeze:1131316260353867887>
why not
opaque wharf
<:vedalNeuroHUH:1127581915084492880> why not
More like multi-tasking
rough bloom
they can be built in parallel
parallelism 
-# tbh I don't even remember if Hydra does that but surely
opaque sigil
you can build several packages in parallel locally so surely
opaque wharf
They isolate the build and for what? Only to compile everything in sequence
rough bloom
They isolate the build and for what? Only to compile everything in sequence <:ev...
ye unrelated 
opaque wharf
postgres really have everything under the sun
opaque sigil
oh also cursed fun fact, apparently hydra is mostly perl
opaque sigil
the first commit is from 2008 so i guess people never really felt the need to change
seems to be robust enough to handle nixpkgs so it can't be that bad 
opaque wharf
I think it is still not as cursed as my old robotic team tooling. It is a CLI script
in PHP
opaque sigil
php for cli is comfy
opaque wharf
What do you mean comfy?
opaque sigil
i don't mind writing php at all it's kind of like js to me, comfy
(doesn't mean i will)
tender river
oh also cursed fun fact, apparently hydra is mostly perl
perl also used by some standard nixos utilities
opaque sigil
aren't all the nixos-* scripts in perl or something
opaque sigil
(doesn't mean i will)
this is a lie actually, i've been wanting to play around with bun's shell module for scripts
tender river
aren't all the nixos-* scripts in perl or something
many are in bash
opaque wharf
this is a lie actually, i've been wanting to play around with bun's shell module...
If you want to interface with hardware, then it's not ready yet. Some stuff like serialport can't be accessed just yet. I think canbus should be fine if you mount the SLCAN as network interface
opaque sigil
or that was a separate project idk
opaque sigil
If you want to interface with hardware, then it's not ready yet. Some stuff like...
this feels like it's not addressed towards me
opaque wharf
Well to be fair, I don't know what else will not work. But you can see the list of unimplemented libuv here
https://github.com/oven-sh/bun/issues/18546
opaque sigil
okay looks like nixos-generate-config is the only perl one
huh what does libuv have to do with anything
tender river
okay looks like `nixos-generate-config` is the only perl one
out of nixos- ones probably, there are also internal script too
opaque sigil
i was talking about this https://bun.com/docs/runtime/shell
tender river
like the bootloader install scripts
opaque sigil
rough bloom
install-grub is perl apparently
systemd-boot isn't listed in the perlless profile 
opaque sigil
or that was a separate project idk
https://github.com/nix-community/nixos-cli tis the one i was thinking of apparently, idk how useful/usable that is
rough bloom
install-grub is perl apparently
systemd-boot isn't listed in the perlless profil...
it's Python
tender river
<https://github.com/nix-community/nixos-cli> tis the one i was thinking of appar...
if you look inside the nixos scripts, most of them are ridiculous things like having to deal with both flakes and non-flakes, and ore reasonable things like error handling, but if you strip them down to their essence each script is like 3-5 commands
opaque sigil
makes sense
tender river
like this, so much code for just running one command
warped narwhal
something feels off here, but I can't quite put my finger on it
mossy fossil
softwaregore moment
opaque wharf
Can I yoink a frame using VLC during playback?
Ohh, so we can
Why is it not on tools, who knows
opaque sigil
https://github.com/Zaneham/BarraCUDA certainly a choice limiting yourself to c when cuda is effectively a c++ superset
opaque wharf
Hmm, how does it compare to ZLUDA?
opaque sigil
this is a compiler, zluda is meant to run compiled cuda as is
by replacing libcuda
i also don't see the point when hipify exists and can already do 90% of the work
but hey, I'm not complaining this is still cool
olive sable
oh
i see now
1080p30 makes core 6 go to 96.7% usage
which is why it crashes after 10 min
verbal grove
hello programmers
olive sable
the moment the core cant keep up anymore it stops
verbal grove
i have a very serious question for you this evening
are AI accelerators something i should look at? or not, because i wanna make myself a new ai-box soon
olive sable
the moment the core cant keep up anymore it stops
wait nah. that might jsut be the loop not blocking
ye
i changed it to poll, now its only 2%
opaque sigil
are AI accelerators something i should look at? or not, because i wanna make mys...
you mean accelerators specifically for ai stuff?
olive sable
NPUs?
opaque sigil
something like that yeah
software support is usually the biggest problem
may as well just get a gpu
verbal grove
you mean accelerators *specifically* for ai stuff?
exactly
verbal grove
software support is usually the biggest problem
software support wont be a problem in my case
olive sable
from what i know, they're not used all that much
verbal grove
yeah
but for like a mini AI box i can just slap 6 nvidia tesla p40 in with a threadripper and be done with it
(wich are not that expensive funnily enough)
olive sable
from what i know, they're not used all that much
i think that is mostly due to software support. idk if its even worth putting the effort in of getting it to work tho
verbal grove
i think that is mostly due to software support. idk if its even worth putting th...
true
but software support just means 4 hours of having to add software support again
olive sable
"4 hours"
verbal grove
maybe four weeks
ive worked with worse
seriously tho the tesla p40 is not that bad
opaque sigil
the p40 is kinda uhhh
I guess if you really need the vram and don't need that much compute it could work
I mean
if you can get them for cheap then sure I guess
verbal grove
yeah, should work. i mean im not thinking of buying a single one... more like six
verbal grove
if you can get them for cheap then sure I guess
182€
i think thats incentive enough
opaque sigil
mind you the p40 also has no tensor cores
verbal grove
thats a bad thing in most cases but i can work with that
opaque sigil
fair enough
opaque sigil
what about renting
could be worth it if you don't need the cards to run constantly
verbal grove
could be worth it if you don't need the cards to run constantly
hmmm
well renting would be nice if i didnt have to run the cards constantly
issue is for what im doing i need them to run constantly (especially since im doing SNN/risc-v tech research and rapid development with multiple people)
and when im training something itl train for up to a month
opaque sigil
opaque wharf
Honestly, if you're running for months, maybe it's better to rent
opaque sigil
a month on what hardware though
verbal grove
a month on what hardware though
atm it depends on what im training, but most of the time quad 3090ti
opaque sigil
verbal grove
Honestly, if you're running for months, maybe it's better to rent
i dont really know if renting is better than owning
verbal grove
<:neuro7:1129322212265050204>
im looking for something to scale up cheaper, 3090ti's that i can find for example on ebay... or something amazon are expensive
opaque wharf
atm it depends on what im training, but most of the time quad 3090ti
Ohh, yeah then that means it's better to buy lol. Because it could get expensive real fast
verbal grove
Ohh, yeah then that means it's better to buy lol. Because it could get expensive...
yeah
most cards i have are bought from amazon but uhh...
a bit much
olive sable
amber fractal
a bit much
If your going P40 hit up ebay
olive sable
<:catdespair:1087521982817509426>
in the FUCK??????
amber fractal
$190USD
verbal grove
If your going P40 hit up ebay
yeah thats what i was thinking
182€ for those
i can buy em in bulk for example
amber fractal
I did so years ago
verbal grove
they're good accelerator cards
olive sable
<:catdespair:1087521982817509426>
ofcourse im not getting 1080p60 output, the resolution it chooses to run the hdmi on is seemingly fully random and ever-changing 
verbal grove
slap em in a large GPU server and make money of em if you want to
verbal grove
ofcourse im not getting 1080p60 output, the resolution it chooses to run the hdm...
rip
maybe i should look at 4070ti prices by now
since i want to buy them in batches of six like the rest of the towers i have
amber fractal
I have used these p40s for exactly nothing, and I now that I'm actually developing a system; I'm better off running on a smol cpu than GPU.
amber fractal
I should preface, I'm not even looking into SNNs or any other alternates that people know of
GPUs genually suck at what I'm doing
verbal grove
¯_(ツ)_/¯
if you wanna slide those p40's over il use em for a jellyfin server
or graphics rendering
verbal grove
k
olive sable
<:catdespair:1087521982817509426>
signal integrity issue?
this is an 8K60 hdmi cable so 1080p60 shouldnt be a problem
and it shouldnt be an issue with the 3090's hdmi port either
ok, new idea
windows
amber fractal
o7 to your sanity
burnt marsh
https://github.com/alyfrillnya/NeuroPlane
anyone have tips on how i should do this?
all i know atm is writing a plugin with the xplane sdk and letting neuro control that with her sdk/api
jagged turtle
https://github.com/alyfrillnya/NeuroPlane
anyone have tips on how i should do th...
abstraction, abstraction, abstraction
burnt marsh
abstraction, abstraction, abstraction
elaborate pls
unkempt citrus
Being an LLM, neuro isn't able to do real time actions properly
burnt marsh
Being an LLM, neuro isn't able to do real time actions properly
i see
unkempt citrus
Usually for something like this you would add a autopilot or something
and then have neuro choose what the goal of the autopilot is
for example cruise, turn
head to this location
crash immedaitely
do a barrel roll, etc
burnt marsh
welp that repo is going into the trash 
jagged turtle
why
burnt marsh
because possum said shed have to say a goal, and thats not really my goal with this project, like i want her to be able to have full control, which is not possible as im being told
unkempt citrus
or raise flaps, but its not going to be able to do it in real time very well
jagged turtle
yeah
burnt marsh
yeah, real time is the issue
and if it doesnt work well it likely wont ever be used by vedal
unkempt citrus
well that depends how chaotic it will get
jagged turtle
if you designed it to be like a copilot, it'd probably work better so that neuro can mess around while vedal steers the actual thing
jagged turtle
and if it doesnt work well it likely wont ever be used by vedal
most likely it'll come down to content
unkempt citrus
Hollow Knight integration yeah provides some limited access to Evil, while Vedal does most of the game play
burnt marsh
unsure if that would make good content, but
unkempt citrus
RImworld integration is similar, Vedal is supposed to play, Neuro becomes the story teller and fucks him up
burnt marsh
would the swarm want to watch vedal and neuro fly a plane
probably not
sigh back to the integration drawing board
jagged turtle
*sigh* back to the integration drawing board
some games have been requested to have an integration by one person in particular in the integration thread #1336984506489376779, though since we don't usually do requests (it takes a fair amount of time to design an integration) they just are left there, you could look into them if you want
burnt marsh
i see, ill take a look
olive sable
windows
same issues
in windows it also makes using your pc miserable cuz every second it does the "new monior connected" thingy where it zooms out all your screens
amber fractal
fiery anchor
oh no
olive sable
lemme try something
amber fractal
Something I got reminded of with the 4k capable port and 1080p capture max is how certian cards can passthrough 4k but still be limited in what they can actually capture
olive sable
when i feed it 4K30 it still works, but thats cuz its doing 1080p30
idk if im sending it 4K and it changes it to 1080p by itslef, or if my pc sees the edid or something and my pc decides to do 1080p
but no 4K is reaching it
amber fractal
it is probably doing 1080p itself
The real question is can it clock higher than 60hz, probably not
as nothing of it's type needs to
olive sable
the output hdmi to my monitor does 144hz
its purely the input thats fucked
im updating the board firmware now
still broken
amber fractal
input has to go into a chip to be processed before being sent as an output. The bois can downscale as I see on a lot of standalons cards, but that is probably what your fighting
sage crag
topipoprrr
kind fable
Omg this is amazing
https://youtu.be/btUbcsTbVA8
i'm sure we can use that to make AI understand more things
jagged turtle
embed failed
cobalt sigil
just chat more to get embed perms
kind fable
well i wasn't here for long
cobalt sigil
kind fable
i like neuro
olive sable
still broken
it would seem liek they marketed 4K input as a hardware capability, but their drivers and kernel dont actually support it well.
so in other words i would need to make my own kernel for this thing
hollow matrix
kind fable
let me guess
written by a bad dev
hollow matrix
I assume
burnt marsh
let me guess
written by a bad dev
or vibecoded
kind fable
or vibecoded
ah, approved by a bad dev
sage crag
sage crag
no
olive sable
kind fable
right now would be the moment where i link to my website where there's a tool which i vibecoded which is good
but sadly oracle don't let me pay them so i can't spin my VPS up...
burnt marsh
you like writing kernels right?
modify the linux kernel?
azure lynx
they care about your privacy. a lot. they don't want you sharing your personal details period.
olive sable
modify the linux kernel?
modify the orange pi 5 ultra version of the linux kernel
kind fable
they care about your privacy. a lot. they don't want you sharing your personal d...
actually that is such a good idea
i'll make a cookie banner on my website where if you accept it, it kicks you, just to teach people to refuse cookies
olive sable
modify the orange pi 5 ultra version of the linux kernel
range Pi 1.0.0 Jammy with Linux 6.1.43-rockchip-rk3588
olive sable
actually, pay me a million usd
a hdmi to usb3.0 converter is 16 bucks. so if you can do it in 15 bucks sure
fiery anchor
that isn't even 15min worth of professional work.
fiery anchor
hobby.
fiery anchor
oddball board, rare usecase... yeah, stuff like that can happen.
maybe retry tommorrow with a full night sleep.
kind fable
sleep is not real
fiery anchor
sleep is temporary coma
olive sable
it would seem that hdmi usb devices also suck
altho they do get to 60hz, they dont go beyond 60hz
azure lynx
4k60 is a lot of bandwidth
fiery anchor
you use the input to stream from one pc to another, yes?
olive sable
from pc to pi, and from pi to projector
olive sable
i ideally get 570x1140x176
without custom edid ill probably have to do 720x1280x180
which is still a lot of bandwidth
fiery anchor
ditch the hdmi in and use a network based protocol.
fiery anchor
there are hardware encoders / decoders. h.264 via udp or sth. like that might work
olive sable
720 × 1280 × 180 × 3 ≈ 497 MB/s ≈ 4.0 Gbps
we can probably compress that a bit tho
if i use GPUDirect RDMA i can do cuda directly to a nic in a different pcie slot
actually, scrap all this.
i have just thought the biggest brain idea
if we use lan, we dont even need to send video frames
we can just send our voxels at only 30hz
amber fractal
voxel lookup table
fiery anchor
and the big pc does the heavy lifting?
olive sable
ye
and the pi still uses its gpu for the fragment shader
we don't even need lan i guess, we can encode voxel data into images
i just don't know how
hallow dagger
what are we cooking if i may ask 
olive sable
volumetric display
hallow dagger
ooooh
fiery anchor
sounds like an simpler architecture, pi is degraded to simpler device with less responsibilities
hollow matrix
Wait, did wacom put this into production in some sort of debug mode?
amber fractal
we don't even need lan i guess, we can encode voxel data into images <:FOCUS:116...
I'd guess one would want 3d aware compression, but who fricking knows
azure lynx
Wait, did wacom put this into production in some sort of debug mode?
probably for the same reason the code doesn't work. Gigya does "Customer Data Platform Solutions".
they wanted to know why it wasn't working ;]
hollow matrix
they wanted to know why it wasn't working ;]
And yet they push to prod with it in a broken state
olive sable
with a voxel height of 160 we have rn, we would need 570*1140*160 bits to send tot he pi
azure lynx
they probably thought "most people don't want spam from us"
amber fractal
I'd guess one would want 3d aware compression, but who fricking knows
Depending on what step you transfer, maybe take advantage of you masking off sections and at least compress some or none.
hollow matrix
they probably thought "_most_ people don't want spam from us"
I think you get a year of clip studio paint if you somehow complete this
fiery anchor
we don't even need lan i guess, we can encode voxel data into images <:FOCUS:116...
fuck knows how to do that, but here is a reasearch paper that sounds like that: https://www.researchgate.net/publication/383659186_VOXEL_VIDEO_STREAMING_OVER_WIRELESS_NETWORKS
azure lynx
they might have a customer support number to call
olive sable
with a voxel height of 160 we have rn, we would need 570\*1140\*160 bits to send...
12.996mb / frame
at 30hz thats 389.88mb/s
a loooot more efficient then the stuff we were doing before
amber fractal
Depending on what step you transfer, maybe take advantage of you masking off sec...
can one transfer the bitmask
hollow matrix
they might have a customer support number to call
Don't need something like that enough to be bothered, it's just shit that it doesn't work
azure lynx
yeah... i've had to hack a company's support form to make it actually work so I could ask them for the open source stuff they used.
olive sable
<:hmm:1303107376295186524> can one transfer the bitmask
we're sending data from the gpu to the pi, so im hoping to still use hdmi. so it depends on if you encode a bitmask in the image i guess
azure lynx
yeah... i've had to hack a company's support form to make it actually work so I ...
which was like step 3 of the rabibit hole in doing something only tangentially related
hollow matrix
Don't need something like that enough to be bothered, it's just shit that it doe...
It gives this warning as soon as the page loads, not sure if this has anything to do with it, but it's probably not supposed to do this
amber fractal
we're sending data from the gpu to the pi, so im hoping to still use hdmi. so it...
The real question is if that makes sense to offset
olive sable
id have to do the math
amber fractal
I'd imagine hit detection is most of it, but one still needs to apply coloring
azure lynx
It gives this warning as soon as the page loads, not sure if this has anything t...
i don't speak SAP. but probably not.
olive sable
I'd imagine hit detection is most of it, but one still needs to apply coloring
the fragment shader stuff you mean? im putting that on the pi
probably
idea rn is to keep the compute shader on pc but the rest on the pi.
hollow matrix
It gives this warning as soon as the page loads, not sure if this has anything t...
I also did try just submitting the form through the console but it just reloaded the page
fiery anchor
hmm. wouldn't it be simpler if you do all the 3D stuff on the pc, render multiple frames into a single image (comparable to a sprite atlas), transfer those with a lossless compression over network, and let the pi encode those into the high frequency images for the projector
amber fractal
Yeah alright, then we can just send bitmask from the compute shader.
Question, would it also make sense to deal with color compression as a compute shader?
that way the pi just has to map
fiery anchor
ye, and could do all the timing stuff
olive sable
i think the pi should be more than capable of doing the slicing stuff.
if we do that on the pc we will need to increase from sending 30hz data to 180hz data
which hdmi would not allow so we'd have to use lan, and so id have to buy a nic and learn some cuda
fiery anchor
couldnt you pack 6 images of the 720p stream into a single 4k 30fps stream? or am i missing a dimension
amber fractal
1080p60 is max
olive sable
nope
amber fractal
olive sable
im pretty sure 1080p30 is the better one of the options in raw bandwidth. and then we dont have to split data into 2 seperate frames
fiery anchor
even with a network stream instead of the hdmi in?
amber fractal
azure lynx
https://cdn.discordapp.com/attachments/1061113840852533339/1456314196655472843/S...
that remnds me of me... except instead of running in place i was kneeling on my chair, spinning it round and round.
olive sable
network stream is limited to 2.5Gbps. which is more than the hdmi input. but would require buying more hardware cuz i cant encode to lan rn
amber fractal
let me do some math on voxels
sage crag
that remnds me of me... except instead of running in place i was kneeling on my ...
topipoprr