#programming
1 messages · Page 145 of 1
why is is literally impossible to install nmap 😭
i get compile errors
and the executable installer literally doesnt even download (the link for it does nothing)
and its not officially on homebrew
actually i think it might be
whats your os?
i would like to present to the esteemed members of #programming
18153 FPS
at this point, it is not possible to go faster on this pc without extra overclocking
i have done it all
i can actually say that now
Nope. You should still be able to get more
litterely impossible
Go raw asm and skip hyprland 

the only way to potentialy do that is raw machine code
any suggestions on other ways to design 3d models for a 3d printer?
Tinkercad works but it feels very clunky
Btw Sam, you're going to college right? Time to learn TeX or typst 

Blender
If you need CAD, then onshape provide a free tier
Fusion360 too
let me be clear, this is the most terrible code ive ever written.
thread-safety? in this house we do thread-slavery.
race conditions? you just gotta catch up bro.
making sure the gpu is ready with the previous frame has been thrown out of the window
how fast is your clockspeed on 100%
ive seen 4.68 at the highest
at idle it did 5.14 cuz i forced max performance, kinda scary
I have faster clockspeed 
which would mean it would probably run faster on your pc
besides the fact i locked it to core 1 and 2 cuz those were the fastest
It doesn't work like that
There are more things in life than clock speed
Unless you have the exact same CPU model
does your GPU support 7 frames in flight tho? cuz thats how much i ended up making it do
and the amount of framebuffers is that + 1 so a grand total of 8
ussualy a game does 2, maybe 3 
in this case it is mostly the clock speed
maybe ram for the vulkan stuff, but i cant really see that
i just see core 1, which is the renderthread, doing 99%, and core 3, which is unallocated so probably something vulkan picked, doing 89%
WHAT?! I thought clockspeed is everything
this is doomed
most of the renderthread has been commented out. at the cost of "potential race conditions" with the gpu 
i aint waiting on those fences
bro trying to max out the cpu on a fucking triangle

i think an igpu might actually be faster on this since its closerby
the gpu is doing jackshit anyways
im getting 40% usage max so the ps5's apu should be able to still handle it
idk
cuz power has to supply the igpu instead of cpu cores
i think potentially the high cpu usage might be the cpu just checking if the gpu is finally done
lemme profile this shit

not even gpu, just bus speed
or bus latency?
latency is probably more acurate
I’m trying to do some work on my art program but there’s a cat on the desk in front of my monitor, a cat trying to sleep on my shoulder and a cat that tried to sit on my lap.
make a vulkan program that directly call hardwares, instead of using the OS as middleman, so it's way way faster 
idk how
cars


but vani said 1 neurillion
are you gonna buy liquid hydrogen to attempt world record overclocking and run your little triangle program

Might as well do them on the E cores if they exist as well

i have 16 full cores here
:waiting:
maxing out efficient cores is crazy
mad insane
Could do frame buffer 
don't question how you get the log numbers
it's not making a game engine anymore, it's having as many fps as possible 
from now on, make a new branch, and don't care about extendability but only fps
while most studios focus on making games as fast as possible
here at bwaa studios we focus on making frames as fast as possible
*slow
depends on how you interpret the frase lol

its private but i can make it unlic
or can you share private prepos still?
cuz tbh this would not look good on my cv
Learning how to minmax mhm
Maybe?
nope
unless you share your github tokens 
just don't include it in your cv 

one branch for the actual game engine and one for the funny
I need one of these commit messages to be "The autism has hit critical speed"
sure companies won't care about your little dev branch
Surely, they also won't care about the people forking for even greater gains

it means your project is popular for sure
is this the actual name you gave for the repo or you changed it
it used to be called "desktop" cuz thats the folder its in on my linux


the commit message is just "18K fps" 
yep
-# 
oh i forgot i did that profraw thing
its in there too now
idk what it actually does but i googled stuff to make the vulkan faster and that supposedly does something
dont ask me what tho
can dlss work with 20k fps
idk
super sampling a triangle 
Maybe FSR can if you can force enable it
lets start at, how do we get dlss on here, and wouldnt the overhead of using it make it worse?
It's at least public now 
I expect a leaderboard soon


i expect quack to pull some threaripper pro max ultra 16900kxdx5D OC out of his dealer to win
Threadripper is bad for single core performance tho
and thread counts
not if you render with all the cores
vulkan should supposedly support that
The cats are finally leaving, but I have stuff to do now so I still can’t work on the art program.
aren't you doing the rendering on gpu
yes
Then invite the cat back
this is extremely limited on bus speed ngl
Add your submission to the leaderboard 
but you need cpu cores to send commands to the gpu and stuff
vulkan can supposedly do that async and stuff with multiple cores
one with smallest motherboard wins

Testing on laptop
pcie slots right next to the cpu
place your pcie slot in the cpu
But what is this leaderboard for?
make the program small enough to run on L1 cache so we don't even need ram
Wait a second, let me ask my sibling to borrow a Legion GO rq
oh ye, the makefile has a "make debug" option. i completly removed all of that functionality
why? i cant remember tbh, somethign to do with me removing the runtime errors
who needs debug if we have brains
Hardware aquired
the compiled project rn 161,600 bytes. shouldnt that already fit inside L1 cache?
There’s still the cat that was on my shoulder, he’s on my lap now. He got lonely for being locked away from me for a few hours today so now he’s clingy.
oh nah, its split into the seperate 16 cores. i only have 64KB per core
if we want it to fit inside L1 we need to find a cpu with more L1 or just nuke more than half the filesize
oh I always have that on
thats fine as long as you dont need to throw anything
I'll be taking the windows debuff
cuz without it makes my std lib on msvc buggy
what
someone should bring your std lib to a hospital
oh ye i remember now why i remved debug
//if (enableValidationLayers && !debugManager.checkValidationLayerSupport(validationLayers)) throw std::runtime_error("validation layers requested, but not available!");
this line has an if with a throw
git isn't installed :RAGEY:
and no runtime errors means they cut of my arms, no throwing
i could just keep the function and not throw, but it kinda feels like having a debug defeats the point of removing all the excepts
ok enough yapping
@olive sable I may be blindge but what deps are being used here
Time to fast track how to properly set up a dev enviroment
i wonder if i could jsut make another renderthread
it will probably immediatly crash since its usign the same recources as the first thred, but it would be interesting
Just don't write 
yep, immediate core dump 
lets ask chatgpt since i cant be bothere to google how to properly do this
chatgpt: "dont"
Short answer: you’re overwriting a live std::thread with another one. That triggers std::terminate() immediately — hence the crash — and even if it didn’t, having two threads both drive the same swapchain/semaphores/fences is invalid and races like crazy.
and races like crazy.
i am somewhat of a racer myself you see
instead of racing myself i make my code race
max verstappen
i was typing something else first but ye
"stappen" is dutch for walking, and "ver" is mostly used to signify an increase/intensifying
so his last name litteraly says "gotta go fast"
ver can also be used negativly but we wont talk about that


ok i figured out while vulkan allows parallellization for a lot of things.
one window → one surface → one swapchain, so only 1 core is allowed to ask for the actual swapchain submiting
ther emight be some deps i forgot about, idk
it might also just be my makefile beig set up for linux
I just need to figure out how to install deps
I'm already running git bash for make at least

what this does mean, is that for VR, you could use a seperate core per eye, and one for the window on your monitor
but since im not doing that rn im sadly limited to 1
hmmmmmmmm
if i make 16 windows, and each one has their own 18k, could i call that 16*18K fps?
Up to the judges
genuine intel
hi vani
glad to know its not like, fake intel
Hey @true hemlock 
I have another benchmark for you
vani, im not at 1 neurmillion fps, but i have reached 18153 fps
im not bound by NDA so 
suspicious intel
we love make (e=2): The system cannot find the file specified.
?
oh thats either clang not being installed, or it cant find the shaders
its probably fine to use gcc tho
Probably need clang yep
actually i wonder if gcc might be faster 
Sam released the max fps on a triange and I'm making a leaderboard
my pb is 18153 fps, apparently we're making this a leaderboard
i could put it inside a readme.md file
gonna try lol
#programming is getting weirder and weirder, I swear
what's the leaderboard rn
Just sam
leaderboard
sam - 18153 fps: linux CachyOs, 5950x, rtx3090
Sam at 18153fps
im not gonna gitpush yet since its really pathethic to be the only one on a leaderboard
Add OS

Not the first time
I'm the top, the first and only

approx this week im building @umbral thorn a pc with a 7950x so she can also tag along

wait is there any dependencies im missing lmao
you should just need vulkan, volk and sdl2
Nope, it is not found for some reason
ah aight
you might need some windows specific stuff in the makefil
Maybe, but I at least need to get it to error in the first place
-lvulkan might need to become vulkan-1on windows
Finally, dep error
might just try later because i might need to go soon though lmao

Now I can again suffer with deps
or maybe if there's precompiled build for it
nope my 13900 system is on windows lol
well shit
wait does the rendering support multi threadding
kinda
might wanna try on my threadripper aswell, its on linux
we use a total of 2 cores
damn nvm then

Prediction suceeded
"We support multithreading, with grand total of 2 cores." sounds funny
was gonna say i wanna try on my epyc 9655 aswell but damn
i wish i could use multiple cores for rendering, but the thing that takes up 95% of the rendertime is only allowed to run on a single core
vulkan is a lot better for multithreading than any other graphics api, but specifically the swapchain is only 1 per window
How is your pc build ?
Day @true hemlock
so the world record now is 0.0550873ms per frame
Update ure bios
(Most risky work)
oh ye i kinda hardcoded the cpu cores it uses since that was the fastes on my pc. it should use core 1 and 2
starting at 0
this is pretty beneficial for intel aswell cough e cores cough

Intel i9 like something processor
LGA 1700 socket
its fine
its an engineering sample, duh
so it'll show up as that
One minor problem
(Neutral mixed line wire )
I will fix BYe
here's the part where github descends in quality

i wonder who keeps killing engineers and looting their samples
ya ask? 
gg microsoft is going to push 100+ worst integrations with their products now

its a rare drop too so he must really be farming them

they're gonna push microsoft copilot-level quality of features into there 
Okay, I think I'm far too dumb to get this working.
does it work with wsl
I can't even figure out how to get the libs to link
should probably release a precompiled build if you wanted server wide leaderboard 
probably

this is my linux build
totaly non-suspicious file
Virus (real)

It's just a flavor of archive
its small-ish
what did you pack in
doesnt fit in L1, but probably does fit in L2
well uh... main.o vk_frames.o vk_command.o vk_device.o vk_sync.o
Where is a usb, I'm loading up an iso
and volk + sdl
does static linking increase execution speed
doe sit?

hmmm
doe sit
Static linking produces a larger executable file than dynamic linking because it has to compile all of the library code directly into the executable. The benefit is a reduction in overhead from no longer having to call functions from a library, and anywhere from somewhat to noticeably faster load times.
it might

i have 512kb per core, less but also fits
if i use static linking it will increase in size tho
only 36MB L3 for the whole chip lmao compared to 5950x 64MB
The speed difference will be nearly imperceptible

so it does
we're talking about 0.05ms here 
Any speed benifit is a benifit
dynamic linking adds a bit overhead, of course
"nearly imperceptible" maybe if you're a plebian running at sub-1000fps. but here every microsecond counts

something something i9 14th?
but yeah even if you scaled it down to 80kb it probably still won't fit lol, other processes still takes the cache
i7 14700f
it definitely will fit any L3 though
make an operating system that only renders a triangle 
which is more than enough
I find random cpu before this drive
if they can make games under 1kb, so can us
i doubt most games under 1kb used vulkan tho 
by using two cores instead of one, you decrease the load per core by 50%
by using three instead of two, 16.(6)%
what
the math is worng but i get what he means
but it isn't
ah yes
m.2 rocket drive will have to do
https://hackaday.io/contest/18215-the-1kb-challenge
they made a standalone brainfuck computer

anywas here im just using 1 core for handling the swapchain, since that is the only thing i cant multithread about vulkan, and another core for handling the rest. that 2nd core only reaches about 6% on my pc tho 
should we allow custom hardware and shit or it's limited to modifying the original code, and has to be on vulkan
So we need that core to be pegged at swapchain then
yeah time to build my own ASIC with 1kb cache
im fine with custom hardware, but i hope nobody here starts amking a core 2 due powered by the sun
only 2 cores, but 5 bilion GHZ
I will do it
you'd need 3 cores probably tho
in theory you can make an ASIC with really simple pipeline and make it clock like crazy

The theoretical speed limit for a computer clock is limited by the speed of light
rush to ibm and borrow those quantum computers, and the fps will be infinite since it's undefinable 
we've talked about putting the rendering on the cpu, but most of the time is spent submitting and presenting the rendered image. so why not put the program on the gpu?
on point but not exactly applicable for per pipeline
I know. Just talking theoretical
that would require the whole os to be on gpu as well. It has been done but it's kinda buggy and slow iirc
people seriously figured out to use a line of pipeline component in a streaming way
Are we allowed FPGA that renders the exact output?
ok new idea, considering the theory of relativity, instead of rendering more frames per second, just make time for the pc faster than for us.
send the pc through space at 99% the speed of light and the fps will be billions
despite a core being relatively large and seem pretty impossible to reach high clock speed if compared to lightspeed, people fuckin made component queue system to maximize pipeline throughput
if you use vulkan
but it's the fps on our perspective, the fps in the computer remains the same
hmm

science starts with stupidity fr
we need to core that renders to be going at lightspeed compared to the core that measures the time and fps
hitting trees with stones gone hard

LMAO
Quick mafhs shows that a light travelling a planck distance would need 5.39124645 × 10^-44s. Double that and we round it to 1*10^-43
im thinking of a small particle acceleratpr with a core going through it and a core in the centre 
bus speed limit goes real
The maximum theoretical clock will have the speed of 10^43 Hz
shit shit shit
hmm
quantum tunneling
fuck it
entagnlement

entengle both sides of the pcie bus
just do quantum computer at this point
instant data transfer
if we can somehow make planck size computers lol
i like how we were taking this seriously at first and now we're just doing sci-fi shit
Be the change you want to see in the world! 
wait for 100 years so we have commercial quantum computing available
i've kept really thinking of the possibility of this ngl
not only you have practically zero latency
You still need to send a signal for when the computation is done
Portal ahh thoughts
you can do practically infinite size of memory aswell
No, the latency will still be limited by the speed of light
But the bandwidth is limited by how many qubit you can cram
information travels faster than speed of light
the speed of light is a large enough limit imo
It's not 
information as in quantum mechanics
you can have insane bandwidth too if you figured out a signal encoding that works with quantum entanglement
299 792 458 m/s. idk how to convert that to bits per second but im pretty sure a couple million bits fit inside a meter
Yeah, as I said, the latency is limited by the signal that always need to travel by the speed of light. But the bandwidth is limited by how many qubit you can cram
I was always wondering if one day we can have entangled storage rather than compute
quantum entanglement is instantaneous..
That'd require a FS that can handle race conditions tho
10^43 Hz is the limit
I know. But how do you know when to observe and collapse the state?
Entanglement can help ease the bandwidth
But not with the signal speed
oh nuu I took a look at my motherboard and it broke my pc 
pro series
and maybe an AIO
that metal is really chunky
im tempted to get the 2x64gb dimms
where are the armoured ram slots
but im broke so might get 2x32, then get 2x64 later and use the 2x32 for my own pc lol
my pc is still on ddr4
Quantum things just reminds me of how bad Vedal play Outer Wilds 
that's a big southbridge
e
so, there are no restriction towards ram, registers, register sizes and X3D cache? just the ROM?
imagine if the size limit is 1GB
epyc 9684X owners be winning
Failed to create vulkan instance my beloved
1gb limit is non-limit, you can just pop a cd game
i just need the mobo
my current mobo nottlebecking my cpu and ram
specs?
random question: why is it every time i do a ctf challenge i always get a massive headache while doing it
this only happens then
Thats because youre thinking
its a very dangerous activity
you should leave it for trained professionals
r5 3600, 3600mhz ram
like i have NEVER seen my cpu boost higher than ~4.05ghz and ram is at 2933
shit, i didnt know, thanks dawg 👍 won't use my brain from today ownards
Thats the spirit
probably not enabled expo
cpu clock isn't board dependent unless if the vrm is shit. what's the motherboard?
enabled
mobo is asus a320m-r
hold on
it's called Tention Headache
you can try checking the power draw on the cpu under load

@olive sable We love errors on the binary

cachyos because yes
i completely removed error handling and such so makes sense
this is my most "works on my pc" ah code so far
Some vector error, can't create larger than max size
bwaa
i put it at 7 cuz a latency of 7 frames is unnoticable at 18000fps, and its gave me an extra 100 fps
but on laptops its probably not supported
Here I was using the binary so I didn't have to compile

i def should linux my briefcase just for this fr fr
this one has 3
Time to type another url into firefox yippie
It’s usually that they modulate the fuel amount
Wait, there is a flash drive next to me
wasnt the last one 160kb. how is this one now 77?
I'm very stupid sometimes
Why not Safari?
are you serious? safari?
not having a linux port is not event he problem for me
i just dont understand why you'd use safari
Yes, but it hasn’t gotten updates as far I know, I have it installed and it’s hella old
You can always use wine
the none-iphone safari versions all suck, every single one. even for mac its ridled with bugs
Well, Safari doesn’t receive updates as far I see
I use Google Chrome some times
chrome is okay
Actually it’s my main
I got these browsers installed: IE11, Microsoft Edge, Google Chrome, Safari
I was doing that from memory, so I might have missed a few
uhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhhh
It still crashed with the same error

does it give a location or?
Nyope
bwaa
bwaa
I at least have selected GPU console text

Well, the part I’m always wondering with is if my PC starts when the startup screen freezes
First try: nope

the window gets created at the very start, so it could be anything vulkan related
I meant if it does this
Because for some odd reason my pc does this
And another boot fail 
Damn
if the 18K fps version isnt working, the 16K fps version is the first commit into the repo
and that one does have fully intact debug
fine lmao
Sam, I think you need to debug english.lang
Oh wait, I forgot everyone stores languages differently in their head
I beg on my knees to not have to suffer tonight for a meme
i just need to figure out how to compiel an old commit in vscode
but it’s kind of odd that’s just since I installed my new ram kit that the issues appeared
But oddly, it is only on start up
Best bet seems to be pull down a copy and revert a commit
Honestly, I’m not really into compiling C++ Code
Debug info version!
Finally it started
If the issue only started because of new ram then I kinda doubt it’s an os issue.
Dance reliability score. Dance:
Ure safe
Large window, still fails
why is it smaller than the previous executable you sent 
Lovely
bwaa
older version, idk
100kb less 
the window starting means sdl works
oh ye for debug version you probably need the vulkan sdk
"=> Network management is gone <="
Anyways, Back to that.
I have IE11, Microsoft Edge, Google Chrome, Brave, Firefox
@olive sable
I'll check that and if I'm stupid I'm stupid
install opera gx as well
Oh that is just something on Windows that tells how reliable the PC was lately. Idk why it counted my Vacation days though
and all chinese browsers
bro is a collector
you only need one lmao
Windows saying updates 10 pro to 11
Constant
vivaldi
IE11 the goat
Well, I used to have another one that was from someone else too, But Idk where it is
Meh, I'll take my sweet time
Wait, this one is a runtime error
Now I'm going the *.zip search
ERR_CONNECTION_REFUSED
NOMAL for me
if the debug works it should probably give decent description ofhte issue
oh
for that is hould probably compile for debug
im stupid
Let me do crut ure file
~ $ curl get https://cdn.discordapp.com/attachments/1071784467036913664/1406554862430453840/vulkan?ex=68a2e3cb&is=68a1924b&hm=758c9bff6a494d087fabf00c0b33ee73c1c541f06bab8fd2e865317fc6a4e5c3&
[1] 12756
[2] 12757
[3] 12758
[2] Done is=68a1924b
~ $ curl: (6) Could not resolve host: get
curl: (7) Failed to connect to cdn.discordapp.com port 443 after 5172 ms: Could not connect to server
I lost that Browser... Anyways
debug build
yay
i should probably stop sending random files here
the mods might getmad
@amber fractal try this one
got it
But these browsers I don't trust: Opera GX and Opera
Chromium is suck
Sure? 
May you enlighten my why?
Let me reinstall chromium
wha, huh
But... I got Safari:
How is it a file open error
the only files its opening are the shader files
wait
which should jsut be in a folder named shaders
let me fetch those then
you only need to spv's
erf
hi konii
Good morning
It's a triangle yippie
helo
3k is not correct because this is still on battery for some reason
if the full game ran at 3K on battery id be more than happy
even 1k is fine actually
1 fps lmao
better than 0.5
too late, i edited my message before you replied
At that point we should change the metric to spf
4k seems like where this ends, but these render layers geez
I HAD one problem
(I'm always limited by data
First' conver
(Txt to json )
I can put txt as data
i live my 18000 fps spf 
IT WAS NO SHADERS THE WHOLE TIME
1 and 2 are the worst
I need to change to a new pcb byb
(Then login computer)
Then work ai task
Training
somehow
hmmm
so the ones that run better for me?
i didnt manage to static link vulkan or sdl2 or volk so far, but i managed to static link SDL3
lemme try that
Still pcb is not available
Bye i come back 5 mintus
Still finding pcb in my electrition shop
Did… someone say… LAYERS???

Mmmmm, coffee 
Alright, going through the executables in order

because I found a method to actually see the fps
I got volk to staticly link now too. And vulkan just cant be staticky linked byt that should not cause any overhead due to volk
Im not gonna put that on git yet tho
Is the homebrew version of nmap official
vulkan can be statically linked if you want to link all of mesa 
For the love of all that is holy, do not do this lmao
100mb binary
Im not gona do it manually
28000fps???
nice screenshot
Didn't want to fps drop
Laptop gpu's must be better for this with the lower latency to gpu then
its probably a combination of bandwidth and latency
Ok so 26377 fps on legion go
want me to run it on my igpu?
Make sure to grab shaders
because man was that fun not being bundled
Uhhh. The code automatically chooses dedicated gpu ig you have one
Hmm
Either go into the code and make that +1000 points negative, or disable your gpu
nice one
{"isFine":false}
the joy of setting up a dev environment

Should I do an env on Python?
leaderboard
Sam - 18153 fps: linux CachyOs, 5950x, rtx3090
Iggly - 30573 fps: linux CachyOs, Z1 Extreme, AMD Phoenix (intergrated)
Strange, when I'm not pulling up the kde overview the fps number starts with a 3
obviously 
not that i have the same thing as you, im using nixos
let me install a waybar rq
Well, I didn't do that. I just do pip install <Package>
It's not a wayland session frick
Q: Does it really need to be Wayland?
it permanently alters your environment
i like how the only copy of that image left for me is a telegram embed of the discord link
Yeah, that my code can run?
@olive sable I'm changing my submission
but if you want to run any other code
it will have poisoned that environment
Okay, how so?

Because the code imports the packages. Not installing any
your use of italics is very annoying
-# Disabled Italics 👍
anyway, say you had two different projects that used the same dependency, but required different versions
Oh yeah. I'm not making more than my Text Converter.
By the way, I was trying to make the actions that happen in italics
you should make an external program that reads the fps in memory, and take the max(and average maybe) fps in 10s duration
or else it's not accurate
nix-shell -p gnumake clang volk vulkan-volk vulkan-headers vulkan-loader SDL2 glm lld
@olive sable is there anything else im supposed to need
because running make just dies
Well, Italics is now off.
We are already using the title bar, memory probing could work if we can even implement it
Not sure if SDL will allow editing the title from not main thread, I'd imagine that has already been done tho
At this point, floating point errors are more of a concern as well
If this ends up working eventually, send me a copy. I want to do it on my main system as well.
Also grab a bar because KDE cuts off the text
Oh and Sorry konii for annoying you with Italics, May I sparingly use Italics when it fits?
and the overview being enabled reduces fps by a lot
-# note to self, 41k covered up by another window
That is certainly interesting 
nix-shell
gnumake
clang

Apparently, it's Sam who dies now
Do you forgive me konii?
No
- konii, probably
Oh nooo, I turned off italics in the way I write and some times the raw output of how I format text comes through when trying to use italics
Brb
italics italics italics
What’s I mean is that when I try to use italics right now it appears as [italics]this[italics end]
I think italics are okay, as long as they are used properly. With the above case, it's probably a no.
I thought I was using italics right…
And I mean on actions like that
Is konii gonna be mad if I start using italics again?
not if you use it correctly
iGPU go fast 
dGPU only did ~20k (probably doesn't help that display output and compositing currently goes through the iGPU)

-# Enabled italics 👍
what if no gpu at all
How do you accidentally test it?
I was swiping the words
messed up the GPU scoring function enough that it chose the software renderer
erf
@sage crag @amber fractal flake
{
description = "Description for the project";
inputs = {
devenv-root = {
url = "file+file:///dev/null";
flake = false;
};
flake-parts.url = "github:hercules-ci/flake-parts";
nixpkgs.url = "github:NixOS/nixpkgs/nixos-25.05";
devenv.url = "github:cachix/devenv";
};
outputs = inputs@{ flake-parts, devenv-root, ... }:
flake-parts.lib.mkFlake { inherit inputs; } {
imports = [
inputs.devenv.flakeModule
];
systems = [ "x86_64-linux" "i686-linux" "x86_64-darwin" "aarch64-linux" "aarch64-darwin" ];
perSystem = { config, self', inputs', pkgs, system, ... }: {
devenv.shells.default = {
name = "my-project";
packages = with pkgs; [
vulkan-headers
vulkan-volk
SDL2
glm
lld
];
stdenv = pkgs.clangStdenv;
enterShell = ''
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$EXTRA_LD_LIBRARY_PATH"
'';
env.EXTRA_LD_LIBRARY_PATH = pkgs.lib.makeLibraryPath (with pkgs; [ SDL2 gcc-unwrapped.lib ]);
};
};
flake = {};
};
}
doesn't include the Vulkan loader because that's not required anymore, remove -lvulkan from the LDFLAGS in the Makefile
im literally a tv

a classic nix shell with shell.nix or something probably works, idk just make sure it uses the clang stdenv, it makes stuff easier 
I use flakes because it's convenient to do nix flake init -t github:cachix/devenv#flake-parts
mabe its just me who thinks flakes are unnecessarily complex and verbose
I can see unnecessarily verbose but not complex
their structure is really simple, you have a list of inputs that get mapped to an output attrset through a function, that's pretty much it
Now I wonder what would happen if the thing decided to run on my 3090 which has no display attached to it and runs on a x4 chipset slot
comparatively to a shell.nix file
true, though those also do a lot less
please add experimental-features to systemwide nix.settings 
(also nix-shell isnt a flake-related command so it wasnt needed here anyway)
why use devenv and flake-parts 
because it's convenient to do
nix flake init -t github:cachix/devenv#flake-parts

if you require a single non-nixpkgs input its generally easier to do a flake
(or if you require multiple outputs)
i did after this, also it was needed here i dont know why

{
inputs.nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
outputs = { nixpkgs }: {
devShells = nixpkgs.lib.genAttrs [ "x86_64-linux" "x86_64-darwin" "aarch64-linux" "aarch64-darwin" ] (system: let
pkgs = import nixpkgs { inherit system; };
in {
default = pkgs.mkShell.override { stdenv = pkgs.clangStdenv; } {
buildInputs = with pkgs; [
vulkan-headers
vulkan-volk
SDL2
glm
lld
];
shellhook = ''
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$EXTRA_LD_LIBRARY_PATH"
'';
env.EXTRA_LD_LIBRARY_PATH = nixpkgs.lib.makeLibraryPath (with pkgs; [ SDL2 gcc-unwrapped.lib ]);
};
};
};
}
also i agree with konii shell.nix would've been better here
you could still use flakes using import ./shell.nix { ... } in flake code
born to watch horse anime forced to read horse blog https://ngnghm.github.io/index.html
Houyhnhnm Computing
epic naming
yeah I get it, a flake is way overkill for this use case of just bringing a few packages into the environment
it's purely because I kinda standardized on this for everything 
please at least add flake-compat for non-flake users for what you release publicly
public flakes are different, ye
{ pkgs ? import <nixpkgs> {} }:
pkgs.mkShell rec {
buildInputs = with pkgs; [
lld
clang
gcc-unwrapped
glm
SDL2
vulkan-volk
vulkan-loader
vulkan-headers
];
stdenv = pkgs.clangStdenv;
shellHook = ''
export LD_LIBRARY_PATH=${pkgs.lib.makeLibraryPath buildInputs}:$LD_LIBRARY_PATH
'';
}
this works
9k
lld/clang/gcc-unwrapped should go to nativeBuildInputs, also stdenv = pkgs.clangStdenv doe nothing, you have to mkShell.override { stdenv = pkgs.clangStdenv; } { ... } instead
{ pkgs ? import <nixpkgs> {} }:
pkgs.mkShell rec {
nativeBuildInputs = with pkgs; [
lld
clang
gcc-unwrapped
];
buildInputs = with pkgs; [
glm
SDL2
vulkan-volk
vulkan-loader
vulkan-headers
];
shellHook = ''
export LD_LIBRARY_PATH=${pkgs.lib.makeLibraryPath buildInputs}:$LD_LIBRARY_PATH
'';
}
also works

i dont like the LD_LIBRARY_PATH here
gross
i hate how i have to do it with rust sometimes too
the compiler just expects the library to be there without setting rpath 
pkg-config 
LD_LIBRARY_PATH=/run/current-system/sw/lib
protip never add emscripten to nativeBuildInputs
emscripten heavy
emscripten needs a diet
actually this is more useful
why on earth do you need nodejs and openjdk

oh you got it to work?
Apparently they use a closure compiler for optimising size 
add me, konii - 8894 fps: linux NixOS, 7600, amd raphael (integrated)
Now I wonder how I'll fare with lucienne
your cpu is worse than mine why is it beating me by 22k fps
Oh, I thought that was a reply to Sam 
wait how did you guys got it to work
AMD Graphics (Ryzen 7000) vs 780M: technical specs, games and benchmarks.
do you use nix?
Did you convert from float16 to int?
damn nix
shell.nix
{ pkgs ? import <nixpkgs> {} }:
pkgs.mkShell rec {
nativeBuildInputs = with pkgs; [
lld
clang
gcc-unwrapped
];
buildInputs = with pkgs; [
glm
SDL2
vulkan-volk
vulkan-loader
vulkan-headers
];
shellHook = ''
export LD_LIBRARY_PATH=${pkgs.lib.makeLibraryPath buildInputs}:$LD_LIBRARY_PATH
'';
}
you can try and replicate this on windows or whatever if you want 
what do you mean?
Well, I was reading it as 8.893 FPS... So I was thinking you rounded up.
it doesnt really have anything to do with converting from float16 to int
i did rounding to reasonably close number

Shuni - 27158 fps: Linux NixOS, 7950X, AMD Raphael (integrated)
It was sort of a attempt to make a coding joke...
excuse me

are you running it with make run
no

how are you running it
./output/release/vulkan 
[nix-shell:~/code/very-fast-but-very-bad-code-vulkan]$ ./output/release/vulkan
Selected GPU: llvmpipe (LLVM 19.1.7, 256 bits)
llvmpipe 
did i need to add mesa
I adjusted the GPU selection code to favor the iGPU over the software renderer
git diff?
diff --git a/src/vk_device.cpp b/src/vk_device.cpp
index 1a4d126..e973df0 100755
--- a/src/vk_device.cpp
+++ b/src/vk_device.cpp
@@ -101,7 +101,7 @@ int DeviceManager::rateDeviceSuitability(VkPhysicalDevice physicalDevice)
VkPhysicalDeviceFeatures deviceFeatures;
vkGetPhysicalDeviceFeatures(physicalDevice, &deviceFeatures);
- if (deviceProperties.deviceType == VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU) score += 1000;
+ if (deviceProperties.deviceType == VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU) score += 1000;
score += deviceProperties.limits.maxImageDimension2D;
return score;
will also make it use the iGPU over the dGPU
Hmm... Should I run Minecraft that the rendering runs on the CPU too? (Just for fun by the way)
And I re-iterate: it is just for kicks n giggles
Sam disabled auto CRLF conversion? 
this is a minor improvement
oooo, What GPU btw?
read
okay
(that may take a while)
Can I has thing to test? And how do I test it on Linux Mint?
gamemoderun output/release/vulkan
@rough bloom try running it with gamemode
programs.gamemode = {
enable = true;
enableRenice = true;
settings = {
general = {
renice = 10;
softrealtime = true;
# may break things
igpu_desiredgov = "performance";
igpu_power_threshold = -1;
};
# danger
gpu = {
apply_gpu_optimisations = "accept-responsibility";
gpu_device = 0;
amd_performance_level = "high";
};
};
};
here's how i have it in my configuration.nix
Damn, and all that from the AGP of the CPU
How do I get on funny leaderboard?
makes no difference, probably because I haven't configured it at all (and probably won't because I don't think I need it)
also it's slower now because I have more stuff open in the background
git clone https://github.com/samvanmaele/very-fast-but-very-bad-code-vulkan
well, its more likely that kde's power profiles are messing with it
just opening YouTube or something dropped it from ~28k to ~20k earlier
30 

And now what do I need to do in order to build it?
I use Hyprland and have no power management daemon or anything like that enabled, unless it comes default with NixOS/systemd
sudo apt update
sudo apt install -y clang gcc lld libglm-dev libsdl2-dev libvulkan-dev vulkan-utils libvolk-dev
make run
probably this
isnt this for CFS
Hmmm, I have Internet Explorer 11, Microsoft Edge, Google Chrome, Safari, Brave and Firefox, What browser (except anything opera) do I miss?
Yeah, it was said at #announcements
early strim because Japan time
Neuro will never be as perfect as Evil 
Should I get Twitch Turbo?
Packages: 1387 (nix-system) 
rip
For some reason my head switched to DE_DE (aka German) on the last word...
By the way, Hi 🦆
why tohugh
so i can use my computer while i compile stuff
could use ananicy 
What is this meant to mean?
I somehow misread it while glancing
ananicy-cpp

no one uses the python one
Hello chayleaf
It's definitely not building
in the makefile remove this underlined red part
ancient clang version because Mint 
disable profile-guided optimization like konii said and try again
the fact that cachyos doesnt use system76-scheduler makes me believe its not strictly super useful
exit your terminal and open it again
Did not work
you did sudo apt install -y clang gcc lld libglm-dev libsdl2-dev libvulkan-dev vulkan-utils libvolk-dev right
try sudo apt install -y libvulkan-volk-dev
I installed some modules manually but I believe they should be installed
uhhh
konii do you have the shell.nix 
sick, thanks
if you can't get the system provided package to work then you can try putting #define VOLK_IMPLEMENTATION at the start of src/main.cpp
then remove -lvulkan -lvolk from the LDFLAGS in the Makefile
That made it more broken



his pc





