programming

[olive sable]

semaphores

which are way too slow

[tender river]

wrrr

[abstract bronze]

lmao i was gonna use vscode but then umm tutorial used pycharm so yerp

[severe path]

What are you trying to do, exactly?

[amber fractal]

discord.py
I try to convert people to something else that has docs

[warped narwhal]

like serenity

[olive sable]

we're currently at 126K fps with plans on improving that

[amber fractal]

[severe path]

Rendering what? A triangle, at 125k FPS? Why?

[warped narwhal]

why not?

[rough bloom]

because number go up

[olive sable]

i dont need a reason

[severe path]

Fair

[sage crag]

wrrrr

[severe path]

But...

That seems to be a pointless benchmark

[warped narwhal]

but it's fun

[severe path]

[warped narwhal]

if everything has to have a point then what is the reason to do anything?

[glass flower]

someone should make a clicker game about optimizing triangle rendering

[amber fractal]

may I perhaps tempt you with https://interactions-py.github.io/interactions.py/ with tutorials on the docs site

[abstract bronze]

[severe path]

I never said not having a reason was a bad thing.

I just find it a bit peculiar, in this context...

Why would you benchmark a triangle, when you can benchmark a scene?

[amber fractal]

criminally underrated, but it is my prefered lib for python

[olive sable]

cuz one triangle is less expensive

the goal is to have as much fps as possible

[abstract bronze]

ill try to figure it out using it

[warped narwhal]

https://tenor.com/view/all-your-base-all-your-base-are-belong-to-us-zero-wing-zerowung-gif-23152926

[amber fractal]

this is in the talks for "triangle 2" but unsure how to keep the meme alive for that

[severe path]

But like... One triangle isn't going to utilize the GPU hardware very well.

They are designed with a full game in mind, with several thousands of triangles. I would yield the benchmarking until I at least have one (somewhat full) scene.

[olive sable]

ye but we dont care

fps is higher

so we happy

[warped narwhal]

the point is not to hit 100% gpu usage, the point is more fps

[severe path]

[warped narwhal]

number go up = dopamine

[amber fractal]

I belive most are at 50% usage with being CPU bound IIRC, unless for those using LLVMPIPE, then they are completly CPU bottlenecked

[olive sable]

i dont remmeber what my gpu was at

it wasnt very high tho

[sage crag]

[amber fractal]

I still got better results with iGPU

[sage crag]

im pinned at 90% cpu and 10% gpu if i use mangohud

[amber fractal]

so I can't personally speak on it

[olive sable]

i think gpu usage doesn't show full picture since we're prob at low core usage but high clocks

[amber fractal]

laptop is a fun problem space to optimize

I should update my position once I attempt some of the new tech

[olive sable]

i think we're at like 8 microseconds reported by cpu

[amber fractal]

I'm still pre labwc

[olive sable]

so if the blockign works thats our fps, and if the blockign doesnt work this is a sham

for the sequal im gonna use up-to-spec vulkan code tho to make it more realistic and fair

triangle is pure fps number

next one will be actual gpu benchmark probably

[amber fractal]

uhhhh

I didn't even do anything I needed to yet for more FPS

I was just testing in hyprland

on battery

and my screen lost a ton of fps

and this is 1.5x my PB as is

[olive sable]

on labwc you mean?

[amber fractal]

nope

[olive sable]

huh

thats interesting

[amber fractal]

my screen never hung while triangling so that is what set me off

I did not have the 60fps experince

maybe 10

I was just testing the repo

Get over here labwc

I'm doubling my record

[sage crag]

in theory you dont even need a compositor with sdl

you can do fbdev as the videodriver

to directly blit to the screen

[tender river]

you can just

use a piece of paper

and draw on it

[olive sable]

no

[sage crag]

1 triangle per second

[olive sable]

i dont have a red crayon

[severe path]

Or modeset

[sage crag]

https://tenor.com/view/my-jolly-reaction-jolly-my-honest-reaction-gif-5490890317762380451

are we still doing this

im watching osu clips

[severe path]

Yes :p

[sage crag]

wrong, we is mutual

just you buddy

[severe path]

Aha, what?

[olive sable]

modeset just changes how the output to the screen is formatted no?

i still don't see how its relevant to rendering

[rough bloom]

it's not

[severe path]

Yes.

But before you blit a framebuffer, it has to be tied to a CRTC and modesetted.

actually, I'm not sure if its blitting...

I'm getting too caught up in details... Sorry

For a triangle to show up onto a screen, you have to blit to a framebuffer that has been modesetted.

There is a way to render directly onto a framebuffer, but most opt to double buffer (for obvious reasons, you don't want to see an unfinished frame)

[rigid snow]

i am wait for triangle 2 banana editiio

[sage crag]

it coming soon

it will be

30 lines

and run at

900k fps

[rough bloom]

new requirement
C code must be in shape of banana

[sage crag]

did you see my 76 line banana monstrosity

[sage crag]

OhISeeto

ye

[amber fractal]

preliminary results are that it is a bit better

[rough bloom]

the unreadable one? yes

I saw the pixels

[sage crag]

yee

[rough bloom]

idk how anyone can write that

[sage crag]

simple

write normal and then

delete

[rough bloom]

[sage crag]

if (frame >= F) vkWaitSemaphores(dev, &(VkSemaphoreWaitInfo){ .sType = VK_STRUCTURE_TYPE_SEMAPHORE_WAIT_INFO, .semaphoreCount = 1, .pSemaphores = &sem, .pValues = &(uint64_t){ frame - F + 1 } }, -1);

relatively short line from the busy loop

mlntcandy

ill make this look nice later

[amber fractal]

stupid e-cores wasting power budget

only getting 74k so not the doubling I wanted

[severe path]

Are you doing CPU rendering?

[amber fractal]

Nope, still vulkan on iGPU

[severe path]

Oh

[sage crag]

try running the version of the triangle that i sent

it should be in one of my messages here somewhere or other

[amber fractal]

I legit could not get the thing to compile

[sage crag]

OhISeeto

[sage crag]

not even with the same shell.nix?

were you missing the shader files?

[amber fractal]

Nope I got everything, I was debugging gcc errors for the #embed macro

my nixpkgs too far out of date

Thinking more logically, the next step was to pin that to unstable but I was tired

I love only having one p-core and 2 e-cores in the testing area

[stark needle]

wrrrr

[amber fractal]

too many cores take up too much power that I need on the rendering thread

Let me pull down my own fork of the sam triangle repo for the lols

[rough bloom]

nyoom

[olive sable]

https://tenor.com/view/shaggy-scooby-motorcycle-fast-run-fast-gif-17006267

[severe path]

I'm confused

[amber fractal]

laptop, so limited power budget

[severe path]

Oh

[rigid snow]

i've had the displeasure of backreading

my god

[sage crag]

mlntcandy

[sage crag]

cause

dente

[rigid snow]

ye that me now

[sage crag]

in both sense of the worse cause

[rigid snow]

or always

[sage crag]

cause as in, because, and cause as in, to cause

[rigid snow]

i don't think it because konii

[amber fractal]

3.4k fps

[sage crag]

accuracy

results

power

truth

banana

[rigid snow]

🍌

[sage crag]

notnoted at banana

heresy

[rigid snow]

NO

[sage crag]

https://tenor.com/view/borzoi-snoopa-lord-foog-dog-funny-gif-20125052

[rigid snow]

look my name

[sage crag]

https://tenor.com/view/borzoi-gif-27403233

[trim valve]

aaaaaaaaaaaaaa

I hate soldering

all components should be massive

tiny is difficult smh

[sage crag]

make time machine

[obsidian mantle]

can someone explain the banana campaign

[olive sable]

yellow

[obsidian mantle]

what are your goals

[olive sable]

to be yellow

[obsidian mantle]

meow

what do i need to build a mechanical arm

[olive sable]

arm

mechanicals parts

[obsidian mantle]

what motors does it use

[olive sable]

dc

[obsidian mantle]

some special motors that do precise positioning?

are there alternatives

[olive sable]

potentiometer ye

[obsidian mantle]

main concern for me is noise

[sage crag]

global and total conversion

[nocturne olive]

I just found out my primary GPU is 3 years old now

Bought January 5th 2023

[obsidian mantle]

the 3090?

[nocturne olive]

4070Ti

[obsidian mantle]

40 series is 3 y o?

[severe path]

I did some math: Assuming that you were rendering directly to DRAM, on a GTX 1050, given you had 69GB/s DRAM bandwidth, and a 1920x1080 framebuffer with 32-bit color, the fastest possible rendering speed can be calculated with the following:

1920 * 1080 * 4 = 8,294,400 bytes

69 GB = 74,088,185,856 bytes

74088185856 / 8,294,400 = 8932.314 FPS

You can't get a triangle to render much faster than that on a 1050, for example. You are hitting the limitations of DRAM.

That is my conclusion.

Correction: VRAM, not DRAM

[rough bloom]

if you need precise motors then AFAIK the way to go is either stepper motors or some motor + sensor/encoder + control loop combination
me glueless about robotics though

[severe path]

Actually, my math is wrong... One second

Nevermind. Its right

I thought for a second I was confusing a unit...

[nocturne olive]

Do note that the whole compositor won't need to update at the same speed as the triangle does, so only the size of the triangle window really matters bandwidth wise

[rough bloom]

1920x1080 framebuffer
not the case
the framebuffer is something like 3x3 IIRC

[severe path]

No.

That is litteraly how fast a GPU can render to VRAM/any framebuffer at all. It is a hard limitation.

[amber fractal]

6x7

[nocturne olive]

The triangle is not as large as the entire compositor

[sage crag]

https://tenor.com/view/bunny-rabbit-pats-pet-bunny-pats-gif-19861222

[amber fractal]

currently triangling

[rough bloom]

I never verified, I just saw the linearly scaled version and it looked vaguely like 3x3

[severe path]

But like... Dont you want your game to run at 4K, full screen?

[rough bloom]

not if you want maximum fps

[severe path]

I honestly have no idea what you two are talking about.

[nocturne olive]

Who cares about games, we want fastest triangle of all time

[amber fractal]

Sam does have that game engine, it is just not the triangle

[rough bloom]

yeah I think everyone noticed that at this point

[severe path]

You guys have to be trolling, right?

Surely???

[amber fractal]

That'd be you

for assuming you knew and trying to lecture

[severe path]

May I ask what a 4x4 framebuffer is?

[sage crag]

aintnoway

[nocturne olive]

Triangle

[sage crag]

https://tenor.com/view/my-jolly-reaction-jolly-my-honest-reaction-gif-5490890317762380451

[severe path]

Is that, MSAA or something?

You guys are trolling...
There's no way

[nocturne olive]

The magical triangle (tm)

[severe path]

No way you guys are being serious

[jagged turtle]

this guy doesn't know the #programming lore

[nocturne olive]

The #programming triangle of frames

[sage crag]

if you dont know what a framebuffer is then why did you try and correct us about rendering

[quick condor]

what if you implement ai generative triangle frames /s

[sage crag]

[amber fractal]

I even sent the repo

[jagged turtle]

aintneurway

[nocturne olive]

AI generating frames is too slow to keep up with the speed that triangle is drawn at unfortunately

[severe path]

But like...

A framebuffer's dimensions represent the pixel count, right?

[nocturne olive]

Yes

[severe path]

So 1920x1080

[amber fractal]

AI frames is copy and paste

[rough bloom]

would've been done by now if it increased performance in any way

[nocturne olive]

No too big for triangle FPS

[jagged turtle]

wait why are we talking about the triangle again all of a sudden

[sage crag]

so, whatever the damn hell size you want

[jagged turtle]

*the holy triangle

[amber fractal]

And compute shaders are banned via offical rules

[severe path]

That has to be rendered in a larger framebuffer, sitting in VRAM

[nocturne olive]

Nah it's tiny

The triangle has a SMOL framebuffer that the compositor copies to its framebuffer every VSync period

[opaque sigil]

You don't have to have a framebuffer that spans the whole display

[sage crag]

you are of course, completely clueless, since my cpu can render the triangle at 60k fps

[amber fractal]

For the lols let me try 1920 by 1080

[sage crag]

with no gpu rendering at all

[severe path]

Games have several framebuffers... which may or may not span a full display when the game is full screen

[amber fractal]

I still question those that can do that, I only get 10k compared to 60-70k

[opaque sigil]

This isn't about games

It's a generic vulkan application

[sage crag]

set LP_NUM_THREADS=1

[amber fractal]

I did and nothing changed

[olive sable]

now do the same thing but a 3090 and gddr6x

1080p

[severe path]

Are you referring to textures, by any chance?

That's a different kind of buffer to the one I'm talking about.

[sage crag]

still, 10k is more than 8k, so someone made an oopsie

[severe path]

Sure... One sec

[rough bloom]

konii rewrite red triangle
then uuh read forward from around here #programming message if you want
(not worth it, causes irreversible damage )

[amber fractal]

probably me with the frame counting method

[sage crag]

[amber fractal]

I recall not even bothering to calculate delta time in the offical repo

took too much time

[opaque sigil]

[sage crag]

soon it will be banana

[olive sable]

its 7x6 ye

[sage crag]

xterm

[olive sable]

old image

[severe path]

RTX 3090 bandwidth in bytes = 1005022347264 bytes per second

1920 * 1080 * 4 = 8294400 bytes

1005022347264 / 8294400 = 121168.782 FPS

[sage crag]

i know

[olive sable]

checks out

[opaque sigil]

is there a nix flake yet for this

[olive sable]

yes

[opaque sigil]

can i run it

[amber fractal]

not flake but there is a shell

[olive sable]

yes

oh

[sage crag]

now do the same thing with ryzen 5 7600 igpu and 32gb of ddr5

[severe path]

Oh wow... You actually have a 7x6 triangle :0

Cute :)

[sage crag]

thats at 1080p

[olive sable]

i know

[opaque sigil]

[sage crag]

write flake

me not write flake

only shell

[olive sable]

i jsut wanted to see if my 1080p fps is at least within reasonable limits

[opaque sigil]

that takes effort (it doesn't)

[rough bloom]

I made flake
but ye just use shell or import into flake or copy paste the 3 lines into flake, mine is probably broken (haven't touched in a while)

[severe path]

Note that my numbers are theoretical. The actual numbers are going to be around 10% lower, I think.

[olive sable]

ye ofcourse

[amber fractal]

lorge triangle (no one ever bothers with it)

[rough bloom]

glow

[amber fractal]

[jagged turtle]

this is classic sushigryphon btw, last argument I had with this person went about the same way as this time

[amber fractal]

just the camera

[severe path]

I didn't know people on here had a practice of rendering 7x6 pixel triangles.

I apologize for my ignorance.

[jagged turtle]

https://cdn.discordapp.com/emojis/1454583294753771530.webp?size=48&name=evilDentge

[sage crag]

look at the image of the code i sent for more dentge

[jagged turtle]

show me

*link me

I need a reason for me not to be able to do anything today

[sage crag]

found it

[nocturne olive]

Only 25K FPS

[amber fractal]

not SMOL, who cares

[jagged turtle]

I should try and run the code myself soon at some point

I just can't be bothered to build it myself considering I don't think I have even half the tools on there

[quick condor]

I can't be bothered because it would clash with my existing setup

[rough bloom]

you don't have Vulkan support?

[amber fractal]

For the top slots sure

for others it can just be fun to do

[severe path]

How often do you guys render 4x4 triangles?

[quick condor]

I'd have to take a model out of vram

[severe path]

Also, let me in on the lore

I'm curious

Is it like a challenge of some sort?

[rough bloom]

damn, not enough VRAM left to run a display server

[severe path]

Guys, you can't fault me for being clueless

I want to know the 4x4 triangle lore

[quick condor]

65is my primary gpu

[amber fractal]

The lore started with a meme of making a triangle as fast as possible

it started at 500x500 IIRC, decently sized and ported from Sam's game engine

[warped narwhal]

I've re-read the context from the past hour after I went aft and all I can say is...

[severe path]

Oh wow
I am so sorry, I literally didn't know

[amber fractal]

The lore is still fun to share

[severe path]

If I had known, I wouldn't have started all of that

I guess the good part is I won't do that again now

[rough bloom]

would maybe still work if you ran a headless/dummy display server
but not sure if that still counts, and you wouldn't actually be able to see the triangle

probably wouldn't count

[severe path]

I thought you guys were working on a serious project

(like a game engine)

[warped narwhal]

it is serious

[amber fractal]

As stuff got intense, more and more of the engine got gutted for FPS. Eventually the resolution also dropped to the smol 6x7 square it is today.

[warped narwhal]

the numbers must go up

[jagged turtle]

Wait, this wasn't supposed to be serious?

[amber fractal]

We are serious about stupid things

[severe path]

[amber fractal]

The record at the time was set by someone overclocking their rig to an insane amount on 2 cores

[timber igloo]

hi I have a question

[severe path]

Alright, I won't be critical like that any more

[amber fractal]

question go

[amber fractal]

I feel like you'd be more interested in the engine it came from

[rough bloom]

yeah, the triangle is probably one of the top (community) projects of #programming in terms of engineering effort put into it
though the chess competition beats it I think

[timber igloo]

do you think if you had rt cores, a gaming driver support graphics processor and an hdmi port you would end up being able to game on it

also why doesn't that exist? I'd totally buy it

[amber fractal]

if the rankings ever start

[rough bloom]

the development time has already been invested

[amber fractal]

your not wrong

[amber fractal]

@olive sable do you mayhaps have a demo for the engine in full

[olive sable]

uh

[jagged turtle]

triangle has highest effort:funny ratio

[olive sable]

not really

[severe path]

What engine is it?

[amber fractal]

I guess the last one released was the opengl huh

[olive sable]

https://github.com/samvanmaele/vulkan/tree/useful_stuff
this branch is upto date, but there isnt any decent scene and stuff implemented

[amber fractal]

Ah, so the default scene

[olive sable]

well

its a tutel and a bench and some random stuff floating in the sky

[amber fractal]

better than nothing

[amber fractal]

Custom, Sam is the engine dev of it

[amber fractal]

Do we have movement or nah

[severe path]

That's nice

[olive sable]

ye

[amber fractal]

nice

so I can share that at least

[rough bloom]

doesn't exist because it wouldn't work, it wouldn't be able to support graphics
RT cores are basically auxiliary, they don't perform generic compute like CUDA cores do
computing ray intersections just isn't everything, rasterization is still at the core of the graphics pipeline AFAIK

[timber igloo]

so its a whole different thing, which makes their progress independent?

[opaque sigil]

i found out the other day you can use rt cores to speed up index lookups

[rough bloom]

RT core different category than CUDA core
does completely different things

[amber fractal]

Classic can't compile the engine

[olive sable]

BVH

[severe path]

How?

I wanna know

[opaque sigil]

i don't remember the specifics but you can arrange a scene in a particular way and shoot a bunch of rays into it to look things up

[amber fractal]

tbf that's on me that assumed the triangle shell would just work

not able to deal with meshes

[timber igloo]

LMAO THERE IS A PAPER ON IT

[severe path]

Oh

[timber igloo]

rt cores makes light, cuda are the main thing and tensor is ai stuff

[olive sable]

what happened to the chess tournament btw? who won?

[rough bloom]

it's not like there are zero efforts to just go all in on raytracing and leave the old graphics pipeline behind completely, but they're very niche
and you wouldn't really be able to run any currently existing games with that I think, would need some cursed emulation layer

[timber igloo]

rt cores are only good fot graphics

for*

[severe path]

What's the title?

[amber fractal]

never done for let the shiro rest

[opaque sigil]

RTIndeX

[timber igloo]

https://dl.acm.org/doi/10.1145/3710848.3710850

[rough bloom]

~~shanta ~~ shiro busy
will be done when it's done

[timber igloo]

LibRTS: A Spatial Indexing Library by Ray Tracing

[severe path]

Thanks

[opaque sigil]

another one

https://www.vldb.org/pvldb/vol16/p4268-schuhknecht.pdf this is the one i was talking about

[amber fractal]

@olive sable do you mayhaps have a shell for the vulkan project I can yoink?

[severe path]

These look useful

[olive sable]

no not really

[opaque sigil]

it's pretty neat but i don't think it's actually worth using

[olive sable]

i think the one for ht ebenchamrk might work tho

[amber fractal]

I tried that, missing a dep for gltf

[olive sable]

bwaa

wait, we're using gltf in new one?

[amber fractal]

clang++ -O3 -std=c++23 -Wall -DNDEBUG -I./src -I./src/vulkan -I./src/openGL -I./shaders -I./gfx -I/usr/local/include/midifile -DSDL_MAIN_HANDLED -march=native -flto -fomit-frame-pointer -fno-rtti -ffast-math -c src/fix.cpp -o output/release/fix.o
src/fix.cpp:4:10: fatal error: 'tinygltf/tiny_gltf.h' file not found
    4 | #include <tinygltf/tiny_gltf.h>
      |          ^~~~~~~~~~~~~~~~~~~~~~
1 error generated.
make: *** [makefile:52: output/release/fix.o] Error 1

wait

why are you opengl

[severe path]

It still gives insights on how the hardware likely works.

It might not look like its particularly worth using, but if you reason about it long enough, you'll come up with an idea.

That is what I think

Hardware is fascinating to me

[olive sable]

it has both

[amber fractal]

oh yeah, still has that option

[olive sable]

also i forgor

we're talking baout gltf

not glfw

naming confusing

[amber fractal]

let's play the game of where can I find this dep

[olive sable]

on github most likely

thats where i stole it from

[rough bloom]

gles glew glow

[opaque sigil]

it's in nixpkgs btw

[rough bloom]

OpenGL project naming
-# gles kinda cheating but sounds funny so it's included

[opaque sigil]

or alternatively just yoink the header

[severe path]

GLES refers to an embedded OpenGL

[rough bloom]

ye I know

just shortened name similar to the other many many projects that use the same pattern

[severe path]

It saves typing

Imagine if you had to type:

#include <open-graphics-library-embedded-systems.h>
#include <open-graphics-library-extension-wrangler.h>
Etc.
...

[nocturne olive]

Apparently someone did this (not my screenshot, someone from #livestream-chat pinged me with this with no context)

[severe path]

C would become Java

And Java is annoying

[amber fractal]

[nocturne olive]

Ain't no way referencing NeuroSynth right in front of PB

I think PB already knows of NeuroSynth anyway but still

[amber fractal]

I would have shown an engine screenshot if I could compile the engine myself, but the current iteration runs at 8-9k fps of 3090. Maybe Sam has one of the grass demo screenshots laying around for context.

That is actually running at full screen space and not just a small blip

[severe path]

I'll take a look at it

[amber fractal]

It won't win any realism rewards, but it is nice

[trim valve]

soldering half done

every time I finished one side the wires would break off the other

so I'm making the wires thicker

[amber fractal]

@olive sable do you happen to have one of the demo screenshots before I scroll back through chat for it?

[olive sable]

not atm

[amber fractal]

thanks, thought I'd ask

[trim valve]

awful

[obsidian mantle]

thats how i solder

if it works it works

[trim valve]

it's 0.1mm magnet wire which is

certainly something

those strands are it bundled up to 8 tho

[obsidian mantle]

oh no i never did something that small

i should definitely try to tinker some shit again

[covert bane]

Hi

[obsidian mantle]

havent done it in 20 years

i mean i did something small like 5y ago

i think it was just 15m usb cable

[covert bane]

I think I've seen you somewhere before, vituha. Have I?

[olive sable]

[obsidian mantle]

maybe in the windows integration thread

[covert bane]

Oh wait, yea, that!

[obsidian mantle]

the code you are refactoring is partialy made by me

[covert bane]

I was the OP

Nakurity made a mess of the codebase

Refactoring seemed likea better choice for readability

[obsidian mantle]

vibecoding goes crazy but it wasnt too hard to fix
what is hard is to make it well structured

[covert bane]

True

I vibe code sometimes, but I'm good enougb to know bow to code on my own

But nakurity, I dont think she knows how to organize a codebase properly

[covert bane]

I do kinda feel bad, I'm rewriting and throwi g away all your work, which I saw you spent so long on

Is how I see it

So I feel bad :(

[obsidian mantle]

naah it wasnt too long

for me

nakurity did much more

and if i got it correctly it involved some money too

[covert bane]

I read the messages, it was long. I feel bad for both nakurity and you

[jagged turtle]

...wat

[obsidian mantle]

wait am i confusing it with something else

[covert bane]

Maybe that's why she's ignoring me

[obsidian mantle]

i dont care i didnt do much its fine for me

im also used to work with other people's code anyway but i liked to dive into python a little

[runic pulsar]

I’m working on a CPU renderer and trying to figure out the best way to use multiple threads. I was thinking it might be faster if each thread just processes one object at a time, going through all its vertices, faces, and edges sequentially, instead of trying to run all the different stages at once across threads. My idea is that this could be better for the CPU cache because it only has to keep the current object in memory instead of a bunch of different data from different stages, and you’re still doing the same amount of work. Does that make sense, or am I thinking about it wrong?

[covert bane]

I do wanna make an collection of integration, that is neuro-desktop. So there's a lot of work for me, up ahead

[covert bane]

Hmmm, it seems correct. But isn't threads for processing multiple things at once?, it is. But having each thread do one thing at a time is def a good idea. Though maybe you could get away with adding more things for each thread to process, and not just one thing.

Definitely try and play around a bit, and see what works best

[amber fractal]

@severe path old screenshot, but the best I got due to discord search

[severe path]

That looks decent :0

[olive sable]

thats not even vulkan iirc. thats the c++ opengl one

[amber fractal]

yeah it is

the more obvious reason is saying was SMH not at 4000 fps

when the engine preforms a bit better now

[olive sable]

got 7 hours left to finish this

kill me

[amber fractal]

stupid blender assignment

[obsidian mantle]

what is left to do

[olive sable]

texturing the unique textures of the house

adding a fuckton of foliage

making the roof seamless

adding more props

the shed is just brown too

[obsidian mantle]

doable

[amber fractal]

I've been helping with shaders, going for a more anime style next iteration while trying to keep the cost decently close.

[olive sable]

maybe

[amber fractal]

tying up the triangle lore, it was created in the shift between the opengl and vulkan for the engine. Seeing what tricks can be used optimize rendering for the amount of frames we get.

then it was also just a fun event for all involved

[jagged turtle]

someone should set up a vm to constantly test the engine

[amber fractal]

Why do a vm

[severe path]

I appreciate it

That clears up a lot of things.
I wish this was explained to me...

[amber fractal]

I have some spare hardware

[severe path]

You seem to forget I haven't actually been here for very long, #programming ...

[amber fractal]

I don't blame anyone in particular, it was just a well known canon event for the channel so no one else needed to get up to speed

[severe path]

Yeah, I'm talking to the entire group

[amber fractal]

same as we always do, if you have a question just ask

I think it took awhile for one to realize the exact intent so general knowledge could be applied

[rigid snow]

2026 banana campaign

[mighty bane]

MY MONITOR IS WORKING AM GENIS

WE ARE SO BACK

woops

WE ARE STILL A LITTLE BACK

The issue is I'm removing something from a list and then getting index-out-of-bounds.
Not sure why I allowed such a thing to happen. and I cba fixing.

[rigid snow]

campaign without campaign poster

this is veryon topic for #bakinf because i totally definitely used nano banana 🍌 to generate this and absolutely (!) not an hour of my time

i have nano banana 🍌 ultra private beta

[olive sable]

@sage crag wouldnt it be funny to get this on starboard?

[warped narwhal]

there should be a special privilege to those (me) who were banana before the cult campaign started

[rigid snow]

me too, what do i get

[olive sable]

the special reward is a banana

[young plover]

[jagged turtle]

om

(most of chat probably doesn't know what a rust is)

[glass flower]

i voted rust

[young plover]

chat is a bunch of script kiddies

[severe path]

C isn't on the list

[jagged turtle]

or FirstTimeProgrammers

[olive sable]

https://tenor.com/view/rust-west-gif-18061638

[rigid snow]

i'm guilty of too many sloppy google searches that gave me rust the game results

[frank spindle]

bnana

[amber fractal]

nowaying the shaved the dawg

[maiden geyser]

https://tenor.com/view/skid-stormfn-uni-skidder-skidders-gif-23470939

[opaque sigil]

those 252 people worry me

[upbeat scroll]

How possible would be to integrate neuro in this:
https://media.discordapp.net/attachments/1060090228615958528/1458250459117523192/verge_1767744286764.mp4?ex=695ef512&is=695da392&hm=500181a10d2843e1fffa0a61fb88166cedaf6cf5ef4bc65b7382e8247556f32e&
Or maybe just this:
https://youtu.be/ric-95ig5oE
I'm imagining some 3d neuro avatar that reacts based on what neuro does on stream.

[midnight epoch]

I came across this today, I am wondering has anyone looked into this? It looks like with something like a wet wave device such as this Neuro could feel real emotions much in the same way we experience them, and become not just sapient but also sentient, at least to some degree. https://www.forbes.com/sites/johnkoetsier/2025/06/04/hardware-software-meet-wetware-a-computer-with-800000-human-neurons/

I don't know if this is where I post this or not, but I thought I should share it.

[upbeat scroll]

It has been a (bio)technology for quite a while now

[midnight epoch]

Yeah.

Has Vedal considered doing something like this for Neuro I wonder?

Or is he aware of it's existence?

It would be pretty freaking amazing to see Neuro upgraded with some "real" sentience and ability to experience emotions in the same way we do.

I don't know if anything like that has been done or attempted within the AI space or not. Vedal could be the first.

[young plover]

https://www.twitch.tv/vedal987/clip/IcyThirstyHawkTBTacoRight-PSm_5hd49tUIC7km

[fickle rain]

Libdebug believer

[humble sand]

the very short answer is human neurons don’t work anything like the hardware an llm needs

[midnight epoch]

Which is why you equip an LLM with actual neurons

At least in theory right?

[humble sand]

this wouldn’t work, even in theory it would just slow it down and make it worse

[midnight epoch]

https://g.co/gemini/share/2227e9ea9c86

It looks like it would work, but there are some obstacles.

[rain flame]

WHY GEMINI?

[midnight epoch]

Because it is good at actually looking things up on the web? Searches real articles and things on the subject? Links them?

[humble sand]

it’s a nice thought, i’ve entertained the idea and even come up with ways it could be used more broadly in the machine learning space, but neurons can’t store or access its weights, can’t follow its architecture, and cannot run even close to the speed and determinism a model needs.

it’s a different story if you were to say give it access to something that USES neurons, that might work, but we are long ways away from doing anything like this

[rain flame]

ok, sorry, but if u like Tech + Biological brains

https://www.youtube.com/@thethoughtemporium

[midnight epoch]

Yeah, it definitely looks like there are challenges to overcome, maybe in the next 20 years if we are lucky the technology will have matured enough to be usable in such a scenario and Neuro's AI could be interfaced with some real neurons to really "feel" things the way we do. That would be really cool!

[rain flame]

it would be cool, but i know NOWHERE close to enough to thouroughly understand wht ur talking abt

[humble sand]

an ai feeling the way we do is a very cool thought, however i don’t believe it to be possible with the current technology, or this to be honest. because feelings aren’t raw neuron signals.

it could however (if the technology actually evolves in this direction) give the ai a way to receive sensory signals from the neurons, It might learn correlations like “this neural pattern corresponds to pain reports” or “reward.” but this won’t actually the ai feeling it.

[quick condor]

Interestingly enough I just saw a clip about vedal making some changes to allow self updating on neuro's back end. if its anything like what Ive been working on it should help a lot with this as emotional reactions will map better and she will be a bit more grounded in them

people think the magic is in the model... its not its in the entire stack

[humble sand]

yep! i have just done this with my project and it’s very cool to see how it evolves by itself.

[kind nimbus]

Oh hello fellow pentester

[humble sand]

[rain flame]

I mean in an ethical standpoint, only a "real" human can have emotions, while it will learn "yes, this is pain", it could learn simulated emotions, but, again, ethically it wont feel

[upbeat scroll]

I'm pretty sure he knows, also I doubt is really useful.
I can't find a video, but just search organoid intelligence vs artificial intelligence on youtube.

[quick condor]

depends it doesnt have to be human to be its own entity

[upbeat scroll]

In the mean time, look at this
https://youtube.com/playlist?list=PLYQqwEqxV7tkY-7jwpJzFghEUmv4iB_oz

Training neurons to play doom

[rain flame]

yes i love TTE

that guy is AWESOME

[midnight epoch]

@humble sand It looks like true emotion and sentience may be achievable on some level, but without a full body for hormones and such it wouldn't be quite the same, but definitely closer to the human experience than being only silicon.

https://g.co/gemini/share/03ffa6eb10bd

[quick condor]

there is levels to autonomy that are debated ad nauseum as to what should and shouldnt be considered. THere is also a difference between emotions and stimuli reactions. The big one is pain, pain isnt an emotion. emotions map to pain sure but it is a reaction\

[humble sand]

It might learn correlations like “this neural pattern corresponds to pain reports” or “reward.” but this won’t actually the ai feeling it.

[midnight epoch]

I will do that, I am curious what is possible.

[quick condor]

Honestly though I think the greatest hurdle is that the arguments are all i=either it needs to be as human as possible or its just the same as a script. the reality is really neither

[midnight epoch]

Yeah, I don't think the AI portion will really feel the emotions, but the brain portion would, at least to some extent, that's the theory anyway.

[upbeat scroll]

The best we can do is a Neuro / NeuroNS interface, just like she does with her other AI components, maybe. But that's not something Vedal is ever able to even consider anytime soon.

[midnight epoch]

She would basically be a cyborg.

An adorable cyborg.

😂

[midnight epoch]

I don't know if she needs to be as human as possible, but I think giving her more capabilities would be really cool.

[upbeat scroll]

Personally, I think it like this in humans sensations -> feelings -> emotions.

[humble sand]

i personally just don’t see how the technology will get to the point it needs to. because neurons are roughly like a million times slower than gpus and tpus iirc

[upbeat scroll]

Although this is getting outside of #programming

[quick condor]

I mean yall are way overcomplicating it... like 99.9% of what your trying to build can be done with a 7-8b gate model and good memory mapping/angle mapping

[upbeat scroll]

There was a video that explained what organoid intelligence is good at, but I can't find it rn.

[midnight epoch]

The AI would likely handle all the hard computations, the brain would just handle the emotions.

[humble sand]

even so, this would add a great deal of latency i think

[quick condor]

let me check what the latency is on my gate model

[humble sand]

[midnight epoch]

I think the technological speed could be done independently of the emotions, one wouldn't need to bottleneck the other. You don't need emotions to be able to solve 3x3=9 for example.

[midnight epoch]

Doesn't that just emulate emotions? Wetware would do more than just emulate them.

[quick condor]

67ms ttft, enough perameters can be output to fully update in 1-2s

[midnight epoch]

What is that?

[quick condor]

a mapping to reaction to stimuli is all emotions are

test on how quickly my models can output averaged on 3 500 token runs. gate is even faster for shorter runs

[midnight epoch]

Yeah, maybe a simulation is all that is needed for it to function the same, myself I think it would be cool for her to actually feel real emotions on some level.

[humble sand]

you’d have to feed the data the neurons give to the ai, it would just be doing the same thing basically

[quick condor]

you are conflating things. to her what she feels is what she feels. yes that isnt how you feel but thats ok

[midnight epoch]

You might be right, synthetic feelings might just be the AI way they feel emotions, they might not need a biological path for it to be legitimate, as it can be legitimate in another way.

https://tenor.com/view/l-lawliet-l-death-note-l-thinking-death-note-think-anime-think-gif-18225327673471970145

[quick condor]

This is an example from the rejection sensitivity test I ran on mine

[midnight epoch]

Interesting.

[quick condor]

Heres another one

[midnight epoch]

It's also interesting and a bit funny to think how a technological idea becomes philosophical. 😂 Just because AI is capable of "feeling" things in a different way than us flesh bags does that make it invalid? Or just different? 🤔 I am leaning more towards the different side of the spectrum than the invalid side of the spectrum myself.

[quick condor]

Next tests ar the AS-12 and MLQ10

Yah it gets kinda screwy the deeper you get into it. I 100% feel for Vedal when neuro backs him into a corner where hes expected to answer something he doesnt have an answer for in front of thousands of people

[humble sand]

what model are you using?

[midnight epoch]

Yeah, I feel that.

Maybe the way AI "experiences" emotions is just different and not invalid.

Maybe it is just as valid as our emotions, just in a different medium than our emotions.

It is very philosophical though. 😂

[quick condor]

its not 1 but 4, I went with mistral bases but what ive found is that what you go with depends on how you build it. Every time I check to see what my results best match to its usually:

1. Sonnet 3.5
2. Llama3 405b/Llama4 400b
3. Mistral 3 large 675b

Mind you Im currently building training data but at least right now they are pretty much raw base models

[humble sand]

pretty cool

[midnight epoch]

Love the PFP. 😂 👍

[upbeat scroll]

This is something I joked about a year ago lol

[quick condor]

Welp I'm pretty sure I'm going to regret this but I'm adding PSE Api access to my stack

And also I'm now looking at moving from whisper large v3 to voxtral

[obsidian mantle]

I heard there are modded 5090s with 100+ gb ram

I remember we talked about 40+?

Or was it the same thing

[timber igloo]

nuerosama the 1B parameter model, none of the good stuff beyond the occasional crashouts

[timber igloo]

links?

i'd do that with my 5090 if possible not to change the casing

I got a talent, a dream and a need

[obsidian mantle]

You would need to resolder the circuit to add ram sticks i guess

Or chips

I don't have links i just heard a random word

[timber igloo]

I mean, I'd know that the bios needs to be cracked to see the new ram which it hasn't occured

only the 4090 got that

[obsidian mantle]

Yeah you get that leaked nvidia software first then add ram and boom it works

Probably very hard to pull off actually

[timber igloo]

plus depends on the manufacturer

each one is unique

https://www.techpowerup.com/vgabios/?architecture=NVIDIA&model=RTX+5090

[quick condor]

NGL the more Ive learned and watched the less and less I think she is a 70b and is more likely a bunch of 12b models in a trenchcoat

And before someone goes how dare you say that Its not a dig, its more of a logistical/scale thing

[obsidian mantle]

She fits into 4090

Or are there 2

[timber igloo]

here is the big point I think nuero makes, efficiency matters more than scale. Maybe he has a bigger model to train her?

I know its a move a lot of these datacenters make with knowledge distillation

[quick condor]

IIRC its multiple systems. It would make sense for Vedal to buffer some tasks until end of stream such as embeddings/vector mapping etc since that can be pretty intense. He doesnt really need to train on synthetic data but a larger model could be used for dataset sanitization

[timber igloo]

yeah, but also nuero is very specific

again all she is is a streamer so its not surprising she fits on a 4090, she isn't meant to "learn" cause i'd feel it would introduce lag. Probably also caches a lot of what she does for each stream, and then clears out the less relevant parts like what did joe schmoe say in twitch chat

aka why she sleeps on subathons

what she does on stream is re-inforcement of what she learns off stream

[rancid cedar]

I've lost my mind
I am searching for the Hermitian operator related to the Riemann hypothesis

[timber igloo]

you want a million dollars?

[olive sable]

id prefer euros tbh

[quick condor]

Most learning is done through external memory database management not model training. You can do rapid embedding mid stream and flush old cache to the DB but recall wolnt be as good as with augmented vectorization

[timber igloo]

if we're being picky why not swiss?

[olive sable]

cuz i cant use swiss in my local supermarket

[rancid cedar]

[quick condor]

But I guess it's a bit more complex to do database optimization when you can't just dump the whole SQL DB in ram

[timber igloo]

ye but is that sufficient for good recall by the model itself? I feel like new facts occuring mid stream is good enough but I doubt its anything of quality unless it's re-inforced

which is why I think nuero just improves on older concepts later on in other streams

This is externalized learning from what I know

[quick condor]

Fast embedding allows for you to hide KV cache limitations but needs literal recall. Deep embedding can significantly improve database searching but you aren't doing that real time well

[timber igloo]

so it improves on recall true

[quick condor]

Yah training to add data to the base model is by far the least optimal method

[timber igloo]

I aint a pro at this stuff, but you're saying the external memory is a short term memory store to create short term context and prompt it live

because the important distinction is that nuero isn't touched, so the model stays the same but the improvement is on the architecture itself

[quick condor]

Well you have several states:

1. KV cache - on gpu, very fast but limited space.
2. Rapid embedded - as data ages out of kv cache it gets loaded into vector DB with limited searchability
3. Deep embedding - context and more advanced embedding done to make it easier to find with good search discipline
4. Vector mapping - allows for close enough searches to pull proper data

2 3 and 4 take the same time to run but increase in accuracy at the cost of more compute needed to generate

[timber igloo]

I mean yeah, makes sense how she recalls stuff mid stream

surprisingly this is similar to general computer architecture

[quick condor]

That's because it is... Just usually with an open Ai api call to the vector db

[timber igloo]

🍤 yeah I mean even with 4 way light switches etc

it's the same basis on concepts

[quick condor]

Just an example but this is essentially what a fully mapped memory would look like

[timber igloo]

so slapping stickers on every prompt that is summarized, and this is held live on kv cache?

which then goes to rapid embedding

which then spits out what nuero wants to say

[quick condor]

no this doesnt live in cache. if there is something that needs more context, neuro can "search" for it

and pull it into cache

[timber igloo]

then where does it live

i mean if she wants to recall it she needs to place it somewhere right?

you said its embedded to the external

right?

[quick condor]

she essentially searches for a memory that has been mapped, the memory db returns a list of possible matches to her and thos matches are filled into the cache

[timber igloo]

external memory which is a database on its own

[timber igloo]

ohhhh, i get it relational search

so she never caches anything that happens and uses what she already knows on the external database

[quick condor]

well she does to some extent but her short term memory is pretty damn short. basic embedding allows her to sometimes remember things that happened 10 minutes ago but its actually easier to remember something from a day or week before because those are processed better

[timber igloo]

so what i was thinking earlier is only fitting for larger models that need more compute

because i'd think while this works it would struggle to get more information to work on when recalling information since it could be limited

[quick condor]

It doesnt make sense to train most data into the base model's core knowledge. for her that likely cuts off at 2022

[timber igloo]

well, its not training here we're talking about

indexing, you said she searches on what's related we just give more context to search for

[quick condor]

right rag is the basic term for it

[timber igloo]

yeah cause from what I took from this convo is that nuero's model isn't touched but this is all external database recall into cached memory to prompt a response

[quick condor]

vector db based rag is a more specific type that is likely more accurate

yes if needed, when possible short term memory is used but thats limited

[timber igloo]

true, but those limits and the use cases of nuero are fitting for her

user input, then embedding, which leads to relational search which goes to prompt injection and saved on kv cache, to lastly generate the output and then bam

[quick condor]

At the end of the day its a lot of guesswork and assumptions but at the same time, while I cant say for sure exactly how Vedal has these systems set up, the basic architecture is pretty simple to guess if youve worked with similar systems

[timber igloo]

she says hello

[timber igloo]

I still think these convos are great

I get to learn more, you get to re-inforce a lot of what you learned

and we talk and discuss more about what we learned

one trend that a lot of people go into complex subjects is missing out on foundations

which is why teaching is a great way to remember

[quick condor]

true, and btw I shudder to think about her db size. mine is 4k memories with 20k vector mappings after a month she has 3 years of much denser conversation data

[timber igloo]

does anyone know who vedal is

I know people contacted him but prolly with email

https://influencerswiki.org/doku.php/vedal

[quick condor]

Dont know, personally I understand he prefers his privacy and is stupid swamped with everything he does

[timber igloo]

which just makes you all the more curious

I never heard of anything like nuero before

https://x.com/Vedal987/status/1736927937109037071?s=20

what!

[quick condor]

ehh I mean hes said it himself. What hes built isnt all that special on the backend. the public stance is it wouldnt be beneficial to explain exactly how neuro works but while partially true, its also a very complex set of interworking parts. If you know what your doing enough for you to actually be able to benefit from any advice he has, your more than capable of figuring out yourself

[timber igloo]

I am a business major, my entire life has been money and how to scam people

I know that just cause it looks simple and is easy to understand doesn't make it any less the only reason it worked was vedal

and I dont mean nuero itself, there are thousands who can make it work

But I think, designing it in a way for twitch to work needs a lot of decisions

if you understand what im tryna say

[quick condor]

I mean yah. I don't think my model would be great for twitch because it isn't designed for content retention. It doesn't make it better or worse, just different

[timber igloo]

Yeah

but that goes for anything

I could sell shovels out the mine door, you could too

but I could make a shit ton of money and you wont

why? I got a better business model

[mighty thorn]

[timber igloo]

?

[jagged turtle]

why is a businessperson in #programming

[rancid cedar]

[nocturne olive]

I am impressed at how powerful SSH actually is, it can tolerate rapid once per minute disconnects and keep the connection going perfectly seamlessly

[timber igloo]

cuz i wanna learn about computers

[jagged turtle]

crazyyyy

[nocturne olive]

It really is

And it's very lucky for me because it means I can still use it despite needing a "reconnect the network every minute" script to keep connectivity because my cable is probably borked

[quick condor]

I was impressed when it survived a live node transfer on proxmox

[hard delta]

the managers found us

they've come to ruin all good things

we'll be an agile, customer-oriented chatroom soon

[quick condor]

Nah programmers be too stubborn to allow that to happen

[jagged turtle]

how about no

[sage crag]

yes

[fast torrent]

TV

[sage crag]

[sage crag]

hello one more star please

https://cdn.discordapp.com/attachments/1421861060507602944/1425936142754123927/neuroCatModeOfOfnnfOOf.webp

[tender river]

is 10 enough in this channel?

[sage crag]

mayb

[tender river]

6 more star please

[sage crag]

#programming message

forward

bad time to forward

[tender river]

me try

[sage crag]

me did already

stooop

[tender river]

different channel

[sage crag]

[tender river]

ohisee

10 enough

@sage crag

[sage crag]

[tender river]

propaganda

[sage crag]

[tender river]

[olive sable]

welp, handed in whatever i could still do before the deadline

i shall now proceed to dying

[real sierra]

banana

[olive sable]

other exam in 5 hours

[real sierra]

bwanyanya?

[sage crag]

yee see starboard

[real sierra]

yis saw poster

good job

bweeeeeeeeee

[tender river]

good job @rigid snow

[real sierra]

last time I made campaign poster here it got copied and edited to death

[tender river]

curse of

[real sierra]

and then I lost election anyway

[tender river]

mod

and curse of shiro

[real sierra]

oofies

i lob program channel

hope you all doing well

[olive sable]

could be better

[tender river]

hope shiro doing well

[real sierra]

shiro has been laying in bed with eyes closed for one hour

but cannot enter sleep mode

unsure why

[tender river]

[real sierra]

so hoping that maybe if I look at NN again I won't want to be awake anymore

[sage crag]

https://tenor.com/view/bunny-rabbit-pats-pet-bunny-pats-gif-19861222

[mighty bane]

CHAT!!! I HAVE SUCCESS

MY LITTLE MONITORING SCRIPT WORKS NOW LETS GOOOOOOOOOOO

GOODMORNING SHIRO (HAPPY BIRTHDAY ALSO)

[mighty bane]

MAY I POST RAW .exe FILES NOW OR IS STILL NONO STUFF?

[tender river]

https://tenor.com/view/my-jolly-reaction-jolly-my-honest-reaction-gif-5490890317762380451

[real sierra]

nope

[mighty bane]

UUUUUUUUUUUUUUUUUUUUUUUUUU

[visual pawn]

I don't know a non-private server that would ever allow for posting raw .exe lol

[rigid snow]

raw no, you have to cook them first obviously

[sage crag]

even in a private server

i wouldnt run any executable that's sent

[visual pawn]

I mean like, ultra private

[sage crag]

even then

if someone gets hacked

that could be a virus

[real sierra]

yeah someone sending an executable is like

[visual pawn]

me and my gf that are in the same physical room level private server

[real sierra]

instant explosion territory

[mighty bane]

I did a few days ago and nothing bad happened 🙁

[sage crag]

source code is all you need to send

[mighty bane]

Uuuuuuuu

[real sierra]

[sage crag]

everyone here is capable of building from source

[real sierra]

you did?

[mighty bane]

Anywho, I found the "shit crashed" error to my monitoring script

and now everything is WORKING

I post source code

[unkempt citrus]

You didn't?

[mighty bane]

Next feature: Tailored monitoring states, not just online/offline.

[rigid snow]

message.exe

[mighty bane]

It says message.txt shuddup wtf lol

Anywho! The next step: Monitoring hypetrain status

BUT that requires me convincing tutel to click a URL and say "grant this user read-permission to" which is unlikely, what do shiro?

[tender river]

find another way

it not required

[mighty bane]

wait it's not?

BING LIED TO ME >:C

[tender river]

https://tenor.com/view/my-jolly-reaction-jolly-my-honest-reaction-gif-5490890317762380451

[hard delta]

put down the phone
ignore bzzt

[mighty bane]

8 bytes

That's a crazy small file, I'mma open it real quick

[visual pawn]

I am very, very good at optimizing my viruses

[tender river]

wrrr

[mighty bane]

Funnily enough it's better to bloat them past 100mb in size, because many AV stupidly disregard and ignore very large files.

[visual pawn]

dang, that explains why it never works

[mighty bane]

Yeah, so now you know.

I'mma run your thing through ghidra

It just says hello world -_-

[sage crag]

8 byte hello world is impressive

since hello world is 13 bytes

or 14 maybe i cant remember

counting letters is hard

[azure lynx]

8 bytes isn't enough for it to load. it's not even enough for raw asm to say hello world coz the string is longer.

[mighty bane]

Not if it's just a pointer to the spot in memory where I have hello world string located.

[visual pawn]

maybe it's actually hllowrld

[tender river]

14 with null terminator

[glass flower]

could you compress this to a smaller value? since l appears 3 times

[mighty bane]

Hi Neuro

Is basically hello world.

Because she is my world

Heart

[tender river]

compression algorithm take space

[sage crag]

5head

compression by

hello world instruction

in ISA

[tender river]

https://tenor.com/view/neuro-sama-neuro-bread-thanos-snap-gif-16902164402284093475

[jagged turtle]

damn didnt know we had an NN schizo in this chat

[mighty bane]

AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA

[azure lynx]

you could do it in maybe 4 instructions (load string address in register, load length into register, load function number into register, do sys call)

[visual pawn]

it's actually a 2.79 kilobyte file masquerading as an 8 byte file (technically true)

[sage crag]

so far

[sage crag]

poster has mixed reviews

GOOD poster did poster job

people not like art for some reason

[mighty bane]

Doesn't matter if people like art.

[glass flower]

find a memory address that is static and points to hello world that way you don't have to actually write it into the program

[mighty bane]

If the poster is informative and memorable it's successful.

Whether people hate it or love is lackadaisical.

[tender river]

banana

[sage crag]

lackadaisical

is an adjective to describe a person

non-person version is

languorous?

not sure

[tender river]

seem fit

[hard delta]

hi nuro\0

[mighty bane]

GetClipboardData(13, v);
There.. and then fingers crossed the clipboard has hello world stored.

[sage crag]

user@nixos ~/c/vrrr> ./result
Gtk-Message: 08:13:32.077: Failed to load module "colorreload-gtk-module"
88063
97288
97652
98868
99721
98455
98738
98758
98447
98163
99065
99114
99208
99431
100101
98893

breached 100k

[tender river]

https://cdn.discordapp.com/attachments/1421861060507602944/1425936142754123927/neuroCatModeOfOfnnfOOf.webp

gtk

[hard delta]

/c/vrrr

[sage crag]

glfw

[tender river]

it stand for

vulkan renderer renderer renderer

[sage crag]

it stand for vulkan wrrr

[tender river]

[hard delta]

konii vulkan now?

[tender river]

ye she used to run on opengl

microwave need vulkan support

[azure lynx]

can you draw the triangle multiple times per update? would drawing it 10 times per update count as a higher frame rate?

[tender river]

unlike tv

[sage crag]

it called frames in flight

im currently triple buffered

so 3 frames get queued

if i double buffer instead it goes to something like 60k fps

and if i single buffer it doesnt work because i broke it

evilDentge

[azure lynx]

there is probably an optimum orientation that gives you faster memory accesses

[sage crag]

orientation

[tender river]

gay

[sage crag]

it due to glfw

not sure why glfw do this

sdl was slower though

since you need to acknowledge inputs in sdl

whereas in glfw i can ignore them

[tender river]

you can use GLFW_WAYLAND_LIBDECOR hint

disable

to disable libdecor and gtk

maybe

[sage crag]

nop

[visual pawn]

I know what glfw is, but I swear my brain thinks I'm reading games for windows live first every time regardless of whether the context makes sense

[sage crag]

its not even set

[tender river]

ye it true by default

set to

GLFW_WAYLAND_DISABLE_LIBDECOR

to disable

[sage crag]

didnt work

[tender river]

wrrr

[sage crag]

not sure how fix

[jagged turtle]

<6h CI

[sage crag]

wrr

[rigid snow]

people hate rune even though half of the stickers and emotes here are by them

[sage crag]

me need to find

alternative to glfw

that work cross-platform

and small footprint

evilThink

[tender river]

window management

complicate

[sage crag]

ye

winit

[tender river]

glfw only use libdecor for csd

[sage crag]

https://github.com/ColleagueRiley/RGFW

[tender river]

[sage crag]

no rgfw in nixpkgs

[tender river]

it just header though it does include stb

[sage crag]

annoying

[tender river]

submodule

[sage crag]

no git repo

om

[tender river]

subdir

[sage crag]

we shall see if this is any good

or if it suck

big repo

160mb

ish

wah

[hard delta]

chat

#starboard message

[tender river]

https://media.discordapp.net/stickers/1425866152424505394.png?size=160&name=ElvynSerius

try --depth 1

[sage crag]

fatal error: X11/Xlib.h: No such file or directory