💽Programming Chat v2

[spare quartz]

blah blah blah you're just mad you can't use ipc as beautifully as i do

the os idea i did have one time and is written in a document somewhere details a more extreme microkernel

basically just permission rings but if we added more

give THIS specific process the ability to use io ports, but not that one

[timid quartz]

macrokernel on top you’re just mad you’re not fast and you still haven’t really solved any problems with fault tolerance

[spare quartz]

"fault tolerance" and his system crashes when a vendor makes one oopsie 😭😭

crowdstrike looking ahh 😭

[timid quartz]

your userspace is still left degraded or unusable

[spare quartz]

nope

[timid quartz]

might as well have just crashed

[spare quartz]

not only is your statement HIGHLY dependent on whats crashed its not really true

if a peripheral driver dies thats fine

if a display driver dies thats fine

for you? both would crash

we can just restart it

[timid quartz]

how you gonna restart it if your peripheral dies lol

can’t type “restart” with no keyboard!!

[spare quartz]

... thats

why other processes can see that they've crashed 😭

[timid quartz]

what if your process manager dies :3

[spare quartz]

why would it die

[timid quartz]

cause it could

[spare quartz]

okay

we restart it

[timid quartz]

how

[spare quartz]

by sending a message

[timid quartz]

You can’t schedule a process if there’s no process scheduler

[spare quartz]

you can

and if you're ever at a point where IPC dies, which is usually one of the only things IN a microkernel

then you're just what-a-bouting and obviously theres no recovery

[timid quartz]

I’m just hearing microkernel inferiority tbh

[spare quartz]

yeah ok

tell me when you got your stupid little monolithic kernel not dying from a sneeze

[timid quartz]

oh it doesn’t dw

[spare quartz]

👃

[timid quartz]

It’s actually quite speedy

[spare quartz]

this mf would put his scrollbar code in his kernel to make things faster 😭

[timid quartz]

You also don’t get pretty visuals like this

[timid quartz]

this mf would use gnu hurd

enough said

[spare quartz]

External Dependency

🪦

[timid quartz]

every minute my kernel stays alive proves how good my code is

[spare quartz]

hmmm

I should write my own Intel firmware update

😁

[timid quartz]

lmao have fun bricking your cpu

new vuln: furdown

[spare quartz]

remember… everything’s an external dependency..

[timid quartz]

Your entire kernel is an external dependency

[spare quartz]

no

[timid quartz]

Whereas mine is all bundled together in a nice little package

[spare quartz]

(full of external dependencies)

[timid quartz]

nah

Meanwhile if your kernel has no external dependencies it wouldn’t do anything except IPC

but there’d also be no processes

[spare quartz]

nah

we just do a little outsourcing

the KERNEL doesn’t depend on anything

[timid quartz]

(external dependencies)

[spare quartz]

no.

https://youtube.com/shorts/V5OpYRz5_Ms?si=w_Vu0qxdjIerc7N_

[lavish dove]

thats what they do with waterproof laptops

[lavish dove]

I was reading the intel docs

they are so fucking good

[radiant slate]

Die

[spare quartz]

https://tenor.com/view/ifunny-mr-stark-meme-avengers-gif-21407304

inventing new ways for QEMU to fail but pass on the simulator

jk

[spare quartz]

good news

our bootloader's size limit went frrom 512 bytes

to

523.00800 kilobytes

[timid quartz]

What are you gonna do with all those bytes

[spare quartz]

whatever we want

don't need to load a second stage anymore

[timid quartz]

ngl…I see why eating tide pods became a trend briefly…they feel kinda scrumptious

[spare quartz]

rust dev activities

we briefly had the kernel overwrite the bootloader

[timid quartz]

I mean once the bootloader is done…

doesn’t rlly matter

[spare quartz]

well yeah but this was during section loading 💀

having the loader set to 00s before it can load everything fully and jump is not optimal

[timid quartz]

o h

[spare quartz]

ultimately i also had to make QEMU load in 1 section at a time

or else i'd get a DMA error

[timid quartz]

L

[spare quartz]

not l

just annoying

[timid quartz]

but now are you able to not do that?

[spare quartz]

wdym

[timid quartz]

like

can you load multiple sections with no error

[spare quartz]

yeah

[spare quartz]

basically what this dose

iso_9660_directory_read:
    mov di, 0x7E01
    mov cx, 0xFFFF
    repne scasw
    sub di, 2
    push dx
    mov ax, [di-23]
    mov dx, [di-21]
    mov cx, 2048
    div cx
    cmp dx, 0
    pop dx
    jz .iso_9660_directory_configure_load
    inc ax
.iso_9660_directory_configure_load:
    mov cx, ax
    mov DWORD [si+4], 0x00008600
    push DWORD [si+4]
    mov ebx, DWORD [edi-31]
    mov DWORD [si+8], ebx
.iso_9660_directory_load_chunked:
    call int13_extended_read_procedure
    add WORD [si+6], 0x80
    add DWORD [si+8], 1
    loop .iso_9660_directory_load_chunked
    pop DWORD [si+4]
    ret
pvd_disk_address_packet:
    db 0x10               ; Size
    db 0x00               ; Reserved
    dw 0x0001             ; Sector Count
    dw 0x0000             ; Transfer Offset
    dw 0x07E0             ; Transfer Segment
    dq 0x0000000000000010 ; Disk Block Number

scans for the file name (e.g. "KERN"), uses an offset from the name to get position in file/size

divides size by 2048 to get block count, sets the DAP to use the block position

and then loop adding 0x80 to the segment position of the DAP until block count is zero

additionally at the end it sets the DAP transfer position to the starting value with the pop there

probably not the best code but i tried being frugal

[spare quartz]

LOL

WRONG IMGE!!!

@timid quartz

[timid quartz]

what

[spare quartz]

fixed

do NOT look at logs

[timid quartz]

im looking in audit logs

[spare quartz]

NO

[timid quartz]

interesting.

[spare quartz]

pls tell me you dont actually log images 😭😭

[timid quartz]

:3

[spare quartz]

the video that thumbnail is from would get me banned 900 times over

😭

[timid quartz]

I can't find it on yt

[spare quartz]

lemme get the link

its a video where bayachao actually speaks in english (albeit not the greatest)

[rustic vine]

https://cdn.discordapp.com/emojis/1316212547287781386.webp?size=96

[spare quartz]

no, it's not like that

actually

it is definitely like that

carry on

[rustic vine]

you're lucky nobody is locked in on moderating this place

[spare quartz]

cause we all moderate ourselves :3

simply a fault of windows clipboard for sending it

[rustic vine]

uh huh

everyone except for you apparently

[spare quartz]

SIMPLY

a fault of windows clipboard

[proud creek]

Did I hear moderate

[spare quartz]

no

you heard coding

[proud creek]

[spare quartz]

idk tz wearing a fursuit all day isnt good for acoustics

[proud creek]

[rustic vine]

oops! all code

I made a code joke in a non code channel and it went silent lol

[spare quartz]

whens the last time you shwoered

[rustic vine]

this morning

[spare quartz]

☣️

[rustic vine]

whens the last time YOU showered

[spare quartz]

no comment

[rustic vine]

brah

[spare quartz]

im kiddinggggg

like 12 hrs ago

[rustic vine]

I've always showered every day and people tell me its bad

but idk my hair is fine so I've ignored them thus far

[spare quartz]

i think the only bad thing there is if you're constantly rinsing it in long showers

but different people have different hair

[rustic vine]

ye

[spare quartz]

uuugggggghhhhhhhh

running xorriso and nasm is so tedious

gonna try and make a simple assembler thing soon..

[spare quartz]

neat

[lavish dove]

im using parted and mtools

got fuckin annoyed by xorriso

[spare quartz]

ive still got no experience in how compilers work

kys emoji user

[timid quartz]

I mean you can build an interpreter

So just uh

Instead of executing the code

Interpret it to binary

[rustic vine]

invert the plant

[spare quartz]

oh

wow

you both

asy stuff

[spare quartz]

em, what

you can't do that with assembly

[timid quartz]

no read what I said again

an interpreter executes the code

but as a part of that it has to parse each line

so replace the "code execution" part of an interpreter with a part that translates the assembly into binary

[rustic vine]

just invert the plant

[spare quartz]

...which is a compiler

[timid quartz]

yeah

that's what I was trying to say

[spare quartz]

im not saying i dont know the concept/how to build one

[timid quartz]

you already know how to make an interpreter

so just...

change a little bit

[spare quartz]

im saying i haven't done this before 😭

excluding sql

[timid quartz]

now if you want to do optimizations and jazz

[rustic vine]

🥺

[spare quartz]

i know multiple passes

[timid quartz]

that's an entirely different thing that leans into CS theory

but it's still neat

[rustic vine]

anyone familiar with these numbers

[spare quartz]

i know a bit how optimizations are implemented within compilers

[timid quartz]

color a graph in polynomal time

[spare quartz]

in asm multipass optimization is very simple

[timid quartz]

right but what kind of optimizations are you doing multiple passes for :3

[spare quartz]

but the subject being replaced has 30 years of updates im up against

[timid quartz]

tbh honestly

do some uh

[spare quartz]

https://userpages.cs.umbc.edu/chang/cs313.f12/nasmdoc/html/nasmdoc2.html

match case for "-O"

[timid quartz]

right but what kinds of optimizations do those entail :3

[spare quartz]

anyways since im busy playing wolfenstein i havent been able to work on it much

[timid quartz]

what kinds of things are you doing in your multiple passes :3

[spare quartz]

it.. says it

literally read the thing

[rustic vine]

tough request in todays world

[timid quartz]

mm it's definitely doing more than it's specifying

[rustic vine]

HUH

[timid quartz]

gotta do your constant propogation, redundancy elimination, reaching definitions, dead code elimination

[spare quartz]

the NASM testfiles for opts can be found here

https://github.com/ncoden/NASM/blob/master/test/optimization.asm

[timid quartz]

data flow analysis

[spare quartz]

extrapolate and search the files as you see fit

[timid quartz]

.

[spare quartz]

and most of that is irrelevant in ASM

[timid quartz]

this is what is done in the multiple passes

[spare quartz]

you cannot do "dead code elimination" for example

constants aren't a thing either

they're just a S&R

[timid quartz]

constant propogation isn't necessarily a constant how you would think

[rustic vine]

im gonna wait for the smart people to finish talking so I can ask my dumb question 😃

[spare quartz]

what is your dumb question

we can multitask

[timid quartz]

although this kinda falls under redundancy elimination, you can do stuff like detecting that you compute a+b once and don't ever re-define a or b and then replace every re-calculation of a+b with either the evaluated constant or a reference if you cant evaluate it at comptime

[spare quartz]

can't do that in asm either

[timid quartz]

...yeah you can

[spare quartz]

ADD, SUB, etc. all affect CPU flags

in ASM this is significant

[rustic vine]

im trying to understand how the butcher tableau of RK integration methods were calculated/came-about

idk how it works cause wikipedia just gives them to you as magic numbers and doesn't elaborate

and so does every other paper

[timid quartz]

if your assembly is being outputted by gcc then yeah gcc already does most of this

[rustic vine]

its not a big deal though. I don't need to know, I was just curious

[timid quartz]

but if your assembly is being written directly by humans then there's places to optimize

[spare quartz]

GCC is not an assembler

[timid quartz]

gcc can output assembly stupid

gcc -s

[spare quartz]

it's not an assembler though

[rustic vine]

this code sucks session terminated

[spare quartz]

"smart people talking" and you say this

[rustic vine]

no bruh I swear its simple

the RK4 one looks really simple

but the dormand-price one is like 20 random numbers

https://cdn.discordapp.com/attachments/1264496483957346346/1348140389256794153/image.png?ex=67ce6114&is=67cd0f94&hm=2d24985d1e9bc2dd8759bfd0919392248786c8a76f40255a1d680fcbc2929c66&

this

I think bro just willed this into existence one day

[spare quartz]

where are the values used

[timid quartz]

ew numbers

[rustic vine]

dormand price integration

its the coeffs for it

[timid quartz]

ok well seems like we're talking about different things

you had mentioned that you didn't have experience with implementing a compiler and then linked an assembler

[spare quartz]

the only thing i can think about when given "seemingly random" numbers like this is just as a way to fingerprint, but thats about all i can give

probably not whats going on here

[spare quartz]

because an assembler is a compiler, just with less steps and/or a different outcome

an assemblers job is to turn assembly language into machine code, a compilers job is to do the same, but at a more generic level: it could convert C to IR for example

[rustic vine]

yeah I think I'm just meant to take these coeffs at face value and not try to think about why they are what they are

thats fine

I was just wondering if anyone had any insight

[spare quartz]

IR doesn't have the same "significances" as ASM like flag or data placement, the compiler can put whatever

[timid quartz]

I am well aware

[spare quartz]

therefore, optimizations exist, but they're different or not relevant to assembly

[timid quartz]

the general flow of a high level compiler is to take your high level language, make it IR and optimize the IR, and then take the IR and output machine assembly for it

im taking a class on this fgs

[spare quartz]

i get that 😭 im just trying to make this clear

[timid quartz]

I mean ig not

the fact that you can't optimize assembly in a similar way to IR doesn't entirely make sense

because I can write two different .s files that do the same thing but one is more "optimal"

[spare quartz]

that's too generic though

[timid quartz]

without caring about CPU state or flags too much

[spare quartz]

you can make a superior version of something, thats the end result, but we're talking about the destination to the result

[spare quartz]

the assembler just can't make the type of assumptions a regular higher-level compiler could is the fact of the matter

(it's always about assumptions...)

[timid quartz]

im assuming that i can send a missile to your house

[spare quartz]

nuh uh

hey wait

[timid quartz]

too bad

[spare quartz]

no nvm

i was gonna ask if you had a kernel written in rust i could maybe see

[timid quartz]

no I don't

[spare quartz]

but even if you did have one it'd probably be too complex for a test sample :<

[timid quartz]

it'd probably be too fat

[spare quartz]

i should really write unit tests for our processor

viewing disassembler output and stepping in qemu at the same time makes my brain itch

[spare quartz]

the thing is even something as inoccuous as QEMU is really bad to debug

since it's literally it's own micro-kernel

it does so many instructions that it just causes my output to die

NOT QEMU

god ... GRUB

always get the two mixed

[timid quartz]

well at least with rust

you could be delivered from gdb

and go to a similar hell called lldb

[spare quartz]

what's lldb

llvms knockoff of gdb?

[timid quartz]

gdb for llvm programs

yeah

[timid quartz]

wait...

you're not using kotlin/native are you

for your kernel

[spare quartz]

.. why would i??

and do you know how many hoops i'd have to go thru

ufedmi_k is written in ada

[timid quartz]

idk cause you like Kotlin and you like doing crazy things that are unconventional and challenging

[spare quartz]

um

a cpu in minecraft is very conventional

and two

not challenging. just tedious.

everything in programming is easy its just tedious

[timid quartz]

ok then uh

Write an algorithm that colors a graph in polynomial time

[spare quartz]

i dont know what polynomial time means or what a polynomial is

also what does color mean in this context

or graph

[timid quartz]

This guy doesn’t know graph theory or big O

[spare quartz]

YEAH CAUSE I DONT DO MATH

[timid quartz]

It’s not really math

Ok so basically

[spare quartz]

its all math

once you're theorizing instead of coding, that's math

[timid quartz]

A graph is a collection of nodes linked together by edges

[spare quartz]

what is an edge

[timid quartz]

Like this is a graph for example

[spare quartz]

is an edge another name for reference

[timid quartz]

I mean you could think of it that way but this has little to do with code right now

[spare quartz]

aera i literally can only code 😭

[timid quartz]

An edge just indicates some “connection” between the nodes

Like it could be a road between a city

Or a reference

There’s different ways to think about it

[spare quartz]

well a reference is just another name for a datums relation to another datum

e.g. a pointer is a reference to memory address from a memory address

[timid quartz]

A valid graph “coloring” assigns “colors” to nodes such that no two nodes connected by an edge have the same color

[spare quartz]

ohhh

[timid quartz]

So like for this, a valid coloring is
0: red
1: blue
2: green
3: red
4: blue

Now

Since everything in programming is easy

Write me an algorithm that takes a graph

Oh wait I didn’t explain polytime

[spare quartz]

so this would also be valid right

[timid quartz]

Yes

That’s valid

[spare quartz]

cool okay

[timid quartz]

ok so basically big O is a measure of the worst case runtime of a program relative to the program’s inputs

[spare quartz]

i know what big o means

[timid quartz]

So O(1) or “constant time” means that your program will always have the same runtime no matter the size of the input

[spare quartz]

just not how to actually calculate or applicate it

[timid quartz]

Polynomial time is O(n) meaning the programs runtime scales linearly with the input size

[lavish dove]

are you guys talking about graphs

[spare quartz]

he is

im listening

[timid quartz]

Yeah ATP said everything in programming is easy so I’m challenging ATP to write an algorithm that colors a graph in polynomial time

But had to give some background

[spare quartz]

so how would i know if my algorithm is polynomial time

[lavish dove]

wait so

why would you need a graph in polynomial time

isnt the point of a graph to be simple to read

[spare quartz]

its a challenge

[timid quartz]

Specifically you need to do a 3-coloring

In polynomial time

[spare quartz]

but what does polynomial time actually implicate

[timid quartz]

3-coloring means “use at most 3 colors”

[spare quartz]

as far as i understand running a time benchmark on an algorithm and deriving big o is useless

theres too many variables

[timid quartz]

So practically

Something that would be O(n) is for example idk

[spare quartz]

a loop

over n

[timid quartz]

Iterating over an entire array with a for loop

Yeah

Because the more elements in the array

The more loops you do

And it’s 1:1

[spare quartz]

oh wait so

big o isn't actually definitive of the actual algorithm

its just how it would perform if it were like

[timid quartz]

So for a 3-coloring of a graph, a polytime algorithm would only iterate over each node once

[spare quartz]

an abstract computation taken literally... i think

[timid quartz]

Or well it wouldn’t have any nested loops

[spare quartz]

(wait no, big o is for the ALGORITHM, not the IMPLEMENTATION, at least on the actual execution layer)

[spare quartz]

okay ill try..

lemme open roblox

[timid quartz]

Yeah big O is abstract

[spare quartz]

before i write some lua

let me send an image thats gonna make me explode

[timid quartz]

Well like for example you can look at someone’s code and determine the big O

[spare quartz]

ooooookayyy graph representation,,,

[spare quartz]

i couldnt

[timid quartz]

It’s a thing that’s possible

One class I took was literally so stringent about big O

[spare quartz]

like taking this immediately at face value

[timid quartz]

We had to implement stuff for data structures within certain big O constraints

[timid quartz]

I can already tell your code is O(n) (which it should be tbh) because you’re going through every element of your stream

[spare quartz]

well.. kinda..

an element could itself pull another element

wouldn't that make that iteration... not "1" in n

[timid quartz]

well that wouldn’t really increase the big O

What WOULD increase the big O is if one of your elements were to cause another iteration over the entire stream

So like

[spare quartz]

a recursive function over an array?

[timid quartz]

while …
“case” => while …

[timid quartz]

you probably don’t know sorting algorithms but “naive” sorting algorithms are O(n^2) because, for each element in the array, you’re comparing it to every other element to find the right spot

[spare quartz]

i only know one sorting algorithm

all the others i just know because they're of use for (problem)

and promptly forget

[timid quartz]

what algo I’m curious

[spare quartz]

what everyone should know

bubble

[timid quartz]

ah yes the venerable bubble sort

a classic O(n^2) example

[spare quartz]

(it is EXCEEDINGLY rare for me to need a sorting algorithm which is why..)

[timid quartz]

I’m personally an in-place quick sort fan

But I can’t implement it off the top of my head cause I’ve never needed to

[spare quartz]

ada (23, if you use a compiler compliant with the parallel RFC) makes a parallel divide/conquer sort pretty easy to implement, i forgot exactly which algorithm it is in the example though

[timid quartz]

anyways I’m going to sleep have fun 3-coloring graphs in O(n)

[spare quartz]

first i gotta find out how i wanna implement it first :<

good enough

[timid quartz]

would make Graph a collection of Node

And Node contain a number, color, and array of other Nodes to represent your edges

[spare quartz]

thats too much

at least what im thinking about

[timid quartz]

Well you gotta have the basics at least

You gotta have nodes with room for color in them

And you gotta store your edges somewhere

[spare quartz]

well

a graph is an abstract thing, right?

[timid quartz]

A graph can be used to represent many things

[spare quartz]

local colored = { [color]: { node ref } }

therefore no data needs to be in the node

hm

[timid quartz]

Yeah tbh you don’t need to store data so

[spare quartz]

yknow what i think i can just have graph be an array of graphs

a memory address is good enough

[timid quartz]

also valid because you can consider a graph as a set of sub-graphs

[spare quartz]

i guess that would make

the top node the graph, and the subnodes.. the nodes

[timid quartz]

There’s not really a top node in a graph

You’re thinking of trees

[spare quartz]

well, it has to be contained somewhere

[timid quartz]

A graph is looser than a tree

[spare quartz]

hm

would a tree be a graph IF the "top nodes'" connections didnt count as an edge

[timid quartz]

A tree is a graph

Always

It’s like how a square is always a rectangle but a rectangle isn’t always a square

A tree is just a type of graph

[spare quartz]

okay so this setup is okay then

[timid quartz]

I’m asking you to make an algorithm to 3-color graphs in general

[spare quartz]

yes yes i just wanna make sure i understand everything

[timid quartz]

And a general graph may not have the same restrictions as a tree

[spare quartz]

anyways this suffices

[timid quartz]

I mean if you can understand how to represent edges there

It’s a lot easier for me to even think as just a Node object

And have the edges be an array of Node <-> Node pairs

[spare quartz]

well

the existence of a graph within another graph is enough to count as an edge between those two

[timid quartz]

Mm the only question is

Where does the edge connect

[spare quartz]

between the contained and container edge

[timid quartz]

Because if you have say “A B” and “C D” with an edge between them

What exactly two nodes does the edge connect?

This is important

An edge always connects exactly two specific nodes

[timid quartz]

Right that’s a tree

[spare quartz]

thinking cap time

[timid quartz]

But a graph could also be like

That

[spare quartz]

A

[spare quartz]

this could be represented in this model too

G4 just contains G1

(implementation wise this could cause errors...)

[timid quartz]

Ok well as long as you think you have things represented correctly

[spare quartz]

i just gotta think

[timid quartz]

If you hit a roadblock with your representation then I recommend the approach I mentioned earlier cause it’s very easy to reason with

Anyways I sleep

[spare quartz]

kohaku,,,

bayachao,,

<3

@timid quartz

type Graph = { Graph }
type ColorGraph = { [number]: { [Graph]: boolean } }
local graph: Graph = { {}, {}, { {}, {} } }

local function colorize(g1: Graph, gc: ColorGraph?): ColorGraph
    local gc = gc or {}
    local pc = (gc[1] or {})[g1]
    for _, g2 in g1 do
        local idx = if pc then 2 else 1
        local chk = gc[idx]
        if not chk then
            chk = {}
            gc[idx] = chk
        end
        chk[g2] = true
        colorize(g2, gc)
    end
    return gc
end
-- All elements are iterated thru exactly once, "O(n)"
-- CS is weird

print(graph)
print(colorize(graph))

plz confirm

[lavish dove]

woah

to make a bios for qemu all you haved to do is have 65536 aligned code

it even starts at rip=0

[timid quartz]

Could you draw the graph

I can’t really visualize how you’ve made it

[spare quartz]

Mmm

I’m on mobile rn,,,

The numbers below are the colors

The memory addresses are the nodes

And an edge is the existence of a node in* the containing node

[timid quartz]

yeah I think I just can't really visualize the edges

[spare quartz]

Tried

Oh god that looks awfully compressed

[timid quartz]

ok that's what I thought I just wanted to make sure

your graph is quite trivial

but you did successfully color it

now try coloring

. A
/
B -- C

wtf

ok thanks discord

[spare quartz]

Okay let me
Wake up

[spare quartz]

Not awake yet I lied

OK

now im awake

sooo

[spare quartz]

to do this ill need to do two things (after wolfenstein)

• create a secondary map for [Node] -> color indexing

and

• add a color queue based on g1/g2

[spare quartz]

accidentally discovered katakanam ode

[spare quartz]

@timid quartz okay back to coding graphs

hopefully this is "nontrivial" enouggh for your specifications

[timid quartz]

oh ill give you another one after this

because this one is also fairly trivial

[spare quartz]

okay give it to me right now if its the same topic

[timid quartz]

ugh fine

[spare quartz]

cause chances are ill make this function specifically for that graph

[spare quartz]

DEVIL CONNECTION REFERENCE

[timid quartz]

your algorithm needs to be generic enough to work on any graph given to it

[spare quartz]

ばやちゃお！！！！！！！

[timid quartz]

and if a graph isn't 3-colorable you can uh

return whatever you like to indicate that

[spare quartz]

throw an error

👍

[spare quartz]

anyways lemme see on how i would implement this

this is the type of problems kids always ask about

"when will we ever do this in the real world?"

it is a 1 in 1 billion chance of getting any practical use out of needing to color a devil summoning circle in your code

[timid quartz]

welll

in a compiler, graph coloring is useful

for register allocation

[spare quartz]

yeah but only mega nerds would call that graph coloring

i'd just call it a stack

anyways arrangement set, variable time

[timid quartz]

actually tbh

your algorithm can find the fewest possible colors

[spare quartz]

wdym

[timid quartz]

just

keep doing whatever you've been doing

[spare quartz]

i can't

not without adding unneeded complexity

that algorithm i came up with is specifically for trees with no circular references

[timid quartz]

:3

well since everything in programming is easy just tedious

ill wait

[spare quartz]

it is

you've in fact, given me a proof of that

by having to type out the self referential nightmare of this

[timid quartz]

I just don't think you know enough CS theory to understand what I'm asking you to do

[spare quartz]

you've given me a specification

i can do specifications

i don't need theory

[timid quartz]

well...

knowledge of a certain part of theory would have made this entire problem a lot easier

or at least arriving at the correct answer

[spare quartz]

well, if you could, tell me how this would actually matter in an application

[timid quartz]

a compiler doing register allocation is most forthcoming

[spare quartz]

because this might appear in your theory, but, it's not how something, at least, anything that i could come up with, would actually function

i love discord freezing

[timid quartz]

like when you're transforming IR (which has an infinite number of virtual registers) to assembly (which has a finite number of physical registers), to fit everything in the registers in an efficient manner, one approach is building a graph where the nodes represent variables and edges represent an "interference" (where both need to be in registers at the same time)

you then use the physical registers as colors

if you can color the graph then you're chill

if you can't then you have to "spill" one or more virtual registers onto the stack

[spare quartz]

why can't you just

not have registers exist at the IR level

and have it all be the stack

then during MCG make the top part stack register bound

[timid quartz]

I mean you COULD but consider the case where you want to minimize memory ops

[spare quartz]

yeah

[timid quartz]

and even then you often end up with multiple variables overlapping

even if you do store them all on the stack

requiring some management of physical registers

[spare quartz]

yeah thats why im saying making the top part of the IR stack actually register bound in the mc

and then the parts not in the register bound part of the ir stack are in the mc stack

(this may or may not actually work to be frankly honest: just how i would approach)

[timid quartz]

I mean that's a naive allocation yes

but it's not an optimal allocation

[spare quartz]

i dont see why not

[timid quartz]

Imagine you have a variable in the machine code stack right

[spare quartz]

sure

[timid quartz]

you could end up with a portion of your code where one of the variables stored in the registers gets stored to memory and isn't used for a time

but your allocation would probably just keep it in the register unused

you'd need some further kind of analysis to detect that

[spare quartz]

hmmm.. good point... but we can see where each part of the IR stack is used, and when it is used

if something isn't used very often that's in the upper half of the IR stack, we can lower it and bring up the thing ahead further up

if no further uses for that thing exist in the function anywhere ahead, we can just pop it

[timid quartz]

and now you've taken half a step towards what I mentioned anyways

cause that doesn't entirely solve the above problem it just reduces when it could occur and the impact it would have

[spare quartz]

well i don't see any way the graph would help solve it either

[timid quartz]

with each node+number representing a "virtual register" or "IR variable" (however you want to think about it cause they mean the same thing), with each edge representing variables being needed at the same time ("interfering"), and with each color being a physical register

this identifies that 1, 2, and 4 are never needed at the same time, so the blue register can be used for each of them whenever they're needed

whereas with your implementation (and still assuming 3 physical registers), you'd probably just have 1 and 2 permanently stored in registers, and use the last register to swap between 3, 4, and 5 that are stored on the machine stack

which while it works is less optimal

[rustic vine]

https://cdn.discordapp.com/attachments/988661200827072592/1348380474803880088/tempFileForShare_20250309-154229.jpg?ex=67cf40ac&is=67cdef2c&hm=a04a5556d9fb1933790548a64ff5275dfdb1770fbd471e66333bd33f9ed48346&

[spare quartz]

hmmm

the allocation in each register would depend on how exactly 1/2 are used

if its in a spot of rich use of 1/2, then yeah they'd probably be stored for the entire duration

but if that suddenly becomes 1/4, 2 falls

[timid quartz]

but at this point you're basically doing what the graph does

[spare quartz]

i dunno i just can't think about how you can with graphs

[timid quartz]

but arguably less convoluted

well

the GRAPH is less convoluted

imo

I can reason with the graph a lot easier

[spare quartz]

for me the IR/MC stacks are just easier to keep in my head

i still have not typed out the pentagram vars yet 😭

[timid quartz]

and yes this is theory but this also gets applied via programming

so yknow...either doing register allocation via graph coloring would be really tedious for you...

or it wouldn't be as easy as you're saying...

[spare quartz]

it's as easy as im saying

just tedious again 👍

[spare quartz]

as you've said... what im doing is effectively what you're doing

just in another format

[timid quartz]

it's effectively the same thing

but more convoluted and probably less optimal

anyways you can stop with the graph coloring now

[spare quartz]

I DIDNT EVEN GET TO WRITE ANY CODE 😭

[timid quartz]

well if you want

[spare quartz]

stupid cs students tricking me into talking over actually doing my job

[timid quartz]

but I'll just go ahead and tell you

I had full intention of watching you burn yourself out solving a problem that people smarter than either of us combined are 90% sure is impossible

[spare quartz]

well those people are stupid

[timid quartz]

our current math and theory knowledge holds that finding the minimal coloring of a graph is not possible in polynomial time

you can look up NP-completeness if you want the theory

[spare quartz]

well

at least tree coloring is possible...

[timid quartz]

tree coloring is trivial

because a tree is basically a bipartite graph

and a bipartite graph is always 2-colorable

[spare quartz]

thats a word i do not know about funny man

[timid quartz]

(bipartite meaning you can separate the nodes of a graph into two "sides" where all the nodes on each side have no edges between them, only edges to the other side)

[spare quartz]

hmmm yes yes

i do not understand a word

[spare quartz]

anyways lemme try now

[timid quartz]

this graph is bipartite

[spare quartz]

doing something weird within a certain iteration that is itself not iteration

is still polynomial time right

[timid quartz]

mmm it depends

[spare quartz]

like if i have a giant if chain, but none of the results causes an iteration, just some arithmetic or whatever

[timid quartz]

considering what we know it would probably make it not polynomial

[spare quartz]

but only given another condition

would that be polynomial

[timid quartz]

are these if statements traversing nodes

[spare quartz]

no

operations within the same and container node

[timid quartz]

mm it might remain polynomial but might not be correct in all cases

well

would PROBABLY not be correct in all cases

someone would probably find a counter case

[spare quartz]

thats a lot of probablys

[timid quartz]

the gist of NP-completeness is this

• some problems are solveable in polynomal time (we call the set of those P)
• some problems are verifiable in polynomail time (we call the set of those NP)
  • verifiable means that someone can check it for correctness, so like they give you their "solution" and you can check it for correctness in polynomial time
• some problems are neither (we call the set of those NP-Hard)
• NP-complete is the overlap between NP and NP-Hard
• somebody smarter than us made a giant proof for one particular problem showing that it is NP-Complete
  • since then we've found ways to "coerce" other problems into being that one particular problem, and by doing that, we can say those problems are also NP-Complete

[spare quartz]

just making a reference for myself ignroe

[timid quartz]

if you were to somehow solve the minimal coloring problem in polynomal time you would have proved P = NP and you'd probably make millions

[spare quartz]

could you imagine how much bayachao merch i could buy with that money

[spare quartz]

anyways rx'd ... who made that problem if you know

[timid quartz]

uhh the problem itself that has the giant proof is called "satisfiability" or SAT

[spare quartz]

so multiple collaborators i assume?

[timid quartz]

SAT is basicallt

you're given a boolean expression that's expressed as ANDs of ORs

so (... or ... or ...) and (... or ... or ...) and ...

[spare quartz]

oh wait is this logic

[timid quartz]

yeah

and finding a "satisfying" assignment s.t. the entire expression is true is NP-complete

the theorem that pertains to the NP-completeness of SAT is https://en.wikipedia.org/wiki/Cook–Levin_theorem

[spare quartz]

let him cook

curious question

does logic account for like

take (A and B) or C

does both A and B and C have to be computed

or is short circuiting just A and B ok

[timid quartz]

I mean you could but that problem is of the wrong form for SAT

an input for SAT must have the form (... or ...) and (... or ...) and ...

[timid quartz]

this is just trivial

[spare quartz]

trivial in what way

[timid quartz]

well you just find one thing to satisfy either clause of the or

and you're done

solveable in P

you can extrapolate that to any number of (... and ...) blocks

[spare quartz]

im just thinking about if https://en.wikipedia.org/wiki/Lazy_evaluation exists in this math crap

[timid quartz]

it does

but again

for SAT

your problem is in the wrong form

[timid quartz]

which is why solving a problem that looks like (A and B) or C is trivial

because you can assign C = True and be done

[spare quartz]

hmm

[timid quartz]

you effectively "short circuit" the entire expression

[spare quartz]

soo

((A or A) and (B or B)) and ((not ((A or A) and (B or B))) and C)

would this be in the right form

[timid quartz]

im not sure about the nesting

yeah you can't have the nesting

[spare quartz]

computationally it's the exact same as what i've sent

ah ookay

hmmm then..

[timid quartz]

conjunctive normal form

that's the rules for it

[spare quartz]

peculiar

so i can't do not; a and b, but i can do not a; and; not b

is what im hearing

[timid quartz]

yeah

[spare quartz]

i wonder why that is (but i wont instigate on it since this theory stuff isn't something i plan to become a student in, just "know" for whenever)

[timid quartz]

idk some math jazz that math people agreed on

[spare quartz]

this is why we hate mathematicians

[timid quartz]

but the way that you can "prove" that the graph coloring problem is np-complete is by morphing it into SAT

you can consider 3-SAT where the only difference is that each disjunction only has at most 3 variables

for each variable you can construct a node in the graph

[timid quartz]

man I forgot the reduction

anyways

you can morph SAT into looking like the 3-coloring problem

and you can morph the 3-coloring problem into the k-coloring problem where k is some arbitrary number of colors >= 3

[spare quartz]

hold on i need to think about the problem

hold onto this for a second

there is an exploitable pattern here

in fact im pretty sure i've made it, just as a joke once while doing power simulation

just gotta figure out how i did it

also hold this

[spare quartz]

HOLY SHIT

[proud creek]

Orb

[spare quartz]

https://wetdry.world/deck/@[email protected]/114135817446309557

@timid quartz you type of post

[timid quartz]

real

[spare quartz]

ファック

[timid quartz]

Yknow this made me think

Stack cringed architectures like the JVM are basically just bloated Turing-complete reverse Polish notation calculators

[spare quartz]

i will kill you

also wdym by "(reverse) Polish notation calculators"

[timid quartz]

Google it

[spare quartz]

i did

i saw the syntax was "a; b; op"

[timid quartz]

Yes

[spare quartz]

which is how every architecture works

the jvm included

[timid quartz]

Hold

[spare quartz]

PUSH A
PUSH B
ADD <pops A, B>

MOV EAX, 40
MOV ECX, 2
DIV [[EDX]:EAX, ECX]

also twitter is down ☹️

no more bayachao content for me

[timid quartz]

Yeah so

Basically every RPN calculator I’ve seen and used also operates similar to Java with a stack

4<ENTER>
3<ENTER>
+<ENTER>

[spare quartz]

yeah thats how every standard architecture works though

not specially the jvm

[timid quartz]

It just reminded me of the JVM because of this

[spare quartz]

well

[timid quartz]

[spare quartz]

that just means you have to write your own jvm

:3

[timid quartz]

pixiv time

[spare quartz]

she has a bsky so thats good enough

also their new game won an award so thats cool

[timid quartz]

I think the JVM is slightly different and is more similar to an RPN calculator

[spare quartz]

i guess so

[timid quartz]

Because like you said in JVM bytecode you’d do like PUSH PUSH ADD

[spare quartz]

yeah

USUALLY you'd do the same in x86/ARM with MOV reg/reg, but that could be out of order

[timid quartz]

whereas with like x86 you’d do LW into reg LW into reg ADD both regs

I think the difference that I noticed was that the arithmetic operations in the JVM implicitly use the stack whereas other architectures may be explicit

Arithmetic operations among other things

Similar to how if you were to press the plus key on an RPN calculator, you would implicitly use the top two items on the stack

[spare quartz]

honestly im not sure what architecture the jvm is closest to (aside from its discrete details like stack-based)

i know it's kind-of a RISC

since most instructions are a byte long, and those that aren't have discrete sizes

[spare quartz]

also wait

wdym by this

implicit, in what way?

[timid quartz]

stack-cringed*

[spare quartz]

i should NEVER give you another post from fedi again

[timid quartz]

like you do ADD rather than ADD <explicit values>

[spare quartz]

well, the same also happens in x86

[timid quartz]

the use of values from the stack is implicit

[timid quartz]

?

[spare quartz]

a DIV operation for example

it implicity operates and returns in D/A

the only operand is the divisor

[timid quartz]

hm didnt know that

[spare quartz]

MUL too

[timid quartz]

i dont know that much x86 :p

[spare quartz]

you should make your own cpu...

[timid quartz]

[spare quartz]

itlll give you extra crediot..

[timid quartz]

noitwont

[spare quartz]

:<

[spare quartz]

pizza TTL 41 minutes

[timid quartz]

I believe you'll find TTL is not measured in minutes smh

how many hops for the pizza

[spare quartz]

sir

i am not writing an extension to TCP for pizza

this is my own system.

[timid quartz]

well good luck measuring time between multiple hops :3

[spare quartz]

ermm

japan postal service and usps... would like to talk to you..

[spare quartz]

BOOTLAODER DEV RESUMATION time because i have been lazy

[spare quartz]

it just hit me

<CR> isnt useless as a character

its just dated

[pallid loom]

its still used a lot

iirc by default the arduino ide does both \r & \n for newlines

[spare quartz]

not that

most programs see it as virtually the same as <SP>

[pallid loom]

"In both ASCII and Unicode, the carriage return is assigned code point 13 (or 0D in hexadecimal); it may also be seen as control+M or ^M. In character and string constants in the C programming language and in many other languages (including representations of regular expressions[2][3]) influenced by C, \r denotes this character.[4]"

[spare quartz]

im just looking at replicating QEMU and LF* is for y++ and CR is for x = 0

[snow oak]

dear programmers in the chat...

local vm_module = {}

local ReplicatedStorage = game:GetService("ReplicatedStorage")
local RunService = game:GetService("RunService")

local modules_folder = ReplicatedStorage:WaitForChild("Modules")
local events_folder = ReplicatedStorage:WaitForChild("Events")
local viewmodels_folder = ReplicatedStorage:WaitForChild("Viewmodels")

local current_camera = workspace.CurrentCamera

vm_module.functions = {
    on_equip = function(tool: Tool)
        if not tool then
            return
        end
        
        local possible_vm = viewmodels_folder:FindFirstChild(tool.Name)
        
        if possible_vm then
            local real_vm = possible_vm:Clone()
            
            real_vm.Parent = vm_module.objects.Camera
            vm_module.functions.viewmodel_loop(real_vm)
        end
    end,
    
    on_unequip = function()
        local vm = current_camera:FindFirstChildOfClass("Model")
        
        if vm then
            if vm:FindFirstChild("CameraBone") then
                vm:Destroy()
            else
                return
            end
        else
            return
        end
    end,
    
    viewmodel_loop = function(vm)
        if not vm then
            return
        end
        
        RunService.RenderStepped:Connect(function()
            vm:PivotTo(current_camera.CFrame)
        end)
    end,
}

return vm_module

[spare quartz]

which i found peculiar

[pallid loom]

you're looking at doing what

[snow oak]

using this module on the client if you couldn't tell already

[pallid loom]

it'd be nice if you said the problem and added debug statements

[snow oak]

nope

it works

just looking for any tips

[pallid loom]

i dont understand

are you trying to do qemu in mc

[spare quartz]

no that's too simple

[pallid loom]

and to be clear we're talking about

qemu

the virtual machine software

not something else

[spare quartz]

this is our implementation of a BIOS and x86 cpu for our mod, im just trying to replicate QEMU's bios since most operating systems are aqquainted with it

[spare quartz]

and yes

[pallid loom]

why

[spare quartz]

cause its fun

[pallid loom]

incredible

incredible amounts of radiating autism

[spare quartz]

anyways the issue is X isn't being calibrated properly on our simulator, which causes a deficit of exactly 1 blank line

i am not autistic .

[pallid loom]

acoustic

[snow oak]

local vm_module = {}

local ReplicatedStorage = game:GetService("ReplicatedStorage")
local RunService = game:GetService("RunService")

local modules_folder = ReplicatedStorage:WaitForChild("Modules")
local events_folder = ReplicatedStorage:WaitForChild("Events")
local viewmodels_folder = ReplicatedStorage:WaitForChild("Viewmodels")

local current_camera = workspace.CurrentCamera

function vm_module.on_equip(tool: Tool)
    if not tool then
        return
    end

    local possible_vm = viewmodels_folder:FindFirstChild(tool.Name)

    if possible_vm then
        local real_vm = possible_vm:Clone()

        real_vm.Parent = vm_module.objects.Camera
        vm_module.functions.viewmodel_loop(real_vm)
    end
end

function vm_module.on_unequip()
    local vm = current_camera:FindFirstChildOfClass("Model")

    if vm then
        if vm:FindFirstChild("CameraBone") then
            vm:Destroy()
        else
            return
        end
    else
        return
    end
end

function vm_module.viewmodel_loop(vm)
    if not vm then
        return
    end

    RunService.RenderStepped:Connect(function()
        vm:PivotTo(current_camera.CFrame)
    end)
end

return vm_module

changed it around a bit

[spare quartz]

no. ocd and adhd

clearly superior

[snow oak]

am i flooding the chat

i think i am

[pallid loom]

no you're fine

we're just yapping

[spare quartz]

you should look into sending your messages as files but otherwise its fine

[pallid loom]

ocd

sir/ma'am have you looked at your computer.

[spare quartz]

yes i have actual ocd from my dad

[pallid loom]

there are cables everywhere.

[spare quartz]

not the "perfectionist" ocd thats on the interwebs

[spare quartz]

it works 👍

[pallid loom]

somehow...

just a matter of time until something blows up

is the HDD still hanging

[spare quartz]

yeah

[pallid loom]

incredible

[spare quartz]

these SATA cables are no joke

that specific hard drive isnt spinning rn though

so if it does fall it'll be fine

[spare quartz]

anyways few things

you should probably be early returning

[pallid loom]

why is it connected then

[spare quartz]

cause its in a RAID array

[pallid loom]

in what raid array

any raid array keeps all of the disks spinning does it not

[spare quartz]

no

theres no activity rn so why would it

[pallid loom]

oh btw i've just come to a great realisation

my server is under my 3d printer

my 3d printer does resonance calibration

[spare quartz]

also the hanging hard drive is frrom 2007 so ill be surprised if it dies by shock over age

[pallid loom]

which shakes the entire room

surprised my hdds are still running

[spare quartz]

i don't think modern hdds are really susceptible to noise like that

just physical shock

[pallid loom]

wasn't kidding

server is in the cage

https://new.cdn.teo0781.dev/u/240d6360-2113-4d4e-b57e-1c105a8924c9.png

you can see the faint white light of the gpu

[spare quartz]

how does that shake the room

[pallid loom]

you'd be surprised

[pallid loom]

https://youtu.be/gVpGbd0qB3Q?t=287

just click play it's at the correct timestamp

[spare quartz]

o wow

[pallid loom]

keep in mind it's basically on a lever since it's on the 2 stupid boxes

there's also horror stories of printers just fucking vibrating to hell while printing and falling off tables

so that's cool

[pallid loom]

i really dont know how any of this works in its current state

[spare quartz]

its very easy to see how it does work

[pallid loom]

i refuse to believe

the hdd

is that the next advancement in your setup

[spare quartz]

ohgyaygh

that was close

I nearly dropped my 2 liter of soda

[pallid loom]

HOLY SHIT

[spare quartz]

I should dust my pc soon

[pallid loom]

NOT A REAL AMERICAN

!!!!

LITERS?!?!?!?

[spare quartz]

yeah

2.1 quarts 🇺🇸

[pallid loom]

so wait

you have a usb wifi dongle

but use the pci one

[spare quartz]

no

i have a bluetooth dongle

my wifi dongle is in storage

[pallid loom]

i really do hope that this isn't somehow functional
https://new.cdn.teo0781.dev/u/e0f0d141-af24-4ed4-8bdf-fdfc8ab78089.png

[spare quartz]

LOL

im 99% sure

thats hooked up to my tablet

[pallid loom]

drawing tablet i assume

[spare quartz]

yeah

[pallid loom]

what fucking drawing tablet uses hdmi

i've only seen usb ones

[spare quartz]

its has a screen on it

so it uses USB-C & VGA/HDMI

greater color accuracy than my 144 hz monitor but is 60 hz

so i switch between the two for tasks

[pallid loom]

i mean i would hope its more color accurate

[spare quartz]

the difference is pretty small but yeah i'd hope so too

it costs $200 more than my 144 hz monitor

[pallid loom]

why is your windows in japanese btw

curiosity

[spare quartz]

testing something and got too lazy to change it back

additionally i get a cool ime

[pallid loom]

incredible

[spare quartz]

ばやちゃお。。。＜３

[pallid loom]

atp you should buy a framework laptop

i'd want to see in what fucked up way you could use it

[spare quartz]

thats expensive..

how much does one cost

[pallid loom]

depends on what you're going for

minimum of 1k usd

[spare quartz]

waow thats a lot

[pallid loom]

but you're a future defense contractor you have 20 billion in your pockets

[spare quartz]

id rather make a computer out of 250 pi picos

[pallid loom]

Lockheed martin junior

[pallid loom]

please never get into hardware you'll go insane

[spare quartz]

well

i did

[pallid loom]

how far

[spare quartz]

but you'll never guess what i did to my pico after running ada on it

i fried it ☹️

[pallid loom]

are you talking about the fucking aca-

expected

How'd you do it

5v to gnd or something else

[spare quartz]

fucked around with a metal probe on the pins

it was shockingly resillient

[pallid loom]

never really used a pi pico

[spare quartz]

they're cool but mega limited

[pallid loom]

or get this

[spare quartz]

haven't gotten my pcbs manufactured yet 💔

[pallid loom]

20 rpi 5 cluster

[spare quartz]

thats so much

[pallid loom]

4 GB ram versions exist

[spare quartz]

what about 256 mb

[pallid loom]

btw move to #1267583072728121384

[lavish dove]

no way

[lavish dove]

thats more than my computer 😭

[lavish dove]

seabios

qemu's bios is ligit just seabios

[spare quartz]

yeah

[spare quartz]

y synced

just need to fix scrolling

[lavish dove]

https://docs.github.com/en/copilot/managing-copilot/managing-copilot-as-an-individual-subscriber/about-github-copilot-free
its free again

[lavish dove]

the pieceo f shit

[spare quartz]

accidentally maed this cool fx

[proud creek]

LOL

[spare quartz]

https://youtu.be/3-W95H5_lX0

[lavish dove]

omg it looks like a uh

terminal from a game

[spare quartz]

https://wetdry.world/deck/@[email protected]/114150084779943084

post made by a rustlet i decided to send ehre

of course, Ada wins, always has won, but yeah

[proud creek]

❌

[spare quartz]

❓

[proud creek]

https://cdn.discordapp.com/attachments/1335762058934747207/1344049487135510558/ezgif-4-22345e2637-1-1.gif

[spare quartz]

go back to coding tz.

https://imgur.com/2M78rBi