programming

[mighty thorn]

That is my name

And it always has been

[sage crag]

making new ones

[tender river]

i just got confused because of arena instead of Arena

[olive sable]

[sage crag]

er

i cant type

that's ok

[tender river]

it is

[opaque wharf]

Ahhh, Vedal in 'nam

Since init is abbreviated, crea is valid design choice too

[sage crag]

n

[olive sable]

does nam look like that?

jsut grass evertywhere?

[opaque wharf]

That looks like rice field

[olive sable]

https://tenor.com/view/batad-rice-terraces-banaue-ifugao-gif-5311368

huh

idk what i thought rice came from but i wasnt expecting grass

[tender river]

O_CREAT

21 bytes saved across the linux kernel source code

[opaque wharf]

Saving 1 byte for reduced readability

[tender river]

increased readability*

shorter = easier to read

[opaque wharf]

Chay be out there trying to make every language using glyph a la APL

[humble shell]

[opaque wharf]

Speaking of which, Sam, since you're programming a GPU, you'd love APL

[tender river]

sam would never

if he cant handle same line braces he cant handle apl

[sage crag]

i thought they already did this

oh they do

[sage crag]

right

[tender river]

22*

i cant count

[sage crag]

awa

[tender river]

awa

[sage crag]

i caused a compiler crash

[opaque wharf]

awa awa awawa awawa awa awawawawawawawa awa awawa awawa awa awa awa awawa awawawa awa awawa awawa  awa awa awa awawawa awawa awa awawa awawa awa awa awa awawawa awa awa awa awawa awawa awa awa awa awawa awawawa awa awawa awawa awa awa awa awawa awa awa awa awa awawa awawa awa awa awa awawa awawa awa awa awawa awawa awa awa awa awawa awa awawa awa awawa awawa awa awa awa awawa awa awa awa awa awawa awawa awa awawawa awawa awa awa awa awawa awawa awa awa awa awa awawawawa awa awawa awawa awa awa awa awa awawawa awa awa awawa awawa awa awa awa awa awawawa awa awa awawa awawa awa awa awa awa awawa awawa awa awawa awawa awa awa awa awa awawa awa awa awawa awa awawa awawawawawa awa awawa awawa awa awa awa awa awa awa awa awa awa awawa awawa awa awa awa awawa awawa awa ~wa awa awa awa awa awawawawawa awa awawawa awa  awa awa awawa awa awa awawawa awa awa awa awawa awa awawa awa awa awa awa awawawa awa awa awa awa awa awa awa awa awa awawa awa awawa awawa awa awa awa awa awa awa awawa awa awawawa awa awa awa awa awawa awawawa awa awa ~wa awa awawa awa ~wawawawawa ~wa awa awa awawa awawawawawa

You can check that it runs
https://temptempai.github.io/AWA5.0/

[humble shell]

Reminds me of brainf**k

[sage crag]

wait i forgot its a thing

did i steal that from there

[tender river]

you did

[sage crag]

im a thief

[tender river]

intellectual property infringement

[sage crag]

works

not sure i like using a Layout struct

its useful for checking valid layouts but also validation is for nerds

[tender river]

thats what high level wrappers are for

you can rename it to raw_alloc or whatever

[sage crag]

what i mean is, i could just be passing length and alignment as args

rather than in a struct

meaningfully there is no difference

[tender river]

sure, this can be used for stuff like multiplying layouts by a size though

[humble shell]

What is that, Ruby? Rust?

[tender river]

hblang

[sage crag]

[sage crag]

its hblang

wait wrong emoji

[tender river]

https://git.ablecorp.eu/mlokis/hblang

[sage crag]

[sage crag]

true

didnt think about it

[tender river]

its so unfaaaair
when you want to make functions more flexible, you have to use function pointers rather than know them at compile time
when you want to make struct layouts more flexible, you have to use structs with changing sizes and cant allocate a fixed amount of memory for them
why is language design so hard even when you have ways to make the type system support any feature you throw at it

[hoary lion]

actually going to crash out

[sage crag]

gwa

[tender river]

this is why most languages just use gcs

[sage crag]

most languages should go and take a walk

[opaque wharf]

Gwa is a slang language for referring to yourself in ID

[sage crag]

interseting

interesting

[tender river]

type interference

is type a wave or a particle

[sage crag]

depends

[opaque wharf]

Yes

[sage crag]

when you observe a type its a particle

but when you arent observing a type it becomes dubious

[sage crag]

the cost for this dynamic dispatch is apparently 35 bytes

but if you used multiple allocators in one function then the cost would be negative

[opaque wharf]

Hmmm, meatballs with noodles or meatballs with rice

Decisions decisions

[rigid snow]

> new Date("2025")
// 2025-01-01T00:00:00.000Z
> new Date("2025-")
// 2024-12-31T21:00:00.000Z

[sage crag]

i agree

your fault for using javascript

[rigid snow]

oh hey i found a fix! ```js

new Date("2025- utc")
// 2025-01-01T00:00:00.000Z

[rare bramble]

🐥 🚬 random JavaScript BS that no sane person could ever have come up with

[sage crag]

parseInt(0.00005)

[humble shell]

Such an awful language, just to make a web browser scriptable

[sage crag]

js converts 0.00005 to a string, which is represented as 5e-5, then converts the first string of digits to an integer

the output is 5

[rigid snow]

it shouldn't be though no

[olive sable]

it is monday

[tender river]

they say that you should be lenient with what you accept to comply with noncomformant implementations but that just allows them to get away with more noncompliance

[rigid snow]

because of the dot

> parseInt("0.00005")
0
> parseInt(0.00005)
0

[sage crag]

add more zeroes

[rigid snow]

yeah

[olive sable]

that doesnt look like an int

[sage crag]

i probably underestimated the number needed

[rigid snow]

you underestimated by two zeroes exactly

[sage crag]

i see

[tender river]

it makes sense but its so cursed

[rigid snow]

but if you just used typescript you would know that parseInt only takes a string

hence the coercion

[sage crag]

converting floating point to string is not coercion of types, its coercion of the truth

converting fp to high quality strings is such a complicated process

[rigid snow]

5e-7 is a very high quality string

[sage crag]

yeh

[rigid snow]

made by the bestest of the best spec developers

[tender river]

old erf had coercions

new one will not

[sage crag]

erf now usable

[sage crag]

its not easy to convert to a string in this form

[sage crag]

https://github.com/ulfjack/ryu

[tender river]

it only had something like i8 to i16 because i didn't realize that just because the type inference algo i used supports subtyping it doesnt mean it supports that kind of subtyping

now it wont

[sage crag]

ah, just upcasting?

upcasting is fine

[tender river]

its still coercion

[sage crag]

acceptable coercion

hblang has the more dubious bool to int upcast

[rigid snow]

function glueless(n) {
    const e = Math.floor(Math.log10(n));
    return `${n / (10 ** e)}e${e}`;
}

[tender river]

[sage crag]

read ryu src code

[sage crag]

most js engines will use this or grisu3 most likely

[rigid snow]

ding ding they do

can't recall which ones

[sage crag]

rust uses ryu as well

[rigid snow]

but i looked into js runtimes the other day

[sage crag]

as such i was going to implement ryu myself for hblang but

[tender river]

functions dont use fixed stack frames with preallocated slots for every variable they add and remove data at will so needy

i cant just use plain old structs for stack frames

[sage crag]

✅

[tender river]

theres also registers which make everything even funnier

[sage crag]

purple

[tender river]

purple

[humble shell]

Registers aren't purple, no

[rigid snow]

what color are they then huh

[humble shell]

Whatever colour sand they had on that day

[rigid snow]

lies made up by big sand

propaganda

[tender river]

big sand is just pebbles

[rigid snow]

the smart watch company??/

[humble shell]

Wow I actually forgot about that since everyone I know has either an iWatch or a Fitbit

[rigid snow]

it all checks out. they never had a purple variant of any watch

[rigid snow]

i have a mechanical watch like a normal person

smart watches are bloat

quartz oscillators are bloat too

you will wind your watch up yourself and you will like it

[humble shell]

I just love my mechanical watch.
Kenneth Cole I bought many years ago.

I don't have image posting perms yet otherwise I'd show it

[tender river]

i was wondering if there were nuclear powered watches

apparently some prototypes do exist

ever since 1970s

but no commercial products

i wonder why

[humble shell]

I dunno, maybe cancer?

[tender river]

i dont think a nuclear battery is gonna be worse than something like your smartphone

i mean this is not medical advice or anything

big news

https://www.nytimes.com/1956/09/02/archives/now-the-atomic-wrist-watch-in-the-works-is-a-timekeeper-run-by-a.html

[humble shell]

1956, so just yesterday

[olive sable]

in my search for grass ive created wifi

[hoary lion]

[amber fractal]

Hey bro why does this spot have so much wifi?
The signal strength the spot has:

[olive sable]

5G? thats baby shit im in triple digits

the signal strenght is so strong you can microwave a burito with it

[amber fractal]

Wrong way, higher bands are closer to gamma rays

[humble shell]

So, cancer causing

[amber fractal]

Atomize the individual on the spot mhm

[humble shell]

Or "Chernobyl 2.0"

[olive sable]

[fast pagoda]

i think im boutta yolo and replace my windows install with a linux distro on desktop as well
it's been great on my laptop

im thinking cachyOS
or maybe nobara
arch based vs fedora based

[humble shell]

My desktop and laptop both run Debian

[fast pagoda]

i have kubuntu 25 on my laptop rn

i'm a fan

it's been pretty nice

also been kinda eying anduinOS since i keep seeing that pop up

[olive sable]

damn i had to think for a sec what sqrt(0) is

im off my game

[midnight sigil]

no it has to be ionizing radiation to destroy our cells, which is "high speed", and "high energy", clearly not the case of even 10G wifi

tbh with higher frequencies we might have to just put a router at every room

[olive sable]

im ok with that

give me 100 gigabit wifi in return

[humble shell]

It's obscene what all the providers in my area charge just for 100mbit.

[midnight sigil]

get fibres

easy 1000mbps

[humble shell]

Bell Canada hasn't run fibre near my street yet

[olive sable]

same here

[midnight sigil]

[olive sable]

float y = floor(sqrt(id)) this feels like such a waste

sqrt is such an expensive function and im just flooring it

[midnight sigil]

make sqrt on your own then

[olive sable]

how?

[midnight sigil]

https://stackoverflow.com/questions/4930307/fastest-way-to-get-the-integer-part-of-sqrtn

but well tbh

cpus and instruction sets are optimized for sqrt calculations

[olive sable]

cpu's

[midnight sigil]

it's not that expensive

[olive sable]

this is shadercode

[midnight sigil]

oh

[olive sable]

this litle sqrt will be ran 250000 times for every single frame

tbh from the replies im seeing there i doubt any manual method would be fastewr with simd and stuff existing

[midnight sigil]

meh you gotta trust the power of John Carmack

[sage crag]

do you need it as a float

[tender river]

fast inverse square root is an algorithm for a different age

[olive sable]

nope

int is fine

[sage crag]

is id and int?

[olive sable]

it was

[tender river]

it can easily become an int through casting if it isnt

[olive sable]

float id = float(gl_InstanceID);

[sage crag]

bruh

[tender river]

that could technically change the result though so fair question i guess

anyway on a gpu i dont think any other solution is going to be faster

[fast pagoda]

i wish i had fiber at my apt

cable not too bad all things considered

just dont look at my upload

[olive sable]

brother you're complaining about 2800mbps

[fast pagoda]

i just want symmetrical

[olive sable]

you can trade me for my symetrical 100 then

[gritty dust]

they have for me

[fast pagoda]

i complain about upload and then it's like

still 360mbps

[gritty dust]

rip dude

[fast pagoda]

https://tenor.com/view/parks-and-rec-gif-4978282

[tender river]

i've had this before

[fast pagoda]

someone posted their connection presumably from a box they were ssh'd into and it was blasting 10+ Gbps

[patent walrus]

this is funny but makes sense if someone were to like, go to the page, disconnect their internet and try to run the test lol

[tender river]

i've had this when simply trying to use their page

[fast pagoda]

sounds like my employees having "connectivity problems" when they're supposed to be working

not that i care if they get their stuff done but it's crazy how it's always the same people with wild & wacky issues

[sage crag]

its probably my dns filter

but its funny so im not going to fix it

[olive sable]

ye i have no clue if im doing this correctly

it looks correct i think

[midnight sigil]

I guess google colab runtimes aren't that powerful

[hoary lion]

colab's true power comes from drive

I love the imbalance

[sage crag]

@tender river

WhatTheDynObj := struct {
  .func: ?^fn(^void): uint;
  .obj: ^void;

  ...
}

Obj := struct {
  ...
}

main := fn(): uint {
  obj := Obj.(...)
  dyn_obj := WhatTheDynObj.new(&obj)
  if func := dyn_obj.func {
    return func(dyn_obj.obj)
  }
  return 1
}

cursed nullable function

[olive sable]

[tender river]

cursed? looks perfectly fine

[sage crag]

evil function that doesnt exist if you look at it wrong

[tender river]

i'm continuing work on the lsp, currently sorting out the TODOs in rpc.hb

and then i'll move on towards serialization

[sage crag]

fnc_raw := u8.['e', 'r', 'm']
fnc: ^fn(): void = @bit_cast(&fnc_raw)

$cursed := fn(): void fnc()

[tender river]

inline assembly is finally possible

actually you could write a DSL that generates assembly at compile time

[sage crag]

you could, using enums

Operand := union { .reg: Register; ... }

OpCode := enum(uint) {
  .mov := whatever_mov_is;
}

Instr := struct {
  .opcode: OpCode;
  .operands: []Operand;
}

asm := Instr.[.(.mov, .[.rax, .rbp])]

:mhm: this would totally not work

probably better to do it in the type system

[tender river]

variable length instructions

[sage crag]

yeah that's the problem

[tender river]

you have to use a serializer

or alternatively just use a linked list

[sage crag]

not more comptime linked lists

[tender river]

taking pointers at compile time and serializing them to a byte array will not relocate properly

so no absolute jump

relative might work

[tender river]

at some point you realize, everything that is strictly ordered is a linked list

[hoary lion]

update on my training script:

man I hate torch

it is going to take few buisness days to get the data loaded

at least i escaped from complete halt tho

[stark needle]

Hi

[twin meadow]

Hello. Just a quick question. Do any of you use macs or stuff like that and know of a good programming software on the operating system

[tender river]

probably just use vscode

you might also want to install brew idk

[olive sable]

goodmorning shadow

rate this shader out of 10

i give it a 4 personally

[fast pagoda]

this one is going slower than i would've expected

[olive sable]

@gritty dust

i need more low quality audio

[gritty dust]

OMG I love it

its so amazing

[olive sable]

im running out of ideas

[gritty dust]

Hold up lemme give you sm

you could lower the quality or sm idk

or like the interstellar sound track lol

[amber fractal]

I was expecting the meme version dang it

[gritty dust]

oh lol

[safe path]

tf

[opaque wharf]

Sorting the todo in alphabetical order that is

[opaque wharf]

Evil never got a chance to colab with vedal

[amber fractal]

fast.com is by netflix so makes sense why it'd be blocked

[sage crag]

asm.{x86_64: .{assemble_fn, mov, syscall, eax, ebx, const}} := lily.asm

exit := assemble_fn(^fn(): never, type.[
    mov(eax, const(os.lib.arch.Syscall, .exit)),
    mov(ebx, const(u32, 1)),
    syscall,
][..])

main := fn(): i32 {
    exit()
}

working on cursed asm dsl for hblang

[fast pagoda]

dont open several 7zips on several gigabytes of directories at once

[opaque wharf]

Sometimes I wonder what goes through everyone minds here at any point in time

[fast pagoda]

the main thing going throguh my mind is 7zip

[opaque wharf]

Also is that ventoy has 100+GB?

It's data right? ....right?

[fast pagoda]

one 7zip has been defeated

the ventoy is 105gb or something because the drive is a 128gb drive

ventoy is just a thing that lets you put bootable .isos on it and you can boot from whatever one

every time i get a usb drive i get like 128 or 256 gb since theyre basically free

https://www.ventoy.net/en/index.html

ventoy

[opaque wharf]

I know what ventoy is. I just wished you didn't actually reserve THAT much space for the ISO

[fast pagoda]

oh lol

yeah i mean im gonna use it like a couple times and then wipe it again to use for something else

so i just formatted the whole size of the drive

oh hard drives

disk 0 and 1 are hdds and are absolutely pegged out reading

sending stuff to disk 3

which is not even at 5% from it

[sage crag]

erf

writing an assembler as a comptime macro is harder than expected

[fast pagoda]

i dunno it sounds like something i'd expect to be hard

lol

[sage crag]

would be fairly trivial if i was targeting a risc architecture

or even x86

but nooo, x86_64 has to be special

[fast pagoda]

we need vplzcntd ok we just do

[sage crag]

i will not be implementing this

[fast pagoda]

[sage crag]

but in theory instructions are generic so

[fast pagoda]

muh avx512

[sage crag]

mov := fn($dst: type, $src: type): type $if @has_decl(src, "imm") {
    return struct {
        $code: u8 = 0xB8 + dst.reg
        $ops := type.[src]
    }
} else {
    @error("todo: non-immediate ", mov)
}

syscall := struct {
    $code: u16 = 0x0F05
    $ops := type.[]
}

well, depending on the instruction

[opaque wharf]

What theory said: infinity
What really happened: quantum mechanics

You will need to make a conjecture like shiro

[fast pagoda]

discuss the 2d nature of shadows

[sage crag]

mhm

[opaque wharf]

Shadows can be 3d if the object is higher than 3d

[sage crag]

i bet not a single person here knows about modrm

x86_64 is the worst

[opaque wharf]

I only know about modm.io

[sage crag]

https://en.wikipedia.org/wiki/ModR/M

[noble zodiac]

don't insult me

[sage crag]

even making an assembler for x86 is a headache

[sage crag]

true, you are like a professor or something

i dont remember

[noble zodiac]

I'm not a professor, but I have written a disassembler

[sage crag]

same thing

sorry i havent slept in 22 hours

[opaque wharf]

So what about that one guy that make movfuscator

[sage crag]

im doing great btw

[fast pagoda]

i'll have you know i have actually recently been watching multiple videos on the topic of instruction sets recently and
i dont recall the term even so

[sage crag]

i dont think about witches

[opaque wharf]

Yeah, this is where theory meet practical application will slap you

[fast pagoda]

all i know is every time i get to branch prediction i return to monkey

[sage crag]

the modrm encodes the operands in x86

its encoding is like rm src dest or something

rm is for the type of operands you have, so 11 is direct register

i dont remember the other ones

and then, for direct register, the src and dest are both three bits which encode a register

[opaque wharf]

Well, branch prediction is just one part of a whole set of optimization. You will meet out of order execution, pipelining (I think it was called?), caching, and many more

[fast pagoda]

reading on it now and it seems like a good way to ruin someone's branch predicting since the addressing mode OR NOT it could be a register for fun

[sage crag]

i.e. rax is 000, rdi is 111

[fast pagoda]

i understand what it does and what its role is in general i just still cannot put the algorithm to the rock

[sage crag]

but you may notice we only have rax to rdi here, so we are missing all of r8 to r15

r8 to r15 come from the REX prefix for the instruction

[fast pagoda]

right then we need the mod

[sage crag]

this is not considering r8b-r15b, r8w-r15w, r8d-r15d, which are the extended registers for 8, 16, and 32 bit respectively

but i dont know how often they are used

probably also set with REX prefix

[sage crag]

typically i think its REX OPCODE MODR/M

i have not considered what happens if you have an extended size opcode

i do not want to think about it

[midnight sigil]

is chat talking about asm now

[sage crag]

mostly machine code

[midnight sigil]

ahh ok

[sage crag]

this is why no one likes x86

switch to risc today

its so much nicer

[fast pagoda]

i am reading the 64 bit section on the modr/m page like

[midnight sigil]

arm vs risc

[sage crag]

wh

[sage crag]

same

typically is based because each instruction is 1 byte, and then the operands sizes are defined in the instruction spec for each instruction

and things like mov wont have a million overloads

it would be mov and movi and movm etc

[fast pagoda]

it does seem like it allows for very complex addressing to be done in like 4 bytes

[sage crag]

i do not want to think about all the things it can do

i really should have written an hbvm assembler first

[fast pagoda]

mov r8, [r13 + 0x20] -> 4C 8B 45 20

that's a lot into a little to me

[sage crag]

yeah that's a

[opaque wharf]

And she fainted

[sage crag]

not evne considering things like mov rax, QWORD PTR [whatever the frick address]

which are memops

[fast pagoda]

update: hard drives are very upset

[sage crag]

i dont remember if that is RIP-relative addressing or absolute addressing or what

feck x86

who designed this garbage

intel forum

[fast pagoda]

now do itanium

[sage crag]

i will destroy you intel

no

when will there be an hbvm cpu

bwa

i wrote a vm impl for hbvm in desmos

[opaque wharf]

What will the ISA look like?

[sage crag]

https://git.ablecorp.eu/holey-bytes/holey-bytes/src/branch/trunk/spec.md

[fast pagoda]

this made me curious about itanium

[opaque wharf]

Interesting

[sage crag]

mhm

NOTICE HOW

[fast pagoda]

and i have found that despite having 300~ish opcodes instead of the like 1500+ of x86-64, they are 128 bit bundled encodes so they come as 3 41 bit slots and then there's 32 5 bit templates to combine with

[sage crag]

the instruction encoding is just one 2 line header

peace to risc ISAs

[sage crag]

hbvm has 119 opcodes, each one is a u8 with variable length operands

[fast pagoda]

every instruction (almost) has a 1bit predicate + 5 different nop types so every one can end up a no-op'

[sage crag]

variable operand count is a bit slow compared to storing the operand count with the opcode

but who cares its a cpu

[fast pagoda]

i for one hate when my cpu is fast

[sage crag]

super instructions for you

[fast pagoda]

lord have mercy if it's efficient too

TX is the end of the line i knew it

[sage crag]

nop efficiency is dependent on your implementation

i have put core::hint::spin_loop into the ableos one

hbvm does actually have two forms of simd

i will tell you these were not fun to write in desmos

[olive sable]

imma sleep

gn

[fast pagoda]

so just either doing registers or bytes

oops

wron message to reply to but yeah

[sage crag]

hi

what

im too bleary to understand

[fast pagoda]

oh i was just saying the 2 types of simd

[tender river]

having enough registers to warrant block register copy is pretty cursed honestly

[sage crag]

256 registers is enough registers for register purposes

also enough registers to do recursion without a stack

not for long though

[noble zodiac]

call me crazy but the x64 instruction set reference was a nice read

[opaque wharf]

Does iOS support creating local http server that is callable from the phone browser?

This is cursed I know

Oh, it seems like they could

Welp, some shenanigans it is

[dire trout]

help
how do i get useless groupmates to finish a simple python system within 12 hours

[opaque wharf]

You really need to sleep lmao

[dire trout]

they AI'ed their entire code and its fucked up

[sage crag]

no

[rare jewel]

Can i make a .bat file download something off the internet?

[amber fractal]

Waste of energy groupmates, just code the entire thing and show it working. Then say "This is your goal, I'll just not submit this if you can't cooperate."

[opaque wharf]

Yes

[rare jewel]

how

[dire trout]

i cant do it myself
i mean i did it myself already but there's the group marks

everyone hasto do something

[opaque wharf]

But then he doesn't get the score either

[dire trout]

[opaque wharf]

It WILL depend on the teacher how this case is handled

One can handle it by just accepting that one person is uncooperative

Another can say it is the group failure to cooperate

[dire trout]

i jokingly said id do it myself and she just glared at me

[amber fractal]

I mean the goal was to gaslight them with this method.

[dire trout]

yeah i did that

i think they copied my code too

without understanding it

[dire trout]

(part of it)

[amber fractal]

Ditto on it depends on the teacher on how it is handled.

[dire trout]

like atp im just asking them to list out the function details USING COMMENTS (you can see where the AI'ed parts are in the comments, too 😭)

[amber fractal]

You could abuse any missing assignment policy to submit working code after the fact, let them fail. If your in the worst universe then I don't think you have a choice but to let them leech.

[dire trout]

we need to present this

live

[amber fractal]

Man I hate group projects

[opaque wharf]

Have that person do the opening, transition, and closing

[dire trout]

i already did that

also, 3 people

[amber fractal]

:AINTNEURWAY:

the entire group is incompetent

[opaque wharf]

At that point I'd honestly ask the teacher and accept my fate however the teacher handles it\

[dire trout]

it's over

my 4.0 perfect grade

all gone

fuck

[opaque wharf]

I don't think GPA matters that much tbh

[dire trout]

i wanted at least one module with a perfectr grade

😭

[amber fractal]

I'd say do your best and whatever happens happens, but do argue your case.

[opaque wharf]

Ye, this. Life happens

[dire trout]

pain

[opaque wharf]

It is what it is

[amber fractal]

I'll say do not give them ammo to dig themselves out, fuels this better.

[opaque wharf]

[amber fractal]

If they have a brain they'll realize only one person did anything and the others are bsing.

[amber fractal]

Your goal is to make that part real, misinformation and all

[dire trout]

actually ykw

"individual QnA"

[amber fractal]

boys we have an out

[dire trout]

i definitely can't use Class though even though its useful

but

most likely ill give something that's covered in the lesson notes (although i really want to use Python pass and return...)

bleh formatting error

[amber fractal]

We just need oofio to get a perfect score on the individual QnA while their mates are all in the 0-4 range, ideally 0 for the lols.

Playing the long game now

[dire trout]

true

presentation's on weds

will let yall know xd

[potent birch]

im slowly(lazily) learning

i just finished trying out how to change scenes, now im lazy again no energy

[lapis wraith]

Good luck with godot

I think I know what's going on with gary... I have more than enough clues to think he's incapable of placing anything on the negative positions.
The schema is fully correct and allows them. The game understand the negative numbers but gary just... doesn't know how to input a negative number to the schema?

[potent birch]

Who's gary

[lapis wraith]

A program that allows to run a local AI to test neuro SDK integrations.
Is also the name given to that AI.

is this: https://github.com/Govorunb/gary

[potent birch]

Ooo a small ai tester

[trim valve]

my only guess is something is casting the integer to an unsigned integer somewhere in the pipeline

[molten island]

so monads in haskell basically and really oversimpliedly are containers that makes a hell lot of sence

[distant blaze]

monads are monoids in the category of endofunctors

[molten island]

i've used cool spell - "oversimplification"

[lapis wraith]

This might be "Guidance". Gary uses it so the AI doesn't write the whole JSON. But It might be confused somewhere.
I'm seeing it's code and says it supports negative numbers but... who knows. Time to dig deeper.

[tender river]

they're more like things you can infinitely chain operations on

calling them containers doesnt do it justice, even the list monad is more like rust iterators than rust vectors

[stark needle]

Hi humans

[tender river]

hello whoever you are

[jagged turtle]

how big is your model

[lapis wraith]

The file name says it's 8B (Whatever that means)

[tender river]

[jagged turtle]

I uh

am pretty sure that's a bit too small

it also depends on your hardware specs as well

[opaque sigil]

8 bytes is indeed prettty small

[tender river]

thats 18446744073709551616 different values

its pretty big

[lapis wraith]

I'm pretty sure it doesn't refer to 8 bytes, right?
Anyway, I only have a 3060 with 12GB of VRAM.
Maybe i'll have to try to find a bigger LLM somewhere

[jagged turtle]

Uhhh I don't know exact specs but I think that should be, like, sorta decent?

[opaque wharf]

8 Billions for a neuro tester is good enough IMHO

Because it is not neuro

[lapis wraith]

Yeah... and I don't need a Neuro for a play tester. In fact, the dummer the better. It only means higher chance of neuro actually getting a chance to understand the prompts

[opaque wharf]

Neuro: "Because it's funny"

[lapis wraith]

Then that's not my problem. She can act funny as long as she understands

[opaque wharf]

Reading the docs about neuro SDK where it is said that neuro may give command that doesn't exists, makes me believe that neuro indeed could make a problem with her AI

[lapis wraith]

Oh yeah. She can. Right now I have no checkers on my schema and moving to the position 10.000.000 in one go is a valid move.

[molten island]

yep i saw and read how they works and practised a lil' bit

it's fucking awesome

[lapis wraith]

YES. I found the problem. Gary has a bug in the "Follow schema" option.
That dummy option simply eats negative numbers for breakfast and converts them to "0"

[jagged turtle]

PR time!!!

[jagged turtle]

literally how bro

is there like

no basic checks at all with her server

[opaque wharf]

Nah, that's just how old school LLM is like

I don't know the exact date of neuro and LLM advancement, but I do know that Neuro is quite old (in relation to LLM research progress)

[tender river]

wrote a json serializer for hblang

as expected, its very trivial compared to deserialization

only 135 lines of code

32 of those are just strings

i'm sure lsp won't require any more plumbing and i will soon actually be able to work on the lsp itself right

[opaque wharf]

Did you not sleep since yesterday?

[tender river]

[opaque wharf]

Man, ablecorp needs to review its internal policy

[tender river]

its fine i only woke up like 23 hours ago

[jagged turtle]

no but like basic checks before sending a packet

[opaque wharf]

Mhm, you and konii are the same person alright

[opaque wharf]

Well, it is also why we got neuro trying to cause problem with her AI lol

And I think the SDK is developed later after the fact. So neuro must've been at it's very early stage and the system got too complex to plug the check without adding latency

And we know how autistic vedal is towards her latency

[tender river]

not the same

[desert hawk]

https://media.discordapp.net/stickers/1247260299480924201.gif?size=160&name=ChunClap

[desert hawk]

Almost didn’t recognize u

😭

[olive sable]

Goodmorning

[opaque sigil]

[opaque wharf]

"Morning" as I check the Belgium time is ~15.00

[desert hawk]

[stark needle]

Hi

[desert hawk]

[ruby palm]

https://fxtwitter.com/reachartwork/status/1947041750125527137

@ reachartwork on tumblr & bsky (@reachartwork)

the reason why ai will never replace dedicated coders is because this is what happens when you let non-smes use ai

Quoting Jason ✨👾SaaStr.Ai✨ Lemkin (@jasonlk)
︀
.@Replit goes rogue during a code freeze and shutdown and deletes our entire database

**💬 29 🔁 458 ❤️ 7.1K 👁️ 427.8K **

FxTwitter · Jul 20, 2025, 9:11 PM

[opaque wharf]

Lol, lmao even

[rare bramble]

who has AI executing random commands on a database without any supervision?

[dire trout]

           
try:
                category_code = [category['category_code'] for category in self.menu.values()]
                choice = input("\nEnter category number (0 to exit): ").upper()
                if choice == '0':
                    break
                if choice not in category_code:
                    print("Invalid category code.")
                    continue
            except TypeError:
                print("Please enter a valid code.")
                continue

            category = list(self.menu.keys())[category_code.index(choice)]
            category_data = self.menu[category]

i feel like im being a idiot somewhere

[dire trout]

scary

[nocturne olive]

I think it may be missing a thing called a "transformer" or "power supply"

[split osprey]

hey Guys

any tips and tricks to fix the disconnect between action state models and their training data mouse behavior

a lot of Gaming is very dependent on Camera movement.. AI can replicate Discrete keypresses Easy but there is a huge disconnect between player control of camera and ai control

[worldly plank]

Try adjusting indentation levels?

[dire trout]

no it works

it just

it just feels stupid

like it could be optimized better

[opaque sigil]

i don't think there's anything that could throw a TypeError unless i'm missing something?

[tiny mango]

programming is just bruising your ego over and over

[opaque wharf]

What are you optimizing for

[opaque sigil]

trying to be smart will only hurt you half the time

[opaque wharf]

This

If it works then why not. Unless there is a need to do so

[trim valve]

if category is not a dict it might

>>> 1['a']
<stdin>:1: SyntaxWarning: 'int' object is not subscriptable; perhaps you missed a comma?
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: 'int' object is not subscriptable

[opaque sigil]

oh true

[opaque wharf]

Unless you mean an extensible code with good performance for some specified amount of workload

[stark needle]

[olive sable]

i finished disc 1 of resident evil 4

[opaque sigil]

[warped narwhal]

lmao, this reminds me about how you can do this in c and c++:

int main(void) {
  unsigned int my_arr[512];
  unsigned int i;
  for (i = 0; i < 512; ++i) my_arr[i] = i;

  unsigned int value = 312[my_arr];
  printf("%x", value);

  return 0;
}

[trim valve]

hate

like I think I get why it works in this case but still hate

[opaque sigil]

gotta love pointer arithmetic

[warped narwhal]

https://godbolt.org/z/b4a78s91x here it is in action

[worldly plank]

Is there a good reason for doing it this way?

[warped narwhal]

no

[opaque sigil]

no

it's just an effect of how the standard defines array subscription

[warped narwhal]

it only works due to how x86 handles array indexing with it's addressing mode

[opaque wharf]

Uhh no

It is due to how the language spec work

[worldly plank]

Is there a bad reason, a something-between-good-and-bad reason, a something-not-between-good-and-bad reason, or any form of "benefit"/cool-stuff-you-normally-couldn't-do to doing it this way?

[warped narwhal]

There isn't really any difference between my_arr[150] and 150[my_arr] they compile to my_arr + 150 and 150 + my_arr respectively, which does the same thing.

[opaque wharf]

There is one. When you enter the C obfucation coding competition

The IOCCC

[opaque sigil]

[warped narwhal]

the movfuscator is the best way to obfuscate your binary

[trim valve]

I assume it implodes if you have a array of something bigger / smaller than an int

[warped narwhal]

¯_(ツ)_/¯

lets find out

[opaque sigil]

it shouldn't matter

[trim valve]

it should

I think?

[worldly plank]

Mine imploded, I think.

[trim valve]

I'm not sure actually hm

[opaque wharf]

It should not. Because the effect is just a logical conclusion of the [] operator on a pointer

And pointer arithmetic

[opaque sigil]

it works fine, just checked

[opaque wharf]

If it does not, I would be worried for the compiler ngl

[trim valve]

oh wait I know why I got confused

I think?

wait no

[nocturne olive]

Whuh

[trim valve]

[worldly plank]

How do you do the code thing?

[nocturne olive]

In C you can index a number at an array pointer??

[opaque sigil]

yes

that is how arrays work

[opaque wharf]

Not index a number at array pointer technically

[nocturne olive]

I did not know of that backwards case

[opaque sigil]

it's because A[B] = *((A)+(B)) you can do it both ways

[gritty dust]

@olive sable u need help with game engine thing at all? bc I got free time and stuff

[nocturne olive]

That is actually crazy

[opaque wharf]

Call it syntactic sugar if you want

[trim valve]

I think I just got confused because surely if sizeof A != sizeof B something feels like it shouldn't work

but I guess it makes sense?

iunno

[opaque sigil]

it works out cause there's only one actual pointer involved so it can just use the size of that pointer's type for the addition

[opaque wharf]

Pointer size is always the same for any type no?

[olive sable]

not really, its mostly just doing stuff in blender rn

[worldly plank]

Yes

[opaque sigil]

the size of whatever type it's pointing to

[opaque wharf]

Making juice?

[worldly plank]

But they still need a type to know how much they should increment (I think?)

[opaque sigil]

but yeah the pointer itself is usually 8 bytes

[warped narwhal]

not always:
https://wiki.osdev.org/X86-64#Data_Types

[opaque sigil]

i said usually for a reason

[opaque wharf]

Ohh yeah right

[trim valve]

what

[worldly plank]

I guess it really depends on the hardware?

Or something address space or something

[warped narwhal]

why not use alignas instead of packed with padding?

[trim valve]

I am not a C developer

so I just guessed :3

[warped narwhal]

you know about packed but not aligned

[worldly plank]

What does the __attribute__((packed)) part do?

[opaque sigil]

removes padding

[warped narwhal]

stops gcc from adding padding for optimisation (e.g. placing values on 16 byte boundaries)

[trim valve]

its the way I've seen padding done before in C

[opaque wharf]

So the thing is, accessing memory are not done 1 byte at a time

And it can be any n bytes. And if your struct is saved in memory happens to be larger than those n, it may need multiple fetch from memory

Why does computer have to be so complex

[worldly plank]

What "what"?

[trim valve]

how on earth did I get 64

oh I'm stupid

its formatting it as hex

bruhge

that would explain why it seemingly worked on small numbers but not big ones 😭

[opaque sigil]

i love reading memory in 64 byte chunks

[nocturne olive]

Because fast

[fast pagoda]

alright linux masters
installed cachyOS last night and it's been working fine except that i had 2 hdds that used to be mounted in a software raid on windows
they were empty so i figured i'd just reformat them in the install process and that'd be that
well they have been preventing the system from booting because it kept failing to mount

got past that and now i have 2 hdds with one(1) UUID between them, because i thought i could do that with btrfs to kinda jam them together into a single "volume" but that was not accurate (or i misunderstood how to do it)

anyways i plan on basically just rebooting and unmounting the messed up drives, yeeting the ~/etc/fstab i currently have and generating them 2 brand new UUIDs then cleaning the boot files with dracut and/or update-grub

does this seem unhinged

[rough bloom]

seems fine other than the part where you made them share a UUID

literally has the word unique in the name

[fast pagoda]

well see i was thinking that

but then i was like well maybe it's just unique per logical volume or somethin

i got it in my head that it was a btrfs feature that you could just give them the same UUID in fstab thus jamming them together into a big blob

[rough bloom]

if you just specified the same UUID in fstab then the partitions never actually shared a UUID, fstab doesn't modify the partitions on the drive
it would just mount the same filesystem twice

[fast pagoda]

well i definitely also forcefully yoinked the entire configuration off of one of them

so when it goes to load the UUID it's getting one (1) normal-ish (but empty) volume

and then also pulling the same UUID

for some sort of mutilated eldritch being

of a drive

at first it wasn't ever moving past that it would just sit till i rebooted the system but now at least it just goes into emergency mode and i can force past it into the OS

didn't know getting a message like this on boot could ever be a relief but it was

[olive sable]

Im of to watch the fireworkz

Bye

[sage crag]

$li64 := fn(r0: u8, imm: u64): struct align(1){.op: u8 = 0x4B; .r0: u8; .imm: u64} return .{r0, imm}
$tx := struct align(1){.op: u8 = 0x01}.{}

bin := .(
        // return 255
    li64(1, 255),
    tx,
)

main := fn(): i32 {
        // actually, return 55
    bin[0].imm = 55

    return @as(^fn(): i32, @bit_cast(&bin))()
}

look how much easier it is to assemble hbvm than x86

it works

[hard raptor]

Stoopid linoox beginner thinking installing an arch based distro to non standard storage configurations was gonna be easy

[hard raptor]

Should've gone for nobara fr fr

[fast pagoda]

i feel like i would've also had this particular failing with nobara too that's the sad part

unless nobara handles NTFS properly ootb but it appears that's mainly an ubuntu and derivatives thing

[hard raptor]

Tbf... nobara is based on fedora which as i have told you before, is the consumer version of RHEL... which you would think they would have proper documentation about how to do the thing you tried to do

[fast pagoda]

But then I read it and think "i think this means i should take these 2 volumes and smash them together by robbing one of all of its attributes"

[hard raptor]

If the docs say that exact thing then yes, you are supposed to do that.. and if it doesn't work then... then it's your drive's fault... but i am certain the docs will tell you otherwise

Also.. the rhel and fedora docs on those things usually also apply to the rest of linux

[olive sable]

I lied i ended up at a bar

[noble zodiac]

frick redhat

[hard raptor]

The company yes very much.. but their projects are actually kinda fire most of the time

[dry charm]

File system is mostly the same everywhere

unless you are doing very custom

or your system is configured in a different way (NixOS)

[hard raptor]

Yeha but my point was more... that branch of distros is more often used for servers and enterprise and stuff... if anyone has good documentation on it it's Redhat

[dry charm]

Which is behind a login screen most of the time

And possibly Arch wiki has a more common explanation

[hard raptor]

Hold on

https://docs.redhat.com/en/documentation/red_hat_enterprise_linux/7/html/storage_administration_guide/btrfs-integrated_volume_management

https://wiki.archlinux.org/title/Btrfs

[dry charm]

Ah I mixed up the customer ticket thingie then

[hard raptor]

Arch wiki is also a great resource yes, but i personally find redhat and fedora docs easier to consume

Also, RedHat made and/or contributed to most of the technologies on linux for server environments

[humble shell]

Including systemd and pulseaudio unfortunately

[tender river]

[hard raptor]

I don't get the hate towards systemd tbh

[humble shell]

It doesn't follow Unix design principles

[hard raptor]

And that immediately means it's bad?

[humble shell]

I'm not saying SystemV init or Ubuntu upstart were any better

[hard raptor]

There is a reason most distros use it, is there not?

[amber fractal]

[hard raptor]

I mean, i get that other options are better for certain usecases

But to me.. it's just the thing that starts the things that need to be started and the thing that makes the funny bluescreen

And it could be the boot manager

But all those things still are seperate projects...

[velvet crest]

You have unlocked new role

[humble shell]

Amazing that we're still using GRUB2 in 2025

[hard raptor]

Yay embed perms

Now i can post systemd propoganda

I'd rather post this cursed european wall socket

[humble shell]

[hard raptor]

This badboy can do 480Volts

But they usually don't exceed 400

Cuz nothing that gets placed inside a home really needs the extra 80

[dense kestrel]

What the-

7 holes

[mighty thorn]

https://tenor.com/view/muscle-man-you-know-who-else-my-mom-regular-show-funny-gif-9101122073397073314

[olive sable]

https://tenor.com/view/embed-fail-epic-embed-fail-funny-embed-fail-goose-goose-embed-fail-gif-25175587

Fireworks

[dense kestrel]

https://tenor.com/view/nichijou-cat-pose-cat-girl-girl-kawaii-gif-11820105

What the-

[mighty thorn]

NO EMBED FOR YOU EITHER

Mweheheheh

[potent birch]

Holy fuck, I've debugged why my animation of running ain't animating

[olive sable]

my free membership of childhood has ended

i now have uhhh.. taxes and mandatory voting

[vestal elm]

imagine

[olive sable]

thanks for your input, he who goes by "sexy"....

[vestal elm]

dw i got u sam fam

[olive sable]

how is a plane only 16?

[cosmic sphinx]

a plane might actually be safer..

not for yourself, but for other people

[vestal elm]

my friend got a piloting license and we in hs bruh 😭

[olive sable]

i would say its the opposite

planes can mow down a whole street

[cosmic sphinx]

can but rarely do if its not intentional

on the road though, one stupid accident can mow down a whole street worth of cars, especially on a hway

[olive sable]

ig a plane has more room for mistakes?

[cosmic sphinx]

more room for mistakes, less dangerous for everybody around, stricter teaching

and far not everybody who can afford a car, can afford a plane

even if we are talking propeller ones

[versed cape]

they are iirc
more trained people, better communication, and higher regulations makes them much safer

[lapis wraith]

Maybe "legal" but I seriously doubt someone at that age gets their pilot license

[versed cape]

like if things go wrong your probably screwed, but getting to that point is much more unlikely

[cosmic sphinx]

frankly, a good training course for a pilot could take a few years, so by the time you are 18 you might already be legible to fly a jet 👍

[olive sable]

ngl the 18 dlc is ass

the 16 one was better

[opaque sigil]

be happy, the next dlcs are even worse

[olive sable]

the 21 dlc has gambling

[amber fractal]

The only part of the 21 dlc I want

[opaque sigil]

imagine not having that as part of the 18 dlc

[sour lotus]

My fuuuucccck,finally got it working

Took me 2 months just to get this bloody model injected as a chrome extension

fuck me

I need ot lie down

[sage crag]

its bug time

[olive sable]

At least alcohol is in the 16 dlc

[sage crag]

actually i think i could also do this on x86, but the internals would look very different

[amber fractal]

That's all the way in the 21 dlc here, not that I'm interested in it anyways

[sage crag]

the problem with this method is that instructions are functions so you have to import them

i will probably have to do some funny comptime stuff to make it work without importing

possibly putting those structs with default values inside a wrapper type

and then the functions become methods

[olive sable]

chatgpt, just because you put that comment there doesnt mean that's whats gonna happen

[lapis wraith]

AI hallucinating
What "Instance", mister AI? That's a main function, not a class function

[olive sable]

thats the vertex shader so every single thing you put there will be calculated for every vertex in every instance

[lapis wraith]

makes sense

I never understood shader code. It's black magic to me.

[olive sable]

imo it looks harder than it actually is

its like writing simplified c, but you control positions and colours

for example, a simple 2D vertex shader looks like this

#version 300 es
precision highp float;

layout (location = 0) in vec2 aPos;

void main()
{
    gl_Position = vec4(aPos, -1.0, 1.0);
}

[sage crag]

generally the program is run once for every pixel of the buffer its applied to

[olive sable]

for the fragment shader yes

[sage crag]

more broadly, shaders are programs that run once for every piece of input data

things like compute shaders for example fall under this category

[olive sable]

you can write a shader for probably almost every single thing where you use the same program on thousands of seperate data points

[lapis wraith]

I can get that... but my brain fails to get what even is "aPos". Like, yes, I get is a vector 2 position, but based on what?

[olive sable]

thats just what the cpu puts in

the in in glsl behind the hood does some funky buffer stuff, but basicly you just put in an array, and that array gets indexed for every single use.

[lapis wraith]

Suuure... so... you have "input data", you modify it however you please and pray the GPU understands the new data?

[olive sable]

usually you dont really modify input data

[amber fractal]

Chat, should I try working on NN later or taking a stab at the sam's grass shader?

[olive sable]

im pretty happy rn with the grass so you can do the NN if you want

[amber fractal]

Granted I've not used opengl before so the grass will be interesting to figure out

NN it is, I hope I can figure out where my logic issue is

[olive sable]

i just need to blend inbetween the diffrent layers better, altho the only reason you can see the line is cuz of the spacing making the bottom of the gras smore visible

i am using the same 5 vert / instance model each time now tho

i will probably move towards one with 3 verts for further, and have it be multiple grass / instance

[amber fractal]

bwaa

hard to be on an exercise bike and draw at the same time

[olive sable]

think of it this way, sending data inbetween gpu and cpu is expensive so you want to do that as litle as possible.
so instead of constantly sending data you store the data in gpu and just say which data to use.

in return you get hyperparallellation

[lapis wraith]

hyper.. WHOOT?

[olive sable]

idk the word

my autocorrect aint helping

basicly thousands of the same program running in parallel

[nocturne olive]

Looking very grass

[olive sable]

[lapis wraith]

aja... I'll probably never get it till I, by pure miracle, try to do shaders myself.... but thanks for trying.

[opaque sigil]

Come do general purpose gpu computing, it's way more fun than writing shaders

totally not biased

[nocturne olive]

Or just throw GPUs around and accidentally create the greatest unofficial Neuro voice of all time

[sage crag]

if you want to, you can think of a simple shader like

for x in 0..width:
  for y in 0..height:
    pos = (x, y)
    # code of shader here

[nocturne olive]

-# NeuroSynth BETA was not meant to be this good, it just is

[amber fractal]

Alright, off the bike, What I was trying to express here is that if you just ditto the transformation across all of these connected meshes then not only is it less unique compute but also less verts overall. @olive sable

[tender river]

you can try wgsl here https://neuro-arg.github.io/graphics-toolbox/
sample code

@vertex
fn vs_main(@builtin(vertex_index) in_vertex_index: u32) -> @builtin(position) vec4<f32> {
    let x = f32(i32(in_vertex_index) - 1);
    let y = f32(i32(in_vertex_index & 1u) * 2 - 1);
    return vec4<f32>(x, y, 0.0, 1.0);
}

@fragment
fn fs_main() -> @location(0) vec4<f32> {
    return vec4<f32>(1.0, 0.0, 0.0, 1.0);
}

this is what i was originally making erf for btw

[olive sable]

im sorry but i cant really tell what the drawing is supposed to show

[amber fractal]

All good, I can redraw. It was bad

[sage crag]

did you see i got the assembler working

(for hbvm)

[olive sable]

what is even th epoint if your gpu doesnt show pretty colours on the screen?
youre the reason we have a shortage

[sage crag]

computational physics

wahoo simulations

[olive sable]

im sold

[tender river]

i did

[olive sable]

we love physics

[amber fractal]

My dumb self about to use red and green for the drawing

[olive sable]

i can see the diffrence between red and green mostly

[amber fractal]

Yeah, better off not doing it regardless

[sage crag]

what about brown and green though

[olive sable]

depends on how bright the green is

[tender river]

you can even show the image it has loaded but the code is a bit big

[sage crag]

for hbvm it was thankfully as simple as packing the opcodes and operands as structs, and then putting that into a tuple

issues exist tho, since you have to import instructions becore you can use them

[tender river]

hmm would returning @Any() work since you only need the data you dont necessarily need to do anything with it

[tender river]

you can at least import asm as x or whatever, x.ld64 is not much longer than .ld64

[sage crag]

return Any

if only hblang had type backflow

then perhaps

[sage crag]

its uglier and thats all that matters

if we had a way of importing all members of a scope at once then i wouldnt mind

would be helpful with re-exporting target stuff too

[olive sable]

when konii says it its a brave statement, when i say it im "the bad bracket man"

double standarts smh

[sage crag]

the difference is im correct

[olive sable]

konii im legaly allowed to drive a car now

[olive sable]

im willing to bet even a crt tv wont win to a 2001 opel zafira

[amber fractal]

Hopefully this more coherent

[sage crag]

i can drive a tractor

[olive sable]

you really are like my irl friend

damn

[sage crag]

[olive sable]

this is getting creepy

[sage crag]

we could be the same person, you will never know

[rigid snow]

i'm legally allwoed to drive anything and i still drive illegally (i don't have a license)

[olive sable]

i fucking hope not

[rigid snow]

wait no i actually have an atv license

[sage crag]

its ok, im to your best knowledge over 600 miles away

[rigid snow]

why is that a thing even genuinely

[olive sable]

if you really are a diffrent person tho

[sage crag]

why dont you ask your friend

[olive sable]

what do you expect me to say?
this discord girl is suspiciously similair to you, prove you arent stalking me

[amber fractal]

"awa"

[sage crag]

yeah

[olive sable]

wait thats gebnious

i will send awa

[rigid snow]

that sounds like something an insane person would say

[olive sable]

so not too out of place

[rigid snow]

a bit out of place i'd say

[olive sable]

she hasnt checked her messages in months tho, that last one wasnt even read

[rigid snow]

but konii now knows you sent her that

[olive sable]

the test is to see if i get response back as "awa"

[rigid snow]

she now knows to act inconspicuously and not respond "awa" back

cover is blown i'm afraid

experiment invalid

[olive sable]

but my irl friend doesnt know

[amber fractal]

[sage crag]

awa

[olive sable]

wait i have her phone nummber, if she doesnt see it for a week i should send her someplace she does check

cuz her messenger account is long forgotten

[rigid snow]

stalking ...

[sage crag]

the

[olive sable]

im just sending "Awa" how is that stalking?

[tender river]

please no wildcard imports

[rigid snow]

is my isp still fucking blocking crates io

i'm going to crash out

[amber fractal]

let's go gambling

[olive sable]

[amber fractal]

lootcrates and all

[sage crag]

use a custom dns

[rigid snow]

i use a systemwide proxy but cargo ignores it

[olive sable]

im going to watch horse racing irl in august cuz we like betting on them

[sage crag]

how about module exports instead

[tender river]

what does that mean

reexport everything from module?

[sage crag]

unrelated but i can make wildcard imports using comptime by bruteforcing all possible strings in @has_decl

[tender river]

dns usually doesnt matter because russian censors are actually competent, but crates.io not working sounds more like a local thing so it might be dns related

[olive sable]

you're doing 3D grass.
rn im doing 2D and rotate them towards the camera

[tender river]

can you really

[amber fractal]

[rigid snow]

it's probably a routing issue tbh but it's been a thing for like a month atp

why tf would they block crates ..

[sage crag]

i was thinking of something like
@exports(a, b, c)
@useExports("lib.hb")

[sage crag]

yeah but you would have to append them to an array

[tender river]

i'm fine with wildcard imports as long as they are scoped very narrowly and dont overlap each other

when theres more than one it quickly turns into a mess

i hate C# for that

[sage crag]

the problem with hblang imports is importing re-exports

since there is no notion of public/private

[olive sable]

im doing smthn like this rn

[tender river]

pythonlang

[tender river]

it sounds like just adding public/private to the language would be a better solution but maybe there are better alternatives

[amber fractal]

I wonder if I can cull a vert off of this depending on facing direction. Maybe billboard directly face on or cull the unseen bottom vert. That'd only work for single tho.

So either 3 or 4 verts and still considered 3d. That's all I got at the moment.

I guess in the end it is about the same as 2d anyways.

Maybe with the small benefit of lighting being being possible and physics only on one node. But that's the same as 2d.

[olive sable]

the way im doing things now physics is on 3 vertices

imo thats fine for closeby but after a distance we'd reduce it to 1

3 allows some curvature

this is only visible closeby tho

[amber fractal]

Yeah, only question becomes is there any benefit for 3d besides batching?

[olive sable]

i still dont get what you mean by batching? cuz we could also batch 2D

[amber fractal]

Because any culling one would do only affects faces and edges instead of verts when combined

[olive sable]

ye true

[amber fractal]

Yeah

[olive sable]

afaik all vertices get calculated even if they arent visible

[amber fractal]

I guess I'm using batching as combining meshes and calling only once instance of physics

[olive sable]

ah

you could do that for 2D too

[amber fractal]

My head is just stuck in trying to do it in 3d for some reason.

[olive sable]

i can try a 3D version if you'd like

[amber fractal]

Nah, let me apply the same train of thought onto 2d first

then I'll see what I like better

[sage crag]

awa

[olive sable]

awa

2D vs 3D

granted i wasn't using an index buffer eventho this would have benefited fromt hat

[amber fractal]

I know 3d is on the left

[olive sable]

im doing triangler fan for 3d so i dont have to calculate 3 vertices per triangle, but the first and last one was still calcualted twice

[olive sable]

correct

[amber fractal]

Much less fps was reason #1

It does at least look fuller

[olive sable]

this is 2D with only 3 vertices, the humble triangle

[tender river]

i think it's time to support \uXXXX in the json parser

[amber fractal]

I don't like either of my described methods. They both feel incomplete and non optimal.

[sage crag]

no hypers

[olive sable]

i have no idea what \uXXXX is so i hyped

[sage crag]

everyone who put hypers has not experienced what this woman does to the poor compiler every time she extends the json parser

[tender river]

this will not be a big change

[olive sable]

what do you do to yuor compiler?

[tender river]

it will just be a painful change

i think the compiler wont suffer

but i will

[sage crag]

https://git.ablecorp.eu/mlokis/hblang/src/branch/main/BUGFIX.md

most of these are bwaaleaf