programming

[amber fractal]

Nope

[olive sable]

ofcourse it got used as the intro to pootis engaged, but the song itself is about thighs

[amber fractal]

Imma just perish

[olive sable]

the song is a banger tho, ill just pretend i don't know

[gritty dust]

Frick u

I was at work

Lol

Merci beaucoup

[olive sable]

cheese has a job?

[gritty dust]

Whattt that's crazyyy I work, wild

[stray dragon]

uhh

yeah that is pretty fast

[maiden geyser]

helvetica staaaandaaaaaard

[maiden geyser]

floating point error vedal

[topaz tendon]

Looks like FanControl got rid of the WinRing0 driver

[maiden geyser]

i don't think this pawn is quite free anymore

[patent shard]

yes, a while ago
Windows started cracking down and labeling anything with winring0 as a trojan

[naive burrow]

Looks like they are training a AI to train it self to do autonomous real world tasks.
Instead of actually training the AI with real scenarios and spend a ton of compute and time, you will be able to pass videos to the AI, and then it will start a process of "imagining" itself doing the task and learn it alone.

In this example they used Minecraft, showing that it was able to mine a diamond by itself, with no AI was able by now. (not even neuro)

For now it's just a technical paper, but it might be interesting to follow the project. Perhaps Neuro can benefit from it in the future and become a better gamer with much less effort from Vedal
https://danijar.com/project/dreamer4

https://youtu.be/oDlBtTcX0g0

[obsidian mantle]

buttons

looks super good though

noooo

[naive burrow]

that's the video generated for training tho

[topaz tendon]

I only noticed it now since I apparently ignored FanControl's update notifications for a while and I've been like 10 versions behind...

Either way, better later than never ig

[obsidian mantle]

Oh

Wait but

Whats the point

Is this video ai generated?

Ai made training video for another ai
Did i get it right

Why did you say it "was able to mine diamond by itself"

Or there is no evidence/footage of the ai actually doing it

[warped narwhal]

I splashed out on my first really fancy keyboard

Do I need it? Probably not. But do I want it... Yes

[opaque sigil]

I always tell myself the custom keyboard I've built for way too much money is worth it because of how much I use it

[maiden geyser]

not even an HE something my head

[warped narwhal]

I don't need the features of HE, and I prefer tactile keys.

But I did look into it (but holy, the price )

[obsidian mantle]

do they make good low-profile (or how is it called? when the buttons are not 3cm towers but flat like in laptop)

[nocturne olive]

I think Windows no longer accepts the WinRing0 driver

[obsidian mantle]

it always seemed to me that they only make "normal" keyboards like that

i mean high quality and all

[warped narwhal]

They do have a couple low profile options, but I can't speak for their quality

[topaz tendon]

Yeah I recall Defender blocking it

[nocturne olive]

Indeed

[topaz tendon]

Haven't had to deal with this personally since I use ESET instead of Defender

[nocturne olive]

I don't have to deal with that stuff anyway since I do not use Windows

[topaz tendon]

I tried Linux several times, it's just not a good fit for my home PC unfortunately

Always something that's borked

I was even installing some Linux distro that seemed to be popular recently (CachyOS) and it really didn't give me a good first impression when the system installer crashed with a Python error soon after it started installing...

[maiden geyser]

it always was in the vulnerable driver list, wasn't it?

[opaque sigil]

Cherry has some really interesting low profile switches but idk any pcb that accepts them

a lot of this is also just the keycaps tbf

[maiden geyser]

linux wouldn't be linux if you could find a distro with just the right colour and taste

[patent shard]

Windows does this now with anything with winring0
you can obviously create an exception and force it, but there's a reason Microsoft added this to Win Defender

[topaz tendon]

I've tried like a dozen by now 😭 only so much time I'm willing to burn for trial and error

[maiden geyser]

https://www.youtube.com/watch?v=FYjs7vsaSEw

this is the fourth time my iems made that sound today

this must be all in le head

[nocturne olive]

Only recently started getting actually blocked

[obsidian mantle]

anybody knows how to stop my program from any point of a program

and not call exit; until i get to the main

on every subfunction

oh shit division by 0 actually works

ok but now it dies before the message arrives

[quiet marlin]

any logs....? without logs there's no way to know what's wrong
could even be a network card issue that crash the system(especially mediatek ones)

[topaz tendon]

I only got a photo of the error, one sec

[topaz tendon]

When I chose GRUB instead of systemd-boot, it installed fine

[quiet marlin]

try normal... well I mean generic distros like archlinux(which cachyos is based on?)?

[median shale]

Meow

[quiet marlin]

that's the fault of that installer I guess

[median shale]

Happy birthday @topaz tendon

[topaz tendon]

I haven't tried bare Arch, but from memory I tried: Ubuntu and some flavors, Debian, Mint, Manjaro, Antergos, Fedora, and probably something else I missed

[topaz tendon]

Thank you <3

[quiet marlin]

did they have similiar errs related with partitions or mounting?

[topaz tendon]

Nah, Cachy is the only one that crapped out during the installation process. Other distros had other things wrong with them, not necessarily errors, just configuration/performance/stability problems

Some of those go several years back so I can't give you the details, but generally I've had issues with getting VRR working, mouse movement in games being iffy, and some other things

[quiet marlin]

could this help...? https://wiki.archlinux.org/title/Variable_refresh_rate

[topaz tendon]

I did research my problems, no worries

If I recall correctly, VRR was something that was temporarily broken by Nvidia's proprietary driver

Trying to get an older one than what's in the repos was a mess so I let it go

[quiet marlin]

mouse problems could be related with so many things
if that's windows games sometimes it's even wine/proton's problem

[quiet marlin]

did the nvidia-open driver help?
it did improve a lot

[topaz tendon]

The main one was TF2, which happens to run natively. I tried several tweaks to change that, incl. disabling mouse acceleration globally in some configs etc.

I don't think nvidia-open was a thing back when I was testing this

[naive burrow]

No, my bad
I got confused because I thought that the print screen was some hallucination of the AI, like in that minecraft gameplay generated by AI. But it's just very low resolution

The print screen that you took was in fact buttons crafted by the AI.

The AI in the video is actually playing minecraft and mined a diamond

[obsidian mantle]

nice

my negamax gives away his queen on turn 5

surely static board was great after that

[nocturne olive]

I don't know, I would throw a special custom throwable and catch specifically that in the main

[obsidian mantle]

what the hell they literally

trade ANYTHING to get a check

queen for rook + check? best move ever

bishop for rook and check? great, too bad i cant trade my queen for that

does this negamax even work

is this a trolling

i think if i tell it to choose the worst move they will actually do better

no it does even worse

[cobalt sigil]

https://www.threads.com/@tecny_guy/post/DPa3P10D9HE?xmt=AQF05HcUBksjlfyOsqOyETSoSooYtMXDoLCCpMjJLvb6jA&slof=1

[rigid snow]

hbd
also interesting pfp choice

[rigid snow]

people in replies saying it’s a damaged screen

[topaz tendon]

Thank you, and thank you :)

[rigid snow]

eliminiminiminate

[tender river]

minimum fieryrage

[topaz tendon]

drop it like an elevator 🗣️🗣️🗣️

[verbal grove]

anyone here good with typescript?

[rigid snow]

maybe me

perhaps

[verbal grove]

well i might need help fixing up an extension

for vscode

if you got time

@rigid snow

[rigid snow]

i think i do have time ask away

[verbal grove]

i think its better if we call since i can stream whats going on

so call me when you're ready

[rigid snow]

server vc doesn't suffice

[verbal grove]

private vc will

might aswell

[hoary lion]

am bored

[uneven pulsar]

yeah this is what i am working on currently

[hoary lion]

devil may cry

[jagged turtle]

[hoary lion]

man this guy got typescript extension

[jagged turtle]

well yeah how else are you gonna mod vscode?

I mean ig you could wasm or just idk distribute binaries from other langs and make ts bindings for it?

[uneven pulsar]

it got automatically downloaded but i didn't care because whatever

helped out alot ngl

[jagged turtle]

actually does vscode's node even support wasm

[obsidian mantle]

how do i sort one array based on another array
they are of equal size

is it o(999)

(idk what o(x) means)

[hoary lion]

erm

I think yall misunderstanding

😭

it was about the name

[jagged turtle]

oh it was added since 2020, interesting

wonder if wasm types were added tho

[obsidian mantle]

i feel like bubble sort is slow and sucks

[shadow sinew]

Use qsort from stdlib?
https://www.w3schools.com/c/ref_stdlib_qsort.php

[obsidian mantle]

but i will need to sort another array

i need 2 arrays sorted by 1 array

hmm actually

it can take function

so i need to give it some kind of pairs

[olive sable]

Goodmorning

[obsidian mantle]

nah this qsort cant sort pairs

i gotta use my shitbox bubble sort i guess

it probably can but im too bad

[tender river]

use std::sort

with lambda

[obsidian mantle]

qsort has lambda too

idk how to set up lambda

why cant i do this

[tender river]

go read docs

[obsidian mantle]

i want it to depointer int

from pair

and it cant do it

[tender river]

wrong code

[obsidian mantle]

also idk why i need to copy them

[tender river]

remove two = lines

use ->

a->first - b->first

[obsidian mantle]

you are so smart

[tender river]

cant . on ptr

[obsidian mantle]

will it work

[tender river]

can -> or *

[obsidian mantle]

sort will probably say something

cant work with pairs wtf is pair

[tender river]

a->b is just (*a).b

[obsidian mantle]

i tried *(a).b

almost 😠

whatever

[tender river]

that same *(a.b)

a pointer -> cant do a.b

[obsidian mantle]

ok so qsort says "i cant take compare which has pairs wtf is that"

[tender river]

read what errors compiler gives

[tender river]

use sort not qsort

[obsidian mantle]

sort says same thing but phrases it differently

or maybe i need change arguments

[tender river]

that different thing

[obsidian mantle]

it doesnt say which argument it doesnt like

[tender river]

read std sort docs

[obsidian mantle]

it just says "bad arguments

[tender river]

and use iterators

.begin() .end()

[obsidian mantle]

this is unreadable https://en.cppreference.com/w/cpp/algorithm/sort.html

i hate this website

[tender river]

it just c++

[obsidian mantle]

ok my pair has no end

and no begin

[tender river]

that not on pair

[obsidian mantle]

ignore unconst array warning

[tender river]

not use array std::vector

vlas not something beginner should use

[obsidian mantle]

i googled vlas this word doesnt exist

[tender river]

https://en.wikipedia.org/wiki/Variable-length_array

this not use

use vector

[obsidian mantle]

its safe here

what can go wrong

does it not deallocate it

automatically when i exit

i dont change its length ever again

[uneven pulsar]

works

[tender river]

you barely know the language

[obsidian mantle]

its bad when you want to change it but its not the case, len_submoves is const within this entire function that uses this array

therefore i can use it freely here

am i wrong

[tender river]

vla not even in c++ standard

just dont use until know what you're doing

[rigid snow]

feet/feet

[tender river]

[obsidian mantle]

i have at least 10 vlas and it works fine

because every array even though being of "variable" length
is declared once and never changes length within scope

g++ builds it and i didnt care before

now idk what to do

vector cant take move

this is stupid i'll just use my bubble search which actually works

and not these lambda bullshit

maybe i'll research it later

end and begin arent even pointers they are just index number what the hell

how does sort know what array to change

of course it cant take pair

it probably uses 5000 line of code decoding of its arguments

[tender river]

wanf use c++ learn its complexity

[timid cloak]

too hard

[shadow sinew]

They're iterators. Sort of an abstract representation of a position in a data structure.

Well not "sort of".

[obsidian mantle]

yeah

so sort() which i call doesnt even know which array it plays with

because it takes end and begin iterators, and function to compare

but its supposed to work because i guess it takes those arguments and strips it from .end and does something else

but then i cant give it just any datatype

it surely can sort normal arrays but when i give it pair it says

"i cant sort this"

actually not waht it says it says it has no end

wait so i can just give it 0 and len_submoves

guess what you are sorting?? idk lets see

yeeeah it cant do shit with it

[shadow sinew]

If you want to sort a regular C-style array you can just give it pointers to the beginning and end. pointers can be used as iterators.

[obsidian mantle]

i have 2 arrays

one is moves

the other is values of moves

i need to sort moves by their values

basically sort 1 array by values of another array

[shadow sinew]

Can you turn it into an array of pairs and sort that?

[obsidian mantle]

i struggle to give it .end and .begin of pair so it understands it

array of pair doesnt even have end or smth i dont get it

[shadow sinew]

Use &array[0]' and &array[len_submoves] instead.

Or array+0 and array+len_moves to save tokens. 😛

[obsidian mantle]

it gets 500 errors trying to parse this anyway

actually just 2

ok i figured it out

so actually compare() function has to take actual datatypes of something being compared and not their pointers

sort() can take pointers and not end() or begin() while all examples ambiguously give .end and .begin examples

thank you all for help

[olive sable]

i just got shared a file in onedrive form one of my throw-away alts

it doesn't even contain a virus, its just a shitpost

the file is from 2018

weird, anyways

[tight swallow]

Just been working on making my own first AI. Found it was not displying the way I wanted. Was using Vseeface, kept tracking my movements and saw no expressions being used.

[obsidian mantle]

it doesnt sort anything whyyyy

it actually does change order but its not correct

what is wrong with this

first of all there is no element move_and_score[len_submoves] but idk if it wants the start of last element or the end of it

i guess it should be start so i put -1 before ]

[obsidian mantle]

then it still gives something like this

[shadow sinew]

len_submoves is a runtime value. In C++ you have to give a compile time constant to size an array.

[obsidian mantle]

why does g++ build it then

without problem

[shadow sinew]

Ah, right.

[obsidian mantle]

why is it even a problem

also i think i fixed it

in compare i had to put < instead of -

and it actually wants &move_and_score[len_submoves]

even though this pointer is out of bounds of array

[shadow sinew]

Yeah, because that's how iterators work. The end element is the one just outside of the range.

[obsidian mantle]

yeah order seems to be correct now

i think i did it

[tender river]

c++ vlas nonstandard extension

[obsidian mantle]

is it banned in competition

[shadow sinew]

Also regarding the < vs - thing that's a difference between the C qsort and the C++ std::sort.

[obsidian mantle]

do i have to switch to vectors

[shadow sinew]

You can stick to arrays if you size them large enough that it doesn't matter. I size all my moves arrays to take 256 elements. That's more than the chess board with the most known legal moves.

[obsidian mantle]

but these arrays are local to each function that uses them

[shadow sinew]

Yeah, they get put on the stack.

[obsidian mantle]

and im not changing their size ever

so current negamax destroys all my previous bots

i guess i wont even submit that garbage then

its not about complexity of cpp its about ambiguousity of things people created using it

oh but actually this negamax is much worse in finding mate

i should work on that

[maiden geyser]

why can't people just use c, are they stupid

[obsidian mantle]

its too limiting

no oop

wtf is that

what is this, 1984?

(im not using anything from oop in my bot though)

unless you counter that sort class-like

[maiden geyser]

wdym "no oop", structs and unions is enough to make everything else

[obsidian mantle]

can you put function in your struct

(genuinely dont know)

[maiden geyser]

you can put a function pointer in it

[obsidian mantle]

mhm i guess

[shadow sinew]

That's pretty much what C++ does. Every object with virtual functions stores a hidden member variable with function pointers. You can do the same with C but it's more tedious.

https://en.wikipedia.org/wiki/Virtual_function?useskin=vector#C

[clear sedge]

have you heard of a vtable

[obsidian mantle]

it probably looks so messy

[clear sedge]

it usually looks like obj->func(obj, ...);

you might choose to have a separate struct for the vtable so the syntax varies on how you approach it, i prefer to declare a Obj_func(Obj *this, ...) and call that so i don't have to worry about the vtable whenever i call things

[true hemlock]

5.2GHz on 772mV stable (kinda)

lets goooo

godbin chip

[clear sedge]

i don't think he checks this channel that often

[midnight sigil]

[clear sedge]

he might be more active in #neurotic-neurons but there's no guarantee he'll see it

remember to not ping him tho

you ping tutel you go bye bye

shiro can attest to that

[stark needle]

Can i try pinging ved ✅

[clear sedge]

do it pussy

[stark needle]

I will send the gif

https://cdn.discordapp.com/attachments/1199359059753697302/1295647276987056149/image-11.gif

it does

[clear sedge]

wait, does he?

[olive sable]

stop being so depressing

i mean, there wasnt really a vibe since chat is dead rn

[midnight sigil]

thx

[clear sedge]

brother you're 13, life hasn't even started yet

[shadow sinew]

He lives in the UK

[midnight sigil]

yuh

[clear sedge]

someone doxx vedal so that we know where he lives

[midnight sigil]

bro got the most typical UK kitchen set for the irl videos

100ms is slower than our reaction speed

[clear sedge]

talk for yourself

my reaction speed is approximately 3 years

[midnight sigil]

you reacted to my message under 3 years

[clear sedge]

i time traveled from the future so i could send that message

checkmate

[midnight sigil]

bro remember everything we sent

[clear sedge]

[midnight sigil]

yuh

https://tenor.com/view/neurosama-gif-4202543522129494896

I actually got a 40.25(iirc) seconds on the classic sheep fricker seed

didn't bother to post it

wanted to get a sub 40 first

40.082 my bad

fucking 2 ticks man

if I can spend less on crafting

[nocturne olive]

Where would I even obtain this specific model?

I have been informed there is an even more powerful model than the one I am using and I need to figure out what it is

[obsidian mantle]

uhh iterative search and transposition tables look scary

[midnight sigil]

it's quite simple

[obsidian mantle]

do i just compare my current hash with all transposition table entries

it seems to be extremely slow

[midnight sigil]

you make a table
search for position in table
if it's stored, find the node type and set its lowerbound(alpha), higherbound(beta) or exact value to the current search(of individual nodes)
store new node to the transposition table

[obsidian mantle]

search for position? how do i search it if hash is random

or is it not

[midnight sigil]

wuh

you just make the hash a key and the nodes the value

hash table

[midnight sigil]

yuh

[obsidian mantle]

slows bot by 10 times

• it looses

[midnight sigil]

you shouldn't iterate through the table to search for the position

[obsidian mantle]

what is the "position" we are talking about

isnt it the board state

[midnight sigil]

the zobrist key

[obsidian mantle]

is my table correct? do i need some "uint64 position" too

so key isnt random

[midnight sigil]

yuh

it represent the board, the unique position(there will be hash collisions but whatever)

you can just obtain the position by transposition_table[zobrist_key]

that'll do the job

[obsidian mantle]

what the hell

but

[midnight sigil]

[obsidian mantle]

doesnt it do the same

[midnight sigil]

well it won't cause you time to iterate 10000 elements per search

it's a hash table, your compiler will do the job for you

[obsidian mantle]

does it create 64x64 table then

so i dont push back

i add to some position

[midnight sigil]

wuh

[obsidian mantle]

like

vector[position]=value

but my vector is empty

i have to make it [64bit]

[midnight sigil]

vector[zobristKey] doesn't exist

[obsidian mantle]

i meant figuratively

[obsidian mantle]

transposition_table[key] = value instead of this

will it work

[midnight sigil]

no you gotta push back

vector[zobristKey] doesn't exist

until you made it to

[obsidian mantle]

but if i push back it all becomes random

[midnight sigil]

it is random

[obsidian mantle]

transposition_table[0] is 3412412313123
transposition_table[1] is 312310391089231032109388091

[midnight sigil]

but it represent a board state

[obsidian mantle]

and it magically finds the [index]

when i give it value from the right

[midnight sigil]

oh wait you're using vector FUCK

[obsidian mantle]

this cant be right i am not getting something

[midnight sigil]

[obsidian mantle]

i can use something else

i used vector because it has nice pushback and is supposed to work well with dynamic length arrays

which is exactly my case since im going to push back thousands/millions of times

[midnight sigil]

you need to implement your own table thingy

[obsidian mantle]

i cant understand how it can work if i dont create transposition_table[18 446 744 073 709 551 616] right away

[midnight sigil]

or std::unordered_map

if you're on C++ ofc

[obsidian mantle]

this looks like something i never heard about

how does it even work

[midnight sigil]

it's json object but a hash map

abc: {123}

you do map[abc] you find 123

[obsidian mantle]

ok where do i get abc from to find it

i only have 123

[midnight sigil]

you don't, just get your value stored as 123, and the key as abc

[obsidian mantle]

but then i have something like

hash1 : {1312}
hash2 : {31232113}
hash3 : {31}

[obsidian mantle]

i give it hash and it has to iterate through all hashes just like i did

since hash1,2,3 are ALSO random numbers

[clear sedge]

for(std::pair<const K, V> &pair : map)

[obsidian mantle]

it would only work if i had 18 446 744 073 709 551 616 length array

so i just get hashX position from it right away

[midnight sigil]

I don't understand

[obsidian mantle]

so i have

a table which has

hash(random number), value(random number)

i want to find if new hash is stored

so i get my current hash

i do not understand indeed

nono its like

lets say i tested 2 boards

my current transposition table is

{1312313132113312, -3333}
{12313193242342341, 4000}

i want to test if my new hash (lets say its 12313193242342341) is in the table

[rigid snow]

[obsidian mantle]

how do i check if its in the table if all table contents are random

and i have random number as input to the search

i mean i know how i do it

i just iterate through entire table

[midnight sigil]

nah you just call std::unordered_map::find, it's O(1)

[obsidian mantle]

how else can it be possible, to find random number in random number table?

[midnight sigil]

if std::unordered_map::find == std::unordered_map::end

[opaque sigil]

sounds like what you want is a tree not an unordered_map

actually

maybe not

[midnight sigil]

true if present false if not

it's that shrimple

for find you pass in your key as argument

[obsidian mantle]

i would understand if it was ordered on push_back but its not

so vector is dead way

[midnight sigil]

define your struct with only the valuation for now.
your unordered_map definition should be unordered_map<uint64, transposition>

[midnight sigil]

vector is not used for such case

you should be using a key-value map

vectors are arrays, here you should be using a dictionary(in Python's language)

yes vectors are not ordered

[midnight sigil]

[obsidian mantle]

so the trick is to put hashes in order

[midnight sigil]

NOOOO

YOU USE AN UNORDERED_MAP

[rigid snow]

[obsidian mantle]

?????

[opaque sigil]

manually sorting hashes is never the answer

okay sometimes it is but not here

[obsidian mantle]

noooo i mean how does it work at all

how can it be possible

find random in random list

without iterating through all

the list has to be ordered!

no?

[rigid snow]

[opaque sigil]

no

[midnight sigil]

^^^ if you want to find is a key present in the map

[patent shard]

Huggingface
Pcunwa>kim-mel-band-roformer-ft
If the instruments are loud compared to the voice, it can start to get muffled
Big beta 6x has a little more fullness for those, but it doesn't have as high of a bleedless score (mel-band-roformer-big)
For removing reverb, search anvuew mel band
However, any voice not in center channel kind of just gets removed..

[obsidian mantle]

(i dont order it i use some unordered_map::std::abcaasd)

which does it for me

[obsidian mantle]

whatever im tryin to do this now

[nocturne olive]

Hmm
How does MB-Roformer-InstVoc-Duality-V2 compare?

[midnight sigil]

https://www.geeksforgeeks.org/cpp/unordered_map-find-in-c-stl/

go grab your Geeks for Geeks here

[obsidian mantle]

so i do something like this

[opaque sigil]

[obsidian mantle]

insert to add entry

[opaque sigil]

you don't need to hash the key yourself btw

unless you want a hashes hash i guess

[obsidian mantle]

im getting confused with names

[midnight sigil]

:YES:

[obsidian mantle]

so i get chess_zobrist_hash

and there is other "hash" in unordered_map

[midnight sigil]

I mean, well it has to

the hash is the chess board hash

[patent shard]

doc I look at says "Outperformed in both metrics by the unwa’s Kim FT"

[obsidian mantle]

ok this worked

now i have to search and its done

[opaque sigil]

ah, then yeah that's fine

[nocturne olive]

Hmmm

[midnight sigil]

idk if that little {} would affect the result or not

[nocturne olive]

https://youtu.be/OnkXWkvLSkc
I'm trying to find the model matching what is used here, and so far my best match is that one I mentioned

[midnight sigil]

it might be treating it as an array instead of two parameters

[nocturne olive]

You got a better matching one?

Target file

[obsidian mantle]

it doesnt work without it

[nocturne olive]

My result

[clear sedge]

remove the {} and make a habit of using emplace in place of insert

[opaque sigil]

i wonder if the zobrist hash itself is random enough to use it as a hash function for the map

though i guess at that point you're better off just making your own hash table

[clear sedge]

(emplace constructs the object for you efficiently, while insert expects an object to be constructed separately)

it also saves 2 tokens

:3

[opaque sigil]

Zobrist hashing is the first known instance of tabulation hashing. The result is a 3-wise independent hash family. In particular, it is strongly universal
hmmm, sounds like it could be fine

[patent shard]

sounds pretty high bleedless, that's good

[midnight sigil]

they are the same smh

[midnight sigil]

no it's not

[opaque sigil]

[nocturne olive]

It has slight manual cleanup (pasting one part from the result over an earlier identical part)

[obsidian mantle]

doesnt work without {}

[midnight sigil]

the collision probability is definitely higher than the built-in hash function for hashmap

[obsidian mantle]

or maybe visual code is lying

[midnight sigil]

yes you're right if you use insert

my bad

[obsidian mantle]

ok ok

[nocturne olive]

But the question is how well does it match the reference?

[clear sedge]

the performance impact in this case is negligible since i think insert has an override for std::pair<K, V> &&
but you do save 2 tokens with emplace :3

[midnight sigil]

or use python, more token saves

[obsidian mantle]

wait so first i have to check if this hash exists at all

and only then i take its value

[clear sedge]

std::unordered_map::find

[obsidian mantle]

what the hell does it return

[clear sedge]

or just use std::unordered_map::at if you're fine with it crashing if the hash doesn't exist

[clear sedge]

an iterator

[midnight sigil]

you will obtain transposition_table::end from transposition_table::find if your table does not have that specific key

you can check afterwards

no biggies

[midnight sigil]

[clear sedge]

for token saving purposes just write auto test = ...; :3

[opaque sigil]

i sure love getting iterators of length 1 back

[clear sedge]

then like caibi said, compare it with transposition_table.end() and then you can read the value with ...::operator*() (aka dereference operator it)

[obsidian mantle]

it is defined i dont know its type

[clear sedge]

susge

[patent shard]

this one has slightly less noise, however, possibly slightly less fullness? hard to tell, and I only listened for like 30 seconds

[midnight sigil]

dereferencing references that points to dereferenced references...

https://tenor.com/view/venom-gif-27487650

[obsidian mantle]

i mean

i dont want to .find 10 times

first to check if its there

then to get its value

but looks like im wrong and getting value does not require .find

[faint sandal]

is aur down

[nocturne olive]

Hmmm

[obsidian mantle]

so this is correct?

[opaque sigil]

[obsidian mantle]

(i check if its in theere)

[faint sandal]

[clear sedge]

that is correct yes, obv you won't get the iterator from it though

if all you need to do is check if it exists, that works

[midnight sigil]

replace asserts to save tokens from if-statements

[obsidian mantle]

i have no idea what these functions return

what kind of datatype is this

truly,

skill issue

[clear sedge]

a node handle

[midnight sigil]

pointers

[opaque sigil]

for the future, if you're ever unsure how to use a function the cppreference has examples for pretty much all of them

[clear sedge]

https://en.cppreference.com/w/cpp/container/unordered_map/extract

we love cppreference

[obsidian mantle]

it doesnt suggest anything

[clear sedge]

you don't -> it

[obsidian mantle]

wait

yeeees

[clear sedge]

yis

[midnight sigil]

you can just do int = transposition_table[key]

[obsidian mantle]

the dot

[clear sedge]

but then that creates a value at that key if it doesn't already exist

which might be an unwanted side effect

[opaque sigil]

[midnight sigil]

you already checked if it exist smh

[obsidian mantle]

i check if it exists first and dont try to read it

[clear sedge]

so much work
just put std::unordered_map::at in a try-catch block and call it a day, ez

[midnight sigil]

extra 1 token

[obsidian mantle]

yea visual studio is clearly blind

it says its not defined

because i forgot }

[clear sedge]

[obsidian mantle]

so you telling me

that this

[obsidian mantle]

is faster than my for_each

while everything is random

[clear sedge]

¯_(ツ)_/¯

[patent shard]

I still like big beta 6x, has so much more fullness.. but the bleedless metric is quite a bit lower

[midnight sigil]

it has time complexity of O(1), your old one is O(n), so, yes

[clear sedge]

you are querying the unordered_map twice

if you want to optimise for speed, assuming the compiler doesn't do it for you, you can store the iterator find returns instead

[obsidian mantle]

i just run it and it seems to be slower than my for_each

maybe i have a different problem somewhere

[midnight sigil]

surely my beloved clang understand my need with -O3

[obsidian mantle]

so this isnt just "take an element from array" actually, this is some super search mechanism behind the scenes?

[clear sedge]

i fear no main
but that thing
-O3
it scares me.

[midnight sigil]

*it probably should

[clear sedge]

it's just doing normal map things
gets a bucket, yada yada, checks if the key exists, if not it creates a new value

[midnight sigil]

too simple to not be recognised by compilers

[clear sedge]

you might replace it with .at(current_hash) and MIGHT notice a speedup, but that depends on compiler flags and optimisations more than anything

plus it'd be a pretty tiny speedup

[obsidian mantle]

WAIT does it mean when i search it next time, it will find this "fake" search?

[clear sedge]

i recommend just storing the iterator

[obsidian mantle]

uh nvm i think i got it

[clear sedge]

there's no fake search, you know the key exists

[midnight sigil]

programmers judging if the + operators are faster than << for std::cout

[obsidian mantle]

ok so i do this first now

[clear sedge]

<< is faster, surely

[clear sedge]

✅

[midnight sigil]

there'll always be stackoverflow fucks saying it's not

[clear sedge]

compare it with .end()

then do auto foo = *something_something;

[obsidian mantle]

the if() isnt a problem it just works like that.
the problem is to get the content

[clear sedge]

don't call that

do int value = *something_something;

[obsidian mantle]

doesnt work

is it me or it looks tilted

[clear sedge]

oh right

it's a pair

[obsidian mantle]

(i guess my value is second?)

[clear sedge]

or something_something->second

[clear sedge]

✅

[obsidian mantle]

yeah it says second is int

and first is uint64t hash

[clear sedge]

first is the key, second is the value

[obsidian mantle]

now it isnt slow

[opaque sigil]

[obsidian mantle]

magic

[opaque sigil]

yeeeeeeah

[clear sedge]

foo->second is less tokens than (*foo).second btw

just saying

[obsidian mantle]

yee im not hungry for tokens yet
trying to add functionality

but i did that ->

[clear sedge]

[midnight sigil]

if you're using alpha-beta pruning, you should be storing more data other than the evaluation, as the node trees are not fully searched but pruned.
See https://www.chessprogramming.org/Transposition_Table

behold the chess fancy terms

[clear sedge]

if you're storing more than 16 bytes of data i'd consider changing your auto value = foo->second; to a auto &value = foo->second for speed reasons

(i pulled 16 bytes out my ass, very arbitrary limit)

[midnight sigil]

my transposition table is {depth, nodeType, eval, bound}, if you want as reference

[obsidian mantle]

i have no idea what to do with that additional information

do i just rebuild the hashed situation

with their alpha,beta, depth, whatver

i thought im supposed to happily return after hash was found

[midnight sigil]

wuh

[obsidian mantle]

the idea is to extend this search in this branch?

store where we stopped?

[midnight sigil]

just read the wiki I dunno how to explain

read Sebastian Legue's vlog and bot source code for reference

perhaps Boychesser as well

holy molly it's 1:32am already

I have school tmr at 7am

cooked

gn programming

[obsidian mantle]

you know all that and you are only in school wtf

i'll never catch up like this

[trim valve]

i love barely understanding the way stuff gets built

it totally does not case me to run in circles

[rigid snow]

it cases you

[trim valve]

I have a .a library, inside are a bunch of .o files
how in the everloving hell can I decompile those?? fwiw ```
$ file main.o
main.o: LLVM IR bitcode

because google is sending me in circles, I couldn't get anything to link with llvm earlier

[opaque sigil]

iirc it was llvm-dis to go from bitcode to .ll

and then idk from there

there was/is an llvm c backend huh, interesting

[clear sedge]

.ll is probably gonna be your final stop, you can't gleam any more meaningful information from just a bytecode file

[trim valve]

I have ida so if I can throw it there that'd be great

[opaque sigil]

https://github.com/JuliaHubOSS/llvm-cbe there's this, wonder how well it works though

[rough bloom]

should be able to convert to a text representation if you have roughly the correct version of LLVM at least
dunno about anything resembling proper decompilation

[clear sedge]

you could rewrite it in c in a way that mimics the outputted .ll

generally, a good chunk of information gets lost in translation, so yeah you aren't gonna get a c file that easily :3

[fringe heart]

greetings people, could you give me some information about how vedal made such a fast responding ai with personality, i have some projects of my own which could greatly benefit from that information

[trim valve]

I'm not expecting a perfectly 1:1 c file

I'm just hoping that I don't have to learn to read raw llvm ir as well

also it looks like this has full debug symbols, so I really should be able to get the source C file back

AVFormatContext oh god oh fuck why is there ffmpeg

[clear sedge]

https://github.com/NotDec/NotDec-llvm2c

i found this

[fringe heart]

rain, do you perhaps have some knowledge about the inner workings of neurosama?

[clear sedge]

no, i don't spend time researching llms personally

i'm only here cuz i think the tutel is entertaining

[opaque sigil]

can confirm this works btw, it's not the prettiest c but it's better than having to read IR (needs llvm 19)

[trim valve]

no windows build instructions

[clear sedge]

[opaque sigil]

it only needs llvm so i doubt it can be too bad

[trim valve]

C:\JL\pi32\bin\llvm-objdump.exe: '.\main.o': The file was not recognized as a valid object file

thank you objdump

very reassuring

[fringe heart]

do you know someone who has more information about what i am looking for?

[opaque sigil]

consider the inner workings of neuro a trade secret

[clear sedge]

idk, email vedal if you feel up to it, he has his email up on his site
i doubt you'll get much of an answer though

[opaque sigil]

it said it's ir bitcode so checks out

[trim valve]

will try in a bit

have been mildly hyperfocused on this for like 2 hours and mentally dying at this point

[opaque sigil]

[trim valve]

where has 400gb of disk space gone

[rare bramble]

trying to solve why my twitch tab takes up 10 GIGABYTES of ram, but it's so much ram, that the debug tools and the tab crash when trying to check the heap

[rigid snow]

ah were

[trim valve]

im going to fucking explode

why computer hate me

[olive sable]

a

[nocturne olive]

7TV? I know it leaks memory on Windows at least

[rare bramble]

might very well be, it's just a little confusing since it would be integrated into the twitch process

[nocturne olive]

Extensions run within their own space

[rigid snow]

it could be the script that embeds into twitch

that's leaking memory

[rare bramble]

heap shapshots are pretty useless, but allocation sampling gave me this:

[trim valve]

explosion imminent

i should sleep

this doesn't feel like it should be as hard as I'm making it

I havve a .a file with the library

all I should need to do it make that into something ida can read

which should just be building an executable that includes it

[trim valve]

ok so

half my issue

is that this library's build tools use clang 4.2.1

so presumably that's the version that the included libraries are

[glad path]

just yell at it until it works

usually fixes stuff for me

/srs

[trim valve]

surely putting all of clang into ida is a good idea

because i can't figure out how to list targets on that version

[rigid snow]

fksdjflsdkjfldksjf

[glad path]

make sure to tell it if it does not work you will be very very sad

and that it should work

it does wonders

-# i have no clue why that helps but it does somehow

-# probably wont do anything for anyone else smh

[olive sable]

tried to take a screenshot, whatever this is happened instead

[obsidian mantle]

wait so i just got a situation
where both sides had 1 king and 1 knight

when one of knights got taken, game ends by insufficient mating material.
??? so while enemy had knight, it was possible to mate him, but when he lost it, it became impossible

[olive sable]

lmao discord cant even embed it

its jsut 2555x2 pixels

[rigid snow]

[olive sable]

how did it go from 2555x2 to 550x0 on discords end?

[rigid snow]

550 is just the width it's rendered as

[olive sable]

ah

[rigid snow]

because that's how wide my chat is

https://cdn.discordapp.com/attachments/1071784467036913664/1424518973407363072/screenshot-2025-10-06_000814.png?ex=68e43e2d&is=68e2ecad&hm=b58da24e689ecc620fb8903896299db53f77d7c63a5d465e7f04ff4fdddbac84&

it embeds fine on mobile interestingly enough

[olive sable]

it basicly just took a 2 pixel wide slice of my screen

[olive sable]

Not for me

[rigid snow]

if this could be considered “fine”. i can’t tap it

[olive sable]

oh

thats not visible on my end

might be a discord ios vs android thing? assuming you're on ios

[rigid snow]

i am. should be the same app tho

[olive sable]

[obsidian mantle]

its hard to navigate in github if you are not using it constantly

or havent used it much

[rigid snow]

make an EXE file and give it to me. STUPID FUCKING SMELLY NERDS

[faint sandal]

how do they know how to compile but not navigate github 😭

[obsidian mantle]

well you dont need github to compile

sometimes people dont build an exe indeed

so you actually have to compile
setting up the ide, dependencies and all

install 50 compilers and libraries for the compilers

i can see most people just not doing it and stopping there

[obsidian mantle]

it makes sense if you actually want people to be "worthy" to use your program

[faint sandal]

but they bothered to set up the building env for it

[obsidian mantle]

which didnt work

you can install visual studio, throw some project in it and just build with random compiler

which will give bunch of errors and it seems where they stopped

[faint sandal]

also i think they tried to build with msb

[obsidian mantle]

is it random enough

so i think my transpositions work but
it gives no performance boost at all
i only use depth+value though

and i have no iterative search maybe they are connected

[midnight sigil]

you messed with it for 4 hours

[gritty dust]

Yoooo guys

Guess what

I got a raspberry pi 5 16gb

[midnight sigil]

now you have a vps

[gritty dust]

lol true ig

[trim valve]

what if

I just build clang 4.2 myself

surely the .o files will read perfectly with no issues whatsoever

(curse whoever removed targets other than their own)

[obsidian mantle]

i havent
it was 30 min max before i wrote that

tbh i worked on the bot for more than 4 hours today

[jagged turtle]

you should've already

[faint sandal]

already

[obsidian mantle]

people just want "download button"

[jagged turtle]

[DOWNLOAD HERE] [SOURCE CODE]

The end user will still click source code lmfao

[obsidian mantle]

because its highlighted with green

and release is written in thin font somewhere bottom right

and release != download so isnt eye catching

[jagged turtle]

oh no I meant on like a dedicated page

using smth like github pages

[obsidian mantle]

you see you click big green button (code)

and it has download zip

"must be download"

thats the problem

[jagged turtle]

???

hello

please read what I just said

although I will agree that among certain groups of end users critical thinking is lacking

[obsidian mantle]

wtf is "github pages"

[clear sedge]

static web pages hosted on github

[jagged turtle]

https://github.io the landing page wont stick but you can see it for like half a second

oh ok cool embed works too

[obsidian mantle]

but when you search for program you get on this page

[jagged turtle]

yeah because the program has been on github for idfk how long

if you make it the github page's problem then people will start linking to that more

[obsidian mantle]

does anyone uses those "github pages" even

[jagged turtle]

yk, assuming you actually mention it

[obsidian mantle]

first time i hear about it

[jagged turtle]

You are sorely lacking in knowledge then

[obsidian mantle]

i am

[jagged turtle]

checks out

considering web frameworks actively have a deployment guide for github pages I'm sure some people use it

anyways

it's not usable for anything requiring a server

or well not super usable

so that's why most people only use it for personal stuff + documentation

and project landing pages

yk, not very interesting stuff interactivity-wise

though ig some games have been hosted there

[obsidian mantle]

oh so i might have not noticed it being github at all

when i landed on such page

because it looks nothing like github

well

nobody notices that either

i mean people who are not using/used github as active developers

so they probably find that fancy "download" button and have no idea it was github

[obsidian mantle]

when we say github we think about that

[clear sedge]

i'm personally moving off github and onto codeberg

github's philosophy no longer aligns with mine

[obsidian mantle]

how would i know about "github pages" then
its not what they teach you in school

now i know

[clear sedge]

github pages are a github thing instead of being a git thing so it makes sense to skip talking about it in vcs lessons

[obsidian mantle]

what lessons

[clear sedge]

vcs = version control system

it's what git is

[rigid snow]

funnily enough we DID get taught to deploy to github pages as part of our git intro

[clear sedge]

that is odd

[rigid snow]

it's not it was part of a web dev course basically

so it does make sense

[clear sedge]

makes sense i guess ¯_(ツ)_/¯

[midnight sigil]

[clear sedge]

this is why you package your python into an exe

https://tenor.com/view/big-brain-markiplier-gif-14835823

[midnight sigil]

speedy python

[clear sedge]

speedy python ❌
python on speed ✅

[jagged turtle]

the suspicious looking github.io in the browser url

[clear sedge]

github pages allow you to provide your own domain

[jagged turtle]

ik

but at least you would know if it says *.github.io in the browser bar

[rigid snow]

you can host your frontend from gh pages. if you need a backend just do that using a separate service

it will still be .github.io but with interactivity

[jagged turtle]

vercel for maximum cost mhm

[jagged turtle]

that's why I said it's not super practical

[faint sandal]

https://azaka.fun/ and basically all websites/blogs I have are all just GitHub pages

gh pages can be used to host many sophisticated sites long as it's all statically built

[jagged turtle]

yeah

[faint sandal]

you can slam an entire blazor project on there if you wanted

[jagged turtle]

I used to use github pages for everything but when I got custom domain I was annoyed that fronting it with cloudflare would just cause SSL issues so I eventually just moved

not necessarily github's fault

[rigid snow]

blazor

[jagged turtle]

or cloudflare's

but I just was super annoyed at that constantly being an issue

[rigid snow]

just don't proxy

[faint sandal]

that's because you had a full TLS chain

[jagged turtle]

I'd rather still proxy so I can set up redirect rules on subpages

don't worry it's the jankiest setup you will ever see

[rigid snow]

just use cf pages atp

[faint sandal]

azaka.fun is proxied and has multiple page rules

[jagged turtle]

where do you think I moved lol

[rigid snow]

[jagged turtle]

oh well

I just moved anyways

bonus is that if I ever switch git hosts I don't have to go rewrite all the links that used to use github.io

I think I need to stop using disposables

you guys ever get a thought that you wanted to write down but when you write it down you don't remember what it is anymore?

[silent cloak]

https://tenor.com/view/python-devs-developers-dev-developer-gif-18258549504742825905

[olive sable]

https://tenor.com/view/kasane-teto-luigi-hatsune-miku-miku-teto-gif-14822580778231932947

[torpid coral]

I have a hard choice:
Go to sleep (it's 3 am, get up at 6 am )
or
Using mystical night enlightenment and start programming

What to choose...

[stray dragon]

had that choice last night, decided to go for the second option, got out of troubleshooting hell
sleep is a mistake

[torpid coral]

I hope this power will help me

[amber fractal]

Man this is hard, I can't find her at all

[olive sable]

me when im colourblind

but ye i can see her lol

[glad path]

thoughts on scp?

am i the only one who has ever used it

surely not

[rigid snow]

i’m having trouble sleeping again fuck

[glad path]

just dont do discord

[rigid snow]

i came onto here because i’m having trouble sleeping not the other way around

[opaque sigil]

rsync and sftp exist

[glad path]

scp works with ssh keys tho

i thiink

the other prolyl do to ngl

[opaque sigil]

they do indeed

it all runs over ssh

[olive sable]

the scp foundation uses ssh keys

[opaque sigil]

there's nothing inherently wrong with scp necessarily, we just have better tools for most use cases nowadays

[glad path]

me when the

[olive sable]

skybox

im goign to bed now

this took way too logn to debug when i just had a multiplication order issue in the fuckign shader

[thick thicket]

[patent walrus]

blender is gonna make me crash out

like why is this texture black while having the correct texture in the icon

even in the paint window it's just fully black, like where did my texture go???? i made it by hand in this window earlier today but blender voided it

[stray dragon]

bad news: my bot has some rather bad memory leak issues
good news: windows can clean up the memory leaks just fine (although it does stress the cpu a bit to clear 50gb of leaked memory all at once)
guess i can safely ignore that problem for now

[obsidian mantle]

Alpha? Or something like that

Some layer blocking everything

[uneven pulsar]

currently learning arrays

[opaque sigil]

[midnight sigil]

[opaque sigil]

Btw pop returns the element that was removed not the new array

[uneven pulsar]

yeah i just learned

also the strategy i was doing didn't work where i would just "numbers.push(1).shift();"

so i have to do it separately

[opaque sigil]

Mhm

I don't remember what push returns actually

[uneven pulsar]

it adds another thing into the array

at the end

[opaque sigil]

I mean yeah

[uneven pulsar]

also shift somehow removes the extra numbers

[opaque sigil]

the js array methods are kind of weird like that, you'll get used to it

Apparently push returns the new length hmmm

[uneven pulsar]

it is what it is

[uneven pulsar]

wow

i did numbers.pop().reverse();

and it didnt work

[opaque sigil]

You're trying to reverse a number

[uneven pulsar]

no

[opaque sigil]

Yes

[uneven pulsar]

i am trying to remove the first number

[opaque sigil]

That's what you're doing

[uneven pulsar]

instead of the last

[uneven pulsar]

nah

[opaque sigil]

Does shift not work

[uneven pulsar]

i did use it

wait shift removed the first number

nvm

[opaque sigil]

[uneven pulsar]

okay idk how this works anymore

[opaque sigil]

It shifts everything by one

[uneven pulsar]

its somehow different with words

[opaque sigil]

wdym

[uneven pulsar]

i dont even know what i am saying

im just gonna fully focus and get the results

my brain cant do discord and coding

[opaque sigil]

Good luck

[uneven pulsar]

thanks!

[opaque sigil]

When in doubt mdn is your best friend

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array

[uneven pulsar]

oh, i see

i will try if confused