programming

[real sierra]

grab new

new .h as well just now

[trim valve]

those work

[sage crag]

shirpo do you want hblang bindings

[real sierra]

token counter doesnt support hblang in reality

so prob cant use

[sage crag]

token counter PR

needs fix

[real sierra]

token counter does need a fix

last updated 4 years ago

trying to update any of the dependencies is hell tho

tried

[trim valve]

(gdb) bt
#0  __syscall_cancel_arch () at ../sysdeps/unix/sysv/linux/x86_64/syscall_cancel.S:56
#1  0x000079721bca00ec in __internal_syscall_cancel (a1=a1@entry=105616026212552, a2=<optimized out>, a3=a3@entry=0, a4=a4@entry=0, a5=a5@entry=0, a6=a6@entry=4294967295, nr=202) at ./nptl/cancellation.c:49
#2  0x000079721bca0847 in __futex_abstimed_wait_common64 (private=<optimized out>, futex_word=0x600ea5b998c8, expected=0, op=<optimized out>, abstime=0x0, cancel=true) at ./nptl/futex-internal.c:57
#3  __futex_abstimed_wait_common (futex_word=0x600ea5b998c8, expected=0, clockid=0, abstime=0x0, private=<optimized out>, cancel=true) at ./nptl/futex-internal.c:87
#4  __GI___futex_abstimed_wait_cancelable64 (futex_word=futex_word@entry=0x600ea5b998c8, expected=expected@entry=0, clockid=clockid@entry=0, abstime=abstime@entry=0x0, private=<optimized out>) at ./nptl/futex-internal.c:139
#5  0x000079721bcac5ff in do_futex_wait (sem=sem@entry=0x600ea5b998c8, abstime=0x0, clockid=0) at ./nptl/sem_waitcommon.c:111
#6  0x000079721bcac698 in __new_sem_wait_slow64 (sem=0x600ea5b998c8, abstime=0x0, clockid=0) at ./nptl/sem_waitcommon.c:183
#7  0x0000600e713cd878 in start_chess_api () at /home/owobred/neuro/other/chess-tournament/target/debug/build/chess-api-8324e014b9d4855d/out/chessapi.c:1544
#8  0x0000600e713cdaa3 in chess_get_board () at /home/owobred/neuro/other/chess-tournament/target/debug/build/chess-api-8324e014b9d4855d/out/chessapi.c:1601
#9  0x0000600e713867e7 in chess_api::Board::get () at chess-api/src/lib.rs:285
#10 0x0000600e71386068 in chess_bot::main () at chess-bot/src/main.rs:7

200iq, I debugged it after it stalled

[sage crag]

sem_waitcommon

!nothing

[stiff mica]

Nothing ever happens, all-in

[trim valve]

@real sierra rust bindings have been thrown into https://github.com/owobred/chess-competition-rust-bindings

unless I have a major skill issue you should just need to run cargo run in the root directory and the bot will be ran

[tender river]

[trim valve]

(or well, not ran)

beware rust people, I have never made bindings before

so it will be scuffed

[sage crag]

are these enums

correctly typed

[trim valve]

hm?

[sage crag]

no repr C?

[trim valve]

(keep in mind I'm currently bringing in my own copy of the api because I am too lazy to fetch it from shrios repo)

[sage crag]

[trim valve]

wait I might be stupid on that front

[sage crag]

ye leaked the bindings

[tender river]

its fine

it uses bindgen

[trim valve]

ye those aren't the ones that go through ffi

[tender river]

[sage crag]

c supposed to be private still

[tender river]

shiro use more macros

[trim valve]

smh

just like

don't sleep all night so you don't remember

[sage crag]

pls use macro

[trim valve]

or I can actively un-leak and neither of you can read the repo

:3

[tender river]

macros

[sage crag]

shimo bitboard.c could be 30% of the size if you use macro

[tender river]

you havent seen this

[trim valve]

what have I done

[tender river]

theres like so many ifs with just this

[rigid snow]

too late already cloned, backed up on 30 bulletproof HDDs on-site, on 30 tape spools off-site and uploaded onto 3 high orbit satellites

[trim valve]

shrio can you just prematurely execute me

[tender river]

[sage crag]

binding

[sage crag]

nyamacros

part of me wants these to be macros too but it doesnt do anything

ye my chess bot now

[trim valve]

konii about the back up the entire chess bot api in screenshots before shiro wakes up

[sage crag]

its ok i already memorised the whole api

@tender river

[tender river]

no codegen

[trim valve]

yeah ok let board = chess_api::Board::get(); is hanging forever

[tender river]

honestly why even have that

Board* is a singleton

so it shouldnt be passed via function arguments at all

functions should just access the global Board* by themselves

hiro can use that function internally if he wants to

[trim valve]

the thing that confuses me is you can free the board

[sage crag]

[trim valve]

#programming message

also I'm mirroring the python api atm

[sage crag]

ye ye but

[tender river]

i would fix the api but

[trim valve]

imo it makes sense if you read the comments

[sage crag]

the api global

[trim valve]

but also yeah its odd

[sage crag]

its ok its just a beta api

[stark needle]

first time i see this

it's an apple llm

[sage crag]

somehow i think it may be apple intelligence

[stark needle]

can't wait for apple ai slop on the iphone

[real sierra]

so are they working?

[trim valve]

ok so fwiw shiro, my threads are stuck on these lines https://pastes.dev/uvwp3eRU3L

[trim valve]

only under a debugger

[real sierra]

[rigid snow]

don't forget there are two models promised, a small one for local and large on cloud

[real sierra]

that shouldnt be possible

[tight sparrow]

programming can i borrow shiro

its important

[trim valve]

i wish shiro

[real sierra]

ok guest additions finally installed on vm

now i can copy paste from it to here

but also wtf has it done to my res

480p screen

[trim valve]

iirc it means you can resize the window

or something

instead of having to manually change res

[real sierra]

yes

this is so much better now

[trim valve]

feel free to test whether or not it breaks for you in the vm

[sage crag]

pls

macro not dangerous macro your friend

[real sierra]

macro danger

[sage crag]

https://cdn.discordapp.com/attachments/1300807614032121950/1409945130810609695/3dgifmaker99373.gif

[trim valve]

marco

[real sierra]

copy paste my friend

[tender river]

shiro cant use macro used to java

[real sierra]

^

bugfix first then code quality

[sage crag]

macro is a micronutrient

[real sierra]

working ugly better than pretty broken

[tender river]

[sage crag]

prevent code duplication prevent stupid mistake

[tender river]

its easier to bugfix deduplicated code

[real sierra]

i dont make stupid mistake

i too good at coding

[sage crag]

[tender river]

[sour mason]

idk if anyone can help, but I‘ve been trying to tokenize a tfds dataset recently but the tokenization functions from tfds and tf dont seem to be available? did they get removed in a later version or smth? anyone knows?

[real sierra]

can i get a call trace here

or

[small light]

why are you nerds such nerds

[trim valve]

that pastes.dev link is a trace

or do you want more

[tender river]

me not nerd

[sour mason]

huh?

[tender river]

haent programed in days

[sour mason]

gotta make a living somehow yk?

[real sierra]

oh it is

i didnt realize

i have concluded this makes no sense

the mutex is blocking despite it being the first time it hits it

[tender river]

[sage crag]

mal (mirw) (mweow) (cat enthusiast)

[real sierra]

who locked my mutex

[sage crag]

sory

not cat myself

this my domain

https://cdn.discordapp.com/emojis/1212937424062119967.gif?size=48&animated=true&name=nuhuh~1

[trim valve]

fwiw shiro I'm just sending ```
uci
isready
ucinewgame
position startpos moves f2f4
go wtime 300000 btime 300000 winc 0 binc 0

[real sierra]

.....

might've found the bug

i could've sworn i had this

[sage crag]

[real sierra]

[sage crag]

uninitialised value

use-before-init

[tender river]

use after init

[pulsar edge]

hi shibo

damn he left

[sage crag]

[pulsar edge]

hi konii

[real sierra]

@trim valve pushed new

[pulsar edge]

hi shibo

[real sierra]

hi scrub

[pulsar edge]

how long is the tourney gonna run for

[tender river]

[real sierra]

a month is the plan

[pulsar edge]

ic

hi chayleaf

[sage crag]

1 month

more than enough time for me to make an abomination

[pulsar edge]

shiro how are your commits this wrong

[real sierra]

wdym this wrong

[pulsar edge]

your commit message says you added a function to the chess api

but it doesn't actually show up until 3 commits later

[real sierra]

see next commit

[real sierra]

i didnt look at my file explorer right

i just pushed chessapi.c

but it was the old copy sitting in the python directory

[pulsar edge]

can I criticize your commit messages

[real sierra]

im the only one looking at them

not meant for understanding by mortals

ls

[pulsar edge]

it would be nice if C changes were C: <message> and the python changes were Py: <message>

[real sierra]

this isnt the terminal

[real sierra]

?

[sage crag]

shrmo give me push access

[real sierra]

the C and python files are totally disjoint

[pulsar edge]

thats my point

[real sierra]

oh wait

no i did include two in a commit

oopie

[real sierra]

[trim valve]

[pulsar edge]

prefix the commit message to make it easier to tell what a commit is modifying

[real sierra]

bred officially the first person to make a chess bot other than me

[real sierra]

will try to do in future

im experimenting with doxygen rn but realizing it might not support every language

i dont wanna use it for only half the docs

[sage crag]

https://tenor.com/view/mini-miniko-minikomew-gif-17857450978747562560

but im purely benevolent

[tender river]

i fully support this

shiro can't keep getting away with

[pulsar edge]

idk it supports C, C++, and C# and thats all you really need

[tender river]

[sage crag]

cant keep getting away

we will catch

[real sierra]

[real sierra]

hows your C# impl going btw

[pulsar edge]

uh I had class this morning

[real sierra]

semi important change for you, a chess_free_moves_array func was added

so you dont need to import libc for free anymore

[pulsar edge]

god I'm gonna have to write a cmake file for the C# impl

eck

[sage crag]

ee gonna steal your code lines

[tender river]

[pulsar edge]

unless you're fine with manually compiling the C dll before compiling the C# library

[real sierra]

doxygen default config has helpful comments

almost too many

thing is 3000 lines with the comments

[sage crag]

i would fix shiro api but im too busy not fixing shiro api

[pulsar edge]

@real sierra

[real sierra]

what mean

id rather people dont need to download a dll

[pulsar edge]

the C# impl needs to link against the C dll

[real sierra]

doesnt sound very portable

[trim valve]

2025-09-15 23:33:30.732<--1:thread 'main' (109966) panicked at chess-api/src/lib.rs:364:18:
2025-09-15 23:33:30.732<--1:unsafe precondition(s) violated: slice::from_raw_parts requires the pointer to be aligned and non-null, and the total size of the slice not to exceed `isize::MAX`
2025-09-15 23:33:30.733<--1:This indicates a bug in the program. This Undefined Behavior check is optional, and cannot be relied on for safety.
2025-09-15 23:33:30.733<--1:note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
2025-09-15 23:33:30.733<--1:BytesRead > 0
2025-09-15 23:33:30.733<--1:n > 0
2025-09-15 23:33:30.733<--1:-- LiesThread.Execute terminated --
2025-09-15 23:33:30.733<--1:thread caused non-unwinding panic. aborting.

I may

have a bug

[real sierra]

that one

is not me i think

[trim valve]

also unrelated but ```
2025-09-15 23:38:52.9791Start calc, move no: 200
2025-09-15 23:38:52.980-->1:position startpos moves e2e4 a7a5 g2g3 g8f6 f1b5 c7c5 h2h4 h8g8 b5a6 g7g6 d1h5 g8h8 h5d5 b7a6 d5c6 a8a7 g1e2 a7c7 e1f1 a5a4 f1e1 h7h6 h1h2 c7c6 d2d3 h8g8 g3g4 d8a5 e2c3 c6e6 e1f1 f8g7 b1d2 c8b7 b2b4 a5b5 c3e2 f6e4 c1a3 g8f8 a1c1 e4f6 e2f4 b5c6 f4e6 c6h1 f1e2 f6g8 h2h1 f7f5 e6c7 e8f7 c1f1 d7d6 d2b3 b7e4 f1a1 e4f3 e2f1 g7f6 c7e6 f3h1 b4b5 c5c4 g4f5 h1e4 e6f4 e4h1 a3b2 a4b3 b2f6 b3b2 f4g6 b2a1q f1e2 d6d5 f2f3 a1e1 e2e1 e7e5 f3f4 h1g2 f6e7 h6h5 e7f8 a6b5 e1d1 g2e4 f8g7 f7e8 c2c3 e8d8 d1c1 d5d4 a2a3 e5f4 c1d2 d8c7 g7d4 e4f5 d4f6 b8c6 f6d4 c6e5 d4a7 f5c8 d2e2 c8g4 e2d2 c7c8 g6e5 c8d8 a7b8 g8h6 e5d7 d8d7 b8e5 h6f7 d2c1 f7h8 c1d2 b5b4 d2e1 f4f3 e1d1 b4a3 e5h8 d7c8 h8d4 c4d3 d4b6 c8b8 b6c5 g4h3 c5b6 h3c8 c3c4 c8f5 c4c5 b8c8 c5c6 f5g4 b6a5 a3a2 c6c7 a2a1n d1e1 g4f5 a5d2 a1c2 e1f1 c2a1 f1e1 f5h7 d2c3 h7g8 c3d2 a1c2 e1d1 c2e1 d1c1 c8d7 c7c8r d7d6 c8b8 d6e7 d2c3 g8a2 c3d2 a2c4 b8b7 e7e8 c1d1 c4b5 d2g5 b5a6 b7b6 f3f2 g5e7 f2f1b d1c1 d3d2 c1d2 a6d3 d2c1 e8e7 b6b4 f1h3 b4e4 e7d7 e4e5 d7c8 c1d1 e1c2 e5e3 h3e6 e3d3 c2a3 d3b3 e6g4
2025-09-15 23:38:52.980-->1:go wtime 300000 btime 300000 winc 0 binc 0
2025-09-15 23:38:52.981<--1:BoardMoves::as_slice {start: 0x5d4872f4d640 len: 16}
2025-09-15 23:38:52.981<--1:info currmove a3b1
2025-09-15 23:38:52.981<--1:bestmove a3b1
2025-09-15 23:38:52.9811---------> Arena:Illegal move!: "bestmove a3b1" ()

this happened

[real sierra]

that might be my fault

ill have a look

[trim valve]

don't worry it very helpfully has the move list

(2k5/8/8/7p/6bP/nR6/8/3K4 w - - 3 101 is the FEN)

currently hunting for the crash again

[real sierra]

adding macros to bitboard.c where reasonable

[tender river]

[real sierra]

i dont suppose you can nest macros

[tender river]

you can

[real sierra]

pogs

[tender river]

macros can call macros

including dependency injection

when a macro is expanded it temporarily gets disabled

as in, a macro cant be expanded in its own expansion

thats basically the only limitation

also you should use ident concatenation in some cases

[real sierra]

gotta admit

this is a bit smaller

[tender river]

you can pass all 4 directions from all_dirs to each invocation

so the macro can access diametrically opposing directions

or 8 i guess

[real sierra]

the bb slide commands still have to be custom tho

since they're tuned per direction

not copy-paste

uh

should i have put the macros in the header

[sage crag]

yes shimo

[tender river]

thats why i recommended you to pass all 8 directions to each instantiation

[real sierra]

[real sierra]

wdym

[sage crag]

noe noe macros accept all sort of thing

including symbol

[real sierra]

no all 8 of these functions are different

[tender river]

#define all_dirs(x)
x(n, e, s, w)
x(e, s, w, n)
x(s, w, n, e)
x(w, n, e, s)

[real sierra]

wtf is that

[opaque sigil]

Thus the macrofication begins

[real sierra]

why would i do that

[sage crag]

#define DO_THING(a, sym, b) ((a) sym (b))

DO_THING(1, +, 2)
DO_THING(1, *, 2)

[opaque sigil]

They're called X macros and they're great

[sage crag]

ye x macros very useful

[real sierra]

oh i see

[sage crag]

i use in vm

[real sierra]

i did opposite

or same?

[tender river]

now pass all dirs to each submacro()

[real sierra]

wat why

no need

[tender river]

so you can access diametrically opposing dirs

[opaque sigil]

The idea is that you have a list of things with the assumption that there's a macro
And then you just define said macro to be whatever you need

[tender river]

so you can rotate the dir by an arbitrary degree in the macro code

[real sierra]

nope im totally lost

[sage crag]

you shimo

not lost

[real sierra]

trying to get me to use more macros when i already finished macroifying everything

[sage crag]

awa

[sage crag]

nop more macros always usable

[opaque sigil]

Just be careful with macros shiro since they don't have symbols

[sage crag]

use more macros to use less macros

[real sierra]

all trolling

[sage crag]

ye it direct code subtitution

[real sierra]

removing all macros

[sage crag]

[tender river]

#define flip(n, ne, e, se, s, sw, w, nw) case dir_ ## n: return dir_ ## s;
switch (dir) {
_all_dirs(flip)
}

[sage crag]

use computed goto

(not do that)

[tender river]

[sage crag]

[real sierra]

what does this even do

[sage crag]

flip

[tender river]

flips the direction assuming _all_dirs passes all 8 dirs to x

[sage crag]

[real sierra]

it doesnt

and shouldn't

[tender river]

you can make it

[real sierra]

why would i do that

[tender river]

so you can dedupe this

[real sierra]

you dont seem to get it

[real sierra]

wait

maybe you do get it

no you dont

im sure you can find some sequence of 3 or 4 macros that generates this but why

[tender river]

ye cant dedupe that

[real sierra]

its just wasting my time

yea this is what i meant by "tuned per function" sorry if i wasnt clear

i didnt mean copy paste and swap a letter

i meant they're actually different

anyway

feel free to figure out which macros make that

[sage crag]

https://cdn.discordapp.com/attachments/1067638175478071307/1406337271241904158/ezgif-3f552200e37817.avif

[real sierra]

probably one for ns modifications, one for ew modifications

[tender river]

i was talking about code like this

[real sierra]

i see now

that makes sense

i still prob wont do that

im more concerned with code readability than macro shortness

[real sierra]

i confuse

[trim valve]

damn tournaments are funny

[real sierra]

do i have separate defines in the header

to just generate declarations

or can i leave those

[tender river]

dont put them in the header if they arent part of the interface

if they're used for definitions but not declarations

[real sierra]

i mean for this

[tender river]

honestly i'd either convert direction to enum or just keep it as is

[real sierra]

keeping

direction as enum no thx

[tender river]

better to keep interfaces explicit

[real sierra]

agree

[tender river]

would you rather direction as 8 functions for each operation

[real sierra]

yes

im not deciding which direction on the fly

it also faster without enum

[sage crag]

then macro

[real sierra]

so if i define _all_dirs macro in bitboard.c

and chessapi.c includes bitboard.h

does it have access to it

[trim valve]

[real sierra]

or must redefine

[tender river]

[real sierra]

are there any legal moves in this position

[sage crag]

it wont be visible by #include "a.h" though

i think

[tender river]

i might have misread

[real sierra]

if there's 0 moves and arena doesnt turn off the chess bot then it will try to dereference the first element of get_legal_moves and segfault

expected behaviour

not worth the tokens to avoid

[tender river]

each c file is completely separate

each header is just code copypasted to include sites

[real sierra]

i was worried about that

worried about the opposite actually

keeps my defines separate which is nice

ok

time to macro nuke the chess api

[trim valve]

https://lichess.org/analysis/standard/6N1/6R1/7P/p1kN4/P7/2p2p2/4p3/2Bn1K2_w_-_-_0_97

several

[real sierra]

keep note of all these bugs bred will fix in a min

[trim valve]

yeah

about to head to bed soon so like

I'll try and find an example

[real sierra]

[stark needle]

[trim valve]

uci
isready
ucinewgame
position startpos moves e2e3 g7g6 g2g4 h7h5 c2c3 a7a6 c3c4 f8g7 f1h3 d7d5 h3g2 g7f8 d1c2 h5h4 c2d3 c8f5 g1e2 f7f6 h2h3 a8a7 g2e4 a6a5 f2f3 e8d7 e1d1 b7b5 e2g3 e7e5 d3e2 h8h7 d1e1 f8e7 h1g1 f5g4 h3g4 b5c4 g1f1 e7a3 f1f2 c7c5 f3f4 d8b6 e2c4 h7h8 b1a3 h8h5 f2g2 b6b7 a1b1 d5c4 g3f5 b7c8 b2b4 c8c6 g4h5 a7a6 e1f1 c6d6 g2g6 d7d8 e4c6 b8c6 d2d3 d6e7 a3b5 e7g7 c1d2 e5e4 b1b2 c6b4 b5c7 b4d3 b2b7 d3e1 g6h6 c4c3 f5g7 d8d7 b7b1 a6d6 d2c1 h4h3 b1a1 d7c8 c1b2 e1d3 h6h7 d3f4 b2c1 d6d4 c7b5 f4e2 f1f2 d4d5 f2f1 e2f4 a1b1 c5c4 f1g1 c8d7 h7h6 f4g2 g7e6 d5d3 h6g6 d7e8 g6g5 g8e7 g5g3 g2h4 g1f2 h4g2 e6f8 d3d4 a2a4 d4d3 b5d6 e8d8 d6c8 d3d2 f2f1 d2e2 b1b4 g2e1 g3g1 e1d3 c1b2 e2c2 g1g8 c2c1 b2c1 d8c7 c8d6 e7g6 g8g6 c7d8 d6e8 c3c2 g6g1 d3f2 b4b2 d8e7 e8g7 f6f5 b2b1 f2d1 g1g6 h3h2 g6g1 d1b2 b1a1 e7f6 g7e8 f6f7 e8g7 c4c3 f8d7 b2c4 c1b2 c2c1r b2c1 h2h1r g7e6 f7e8 c1a3 f5f4 a1d1 c4e3 f1e1 f4f3 d7f6 e8f7 e6c7 h1h2 a3c1 h2c2 f6g8 e3d1 c7d5 f7e6 e1f1 c2h2 h5h6 e4e3 g1g3 e6d6 g3g7 h2g2 g7g2 d6c5 g2g7 e3e2
go wtime 299992 btime 300000 winc 0 binc 0

this crashes the C bot

[real sierra]

ill look into it

[trim valve]

if I had to guess its failing to parse the board or something

oooh wait are you assuming a max line length

[real sierra]

perhaps

does that happen to be around

256 chars

or 1024 maybe

idk what the buffer was

[tender river]

shiro keep it

[real sierra]

ill just make the buffer 1GB

[trim valve]

shruge

[real sierra]

find a chess game that long

[real sierra]

if that doesnt count whitespace

its probably just over 1024

but it should count whitespace so

idk

[opaque sigil]

784 chars (without whitespace)

[trim valve]

oh wait if I copy in the go command with it then its 1028

[real sierra]

[opaque sigil]

wait no

[real sierra]

"surely no game will exceed this! its a whole kb!"

[opaque sigil]

so stingy with the buffers smh

[real sierra]

holy

[trim valve]

tbh I only hit this when pitting two random bots against each other

[real sierra]

honestly impressive

they never hit the 50 move draw?

or the threefold repetition?

ive had my own bots fight lots of times without hitting this issue

[tender river]

[real sierra]

anyway

ill bump it up to 4096

[tender river]

i wonder whats the longest game you could make

[tender river]

is it autoclaimed?

[real sierra]

should be

[tender river]

[real sierra]

50 moves since moving a pawn or capturing a piece is the condition

autoclaimed by Arena*

not the api

api will just not tell the bots

[trim valve]

after increasing the buffer

[real sierra]

how do i macro any of this

this is a problem for later

this is a nightmare to macro

[opaque sigil]

if you can't immediately think of a way you probably just shouldn't

[tender river]

so ~7100 moves is the theoretical maximum

[trim valve]

ok

[real sierra]

the code is just so specialized for performance

[sage crag]

erm shimo

[trim valve]

slight issue

[sage crag]

why not dynamic size

[real sierra]

dynamic size curzed

[trim valve]

arena segfaulted

[real sierra]

ARENA SEGFAULTED?

wtf did you do

[trim valve]

i uh

[tender river]

[trim valve]

just had a tournament going between two of my bots

[opaque sigil]

try passing utf-8

actually

isn't it all ascii

[trim valve]

they were at like 480 moves last I checked

[opaque sigil]

how tf do you run into utf-8 issues

[trim valve]

I wonder if arena has a funny hardcoded limit of 512 moves or something

idk the utf-8 is unrelated

[real sierra]

i need a good logo for the tournament

and for announcement banner

ideas?

can i just put terminator in a rook hat

[trim valve]

uci
isready
ucinewgame
position startpos moves g1h3 b8a6 b2b4 a6b8 c1b2 b7b5 d1c1 e7e6 b2e5 b8c6 e5f4 c6e7 c1b2 g7g6 b2c1 g8h6 g2g3 c7c6 h3g1 d7d5 f4h6 f8h6 b1a3 c6c5 f2f3 d8b6 c1d1 h6e3 a3c4 e3g1 f3f4 e8f8 e2e4 b6d8 d1h5 a7a5 c4a5 d5e4 a5c4 e7c6 h5h6 f8g8 f4f5 d8d4 a2a4 c6a7 f1h3 c8b7 e1f1 d4c4 f1e1 c4f1 h3f1 a7c6 f1b5 g1d4 h6e3 c6e5 a4a5 e5d3 e1e2 a8b8 e3h6 b8e8 a1e1 e4e3 c2c4 d4e5 b4c5 b7d5 f5f6 e5b8 h6e3 b8a7 e1c1 e6e5 c1b1 h7h5 b1d1 d3b4 c4d5 b4c2 h1e1 a7b8 d1c1 c2e1 e2f1 h5h4 b5c6 b8d6 d2d4 h8h6 c6b7 e8b8 d4e5 e1f3 c1c4 f3h2 f1g1 h2f1 e5d6 g6g5 e3g5 g8h8 g1f2 h8h7 g5f5 h7h8 c4h4 b8e8 h4h3 e8d8
go wtime 299992 btime 300000 winc 0 binc 0

much shorter but also crashes

gonna sleep now

[tender river]

good night

[sage crag]

bad night

[tender river]

are you nicogg

[sage crag]

noe

i konii

[tender river]

i'll consider it

[sage crag]

https://cdn.discordapp.com/emojis/1212937424062119967.gif?size=48&animated=true&name=nuhuh~1

i have already considered it for you

no consideration on your part is required

[tender river]

phrrr

[real sierra]

ok ill look into

[real sierra]

doesnt crash for me

big mistake just made

forgot the trailing .

whole project

[tender river]

[real sierra]

good thing i pushed chessapi.h like half an hour ago

[real sierra]

works on python bot too

are you sure you grabbed the new version of the API

am i sure i pushed the new version of the api

back to shrinking my bot

sent

i really thought this would work

[tender river]

as i mentioned macros cant be expanded within their own expansion

you have to create a separate macro for each arity

[real sierra]

unhelpful for shrinking my chess bot

but ty for explain

wait i genius

[opaque sigil]

am i misunderstanding or is the goal just something like this

[real sierra]

thats the goal

[opaque sigil]

or do you want it to be recursive

[real sierra]

but thats just a test

the end use case was gonna be function definitions

then i could leave off the curly braces and return type everywhere

[opaque sigil]

[real sierra]

this is the stage of token saving im at

so close

[opaque sigil]

thoughts?

[real sierra]

look at that token save

but of course they all have the same contents with that definition

[opaque sigil]

nothing stopping you from making it more complicated

pass more stuff to the macro

[real sierra]

its ok

i might not need it

im mangling my everything

im learning the tricks

always use the same generic variable names everywhere

so you can strip common patterns into defines

[sage crag]

https://cdn.discordapp.com/attachments/1067638175478071307/1406337271241904158/ezgif-3f552200e37817.avif

[real sierra]

just a few more tokens

all return values are now named rv

omg im so close now just like 20 more tokens

5 more

this code is a crime

[hollow spruce]

I will hack d bus

[opaque sigil]

what are you doing shiro

[real sierra]

OPTIMIZING

YOU WOULDNT UNDERSTAND

[hollow spruce]

System -end crash soo much

[real sierra]

THIS IS WHAT REAL C PROGRAMMING LOOKS LIKE

[opaque sigil]

i'm a washed c programmer

[real sierra]

i found it

i found the final optimization

[hollow spruce]

???

[opaque sigil]

[real sierra]

🎉

now its broken

[opaque sigil]

[real sierra]

[opaque sigil]

oh

[real sierra]

any way to expand the defines and nothing else?

[opaque sigil]

specific ones? not easily

i guess clangd can do that

[real sierra]

gdb gave function at least

[opaque sigil]

gdb my beloved

[real sierra]

now THAT is weird

is legal move checking magically broke or something now

im so lost

i think the bitboard macros broke the entire API

great

[opaque sigil]

dope

[real sierra]

but its like 200 bytes smaller it was so worth it

[dense kestrel]

wtf

actually not too bad yet

wait

why don't you compile and just give it the binary

i haven't the foggiest idea what the context is

[real sierra]

the context is i need my source code to be at most 1024 tokens

[dense kestrel]

wow

[real sierra]

ah

found it

i assumed !x was the same as x == 0

but !(-1) == true

its still acting

super weird

something else must be bugged

oh

because immediately after saying all that i put in x <= 0

putting my bot against the random bot

if i cant even beat this ive truly failed

well...

[dense kestrel]

wait when is it not

💀

[real sierra]

please read the line

directly below that one

[dense kestrel]

but !(-1) is false

💀

[real sierra]

thats what i thought too

[dense kestrel]

surely this is the c standard

[real sierra]

[dense kestrel]

bruh your thing is nonstandard

The result of the logical negation operator ! is 0 if the value of its operand compares unequal to
0, 1 if the value of its operand compares equal to 0. The result has type int. The expression !E is
equivalent to (0==E).

[real sierra]

its also an enum instead of being strictly an int

god only knows what happens under the hood

[dense kestrel]

https://tenor.com/view/soldier-ww2-traumatized-meme-eyes-gif-12257475272172704406

wow

[real sierra]

i have 8 tokens to improve this bot

how do i spend them

[sage crag]

can confirm

[dense kestrel]

adding newlines at the end of the file

[sage crag]

in fact, -1 is the only case where ! and ~ produce the same output

so

[dense kestrel]

wow real

unless you're on ones complement

[real sierra]

IM SO LUCKY

[dense kestrel]

like a weirdo

[sage crag]

two's complement is surely c standard

[dense kestrel]

should be implementation defined iirc

— the corresponding value with sign bit 0 is negated (sign and magnitude);
— the sign bit has the value −(2M) (two’s complement);
— the sign bit has the value −(2M − 1) (ones’ complement).

indeed

it leaves it up to implementation to choose which one

Which of these applies is implementation-defined

[real sierra]

if I run out of tokens again I think my next victim is );

[dense kestrel]

💀

have you hit all the for and while loops yet?

[real sierra]

for loops?

[dense kestrel]

with macros

[real sierra]

I only have for_0_i(lim)

no other types of for loops exist

[dense kestrel]

#define f(l)

[real sierra]

yes I already did that that's what I'm trying to say

[dense kestrel]

wow i thought you meant that you kept it that long

[real sierra]

no lol that's just the macro name

[dense kestrel]

lol

have you removed all the new lines and spaces and tabs?

[real sierra]

#define tree_child tree->children[i] is also carrying crazy weight

[real sierra]

those aren't tokens

won't have any effect

[dense kestrel]

incredible

[real sierra]

this isn't source code length I'm trying to minimize

just semantic units

[dense kestrel]

wait did you define a struct for that?

[real sierra]

I need a struct for what I'm doing

[dense kestrel]

why not just use adjacency list

with janky arrays

https://tenor.com/view/troll-face-troll-face-rage-comic-gif-2095096620717959646

[real sierra]

beo

you don't even know what the code does

🥀

[dense kestrel]

indeed

but i know that its always jankifiable

[real sierra]

my data structure choice is fine

[sage crag]

#define IF(x, y) if (x) { y; }
this saves 2 tokens per if

[real sierra]

adding more array indexes is not good

[real sierra]

you can only put certain things for y

y cannot contain a ,

or it delimits it

[sage crag]

noe just use __VA_ARGS__ instead of y

[real sierra]

that one might work

nice savings

[sage crag]

#define IF(x, ...) if (x) { __VA_ARGS__; }
IF(a, b)

[real sierra]

slowly but surely we're approaching functional programming

[sage crag]

this is

something

[real sierra]

python programmers with no macros

[dense kestrel]

[real sierra]

but they get sorted()

[olive sable]

Goodmorning

[sage crag]

IF(x, a) has the same number of tokens as if (x) a;

[dense kestrel]

wow

[velvet crest]

You have unlocked new role

[real sierra]

sam

[olive sable]

Me

[sage crag]

you know im just saying, its possible to entirely forego the x, if you

#define IF(...) if (x) { __VA_ARGS__; }

and just assume there is some x defined

[real sierra]

not worth

you then need a bool x = somewhere

or int x =

[dense kestrel]

💀

[sage crag]

possibly, but what about a global bool

[real sierra]

you still need x =

so the two tokens you save are lost

not to mention you only save those tokens in the #define

and the x = is per usage

[sage crag]

#define FUNC(ident, ...) ident() { __VA_ARGS__; }

// int x() { return 0; }
FUNC(int x, return 0)

@real sierra saves 1 token

[dense kestrel]

what if you write x&&y;

or x||y;

[sage crag]

this also makes me realise

#define and &&

save 1 token per &&

[real sierra]

probably doesnt

[sage crag]

it does

i checked

i have tcount

[real sierra]

screenshot

i dont believe

[dense kestrel]

what about using ternaries

x?y:z;

[real sierra]

im using ternaries everywhere

[dense kestrel]

use them for if/else

its still eval'ed

[real sierra]

wtf

blursed

makes me concerned for C

[sage crag]

same should also apply for ||

[dense kestrel]

wow

[real sierra]

there is only one instance of && in my code

and no instances of ||

[sage crag]

rip

[dense kestrel]

how many if statements

[real sierra]

i think i already replaced || with *

[dense kestrel]

or your if macro

[real sierra]

10

[dense kestrel]

replace with
condition&&f(x)&g(x)&h(x)...

i think its more efficient probably

[real sierra]

reasoning?

[dense kestrel]

the && is shorter than if

[real sierra]

can you show me some code

an example

[sage crag]

FUNC(int main, IF(1 and 2, return 0) return 1)

//expands to

int main() { if (1 && 2) { return 0; } return 1; }

[dense kestrel]

int uwu();
bool x = 1;
x&&uwu()&uwu()&uwu();

[sage crag]

you dont need to remove the spaces you know

spaces arent tokens

[dense kestrel]

well i didn't want to type

[real sierra]

clever

[dense kestrel]

wellididn'twanttotype

[real sierra]

but not shorter unless your condition is very simple

[sage crag]

ye since && is terminating

[real sierra]

any complex conditions need () around them

[sage crag]

let's see

[real sierra]

so you're only removing the if

and then adding && (1 token)

so you're net zero

[dense kestrel]

you can remove the parentheses + semi colon into a single macro

[sage crag]

could always use and

[real sierra]

thats why i said (1 token)

[real sierra]

you cant remove the parentheses

only works if x is a variable

[dense kestrel]

#define q )&
int uwu();
bool x = 1;
x&&uwu( q uwu( q uwu();

[real sierra]

if x is a condition, e.g. cond 1 > 0

it prob needs brackets

or does it

[dense kestrel]

i am pretty sure it doesn't

if u have space

[sage crag]

#define IF(x, ...) (x) and {{ __VA_ARGS__; }}

this would work on clang or gcc

[dense kestrel]

and space isn't token

[real sierra]

no parentheses needed it comes later in operator precedence

[sage crag]

the double curly brace is to get the brace to evaluate

auto y = {{ x; }};

ye this is a thing you can do in clang and gcc

[real sierra]

AUTO

I DONT GET AUTO

[sage crag]

strangely these are the same number of tokens

#define IF(x, ...) if (x) BLK(__VA_ARGS__)
#define IF(x, ...) (x) and {{ __VA_ARGS__; }}

[sage crag]

think again

[real sierra]

no

its against my principles

[sage crag]

-std=c23

[real sierra]

its c++ slop

[dense kestrel]

auto bad

[real sierra]

auto bad

[sage crag]

auto not bad

auto useful

[real sierra]

death before dishonor

auto bad

[sage crag]

why auto bad

[real sierra]

javascript feature

[dense kestrel]

its only useful when you need multi-d VLA return value

[sage crag]

it not javascript feature, still strongly typed

[real sierra]

bad bad bad

[sage crag]

i think you are wrong

[real sierra]

no like

no auto

[dense kestrel]

smh why use iterator when you can use pointer

[real sierra]

i spend extra tokens for no auto

[tender river]

shiro java dev

[sage crag]

must have long type name

[real sierra]

[sage crag]

[real sierra]

[dense kestrel]

typedef long long uwu;

wait

bruh

that mistake never occurred

[real sierra]

if you wanted

more details behind why im doing this

i plan to host a chess bot tournament here where your entry is limited to 1024 tokens

[dense kestrel]

wow

[real sierra]

[burnt ridge]

Shiro is intelligent

[dense kestrel]

this is actually very W idea

[real sierra]

you're welcome to participate

[dense kestrel]

surely optimal strat is to generate a seed such that you write a binary purely from rng output from that seed then run that binary

https://tenor.com/view/troll-face-troll-face-rage-comic-gif-2095096620717959646

[real sierra]

no reading/writing files

[dense kestrel]

only would take a few decillion years to get the seed

[dense kestrel]

wow

hmm

[dense kestrel]

write a JIT then do the random number trick and run it from memory

https://tenor.com/view/clueless-gif-24395495

[sage crag]

#define IF(x, ...) if (x) BLK(__VA_ARGS__)
#define IF(x, ...) (x) and BLK(BLK(__VA_ARGS__))
#define IF(x, ...) x ? (__VA_ARGS__) : PARENS

all three of these are the same number of tokens

[dense kestrel]

ooh #define q #define

i think that is fine

[sage crag]

awful

[dense kestrel]

think of all the tokens you'll save

[sage crag]

token counter doesnt think so

every one of these i replace is

increasing the token count

[dense kestrel]

wat

literally how

[sage crag]

ok for just

#define def #define

#define and &&
#define or ||
#define PARENS ()

swapping them to def reduces token count by 1

[dense kestrel]

all future defines also save

so cost is amortized

[real sierra]

[dense kestrel]

well not amortized

[sage crag]

no im counting this assuming both of them have the #define def #define

[real sierra]

* instead of and

[sage crag]

so cost is irrelevant

[real sierra]

save 3 tokens

[dense kestrel]

i haven't the foggiest idea how you all are counting tokens

[sage crag]

#define def #define
#define BLK(...) { __VA_ARGS__; }

but swapping #define to def here increases tokens by 2

[dense kestrel]

[real sierra]

https://github.com/RRethy/tcount/tree/master

easy way to think of it is words + punctuation (roughly)

int x = 1; -> ["int", "x", "=", "1", ";"]

[dense kestrel]

wow

incredible

[sage crag]

i mean in that case just dont use && or ||

and just use & or |

saves the same number of tokens without defines

[dense kestrel]

wait wtf why is brew installing another llvm

i have like installed llvm like 3 times now

💀

ts me when i just want some rust to install

[sage crag]

[real sierra]

2 && 1
2 & 1

[sage crag]

erm yes?

yes im aware that && is a terminating operator

[real sierra]

no the problem is

the first will be true

and the second false

[sage crag]

yes because it

terminating

operator

[real sierra]

then what tihs mean

[dense kestrel]

what

oh yes

my b

[tender river]

shro do you use the comma operator

[dense kestrel]

bozo momento

[real sierra]

considered it but couldnt find a situation that it saved tokens

[sage crag]

in cases where it applicable

[real sierra]

almost every , corresponds to one removed ;

[tender river]

it helps remove {}

also declare variables as early as possible and in bulk

[real sierra]

thats irrelevant i already have #define bclose ) {

bclose my beloved

[tender river]

e.g. not in loop body but outside of the loop

[sage crag]

or just declare fewer

[tender river]

[sage crag]

reuse variables better

[real sierra]

^

if you reuse variables you can bake them into macros

helps reduce tokens and increase macro usefulness

[sage crag]

ye which why i suggested

#define IF(...) if (x) BLK(__VA_ARGS__)
``` but no ![koniibwaa](https://cdn.discordapp.com/emojis/1367518991865610351.webp?size=128 "koniibwaa")

[real sierra]

it good thinking but not quite there

needs more time in oven

[sage crag]

you need more time in oven

[real sierra]

im stuck on hard problem

[tender river]

[real sierra]

how do i go from bitboard to piece type in as few tokens as possible

[sage crag]

this is opaque because i legally dont have your api

[tender river]

ye neither do i

[real sierra]

soon

once bred starts playing with their ai again and finds the last of the edge cases

contest is basically good to goo once this is done

i just want it thoroughly tested

[tender river]

[sage crag]

ye shiro replace all function with macro

save token

[tender river]

goto

[sage crag]

[real sierra]

[sage crag]

no goto, take token

[dense kestrel]

hmm if it were memory or time optim i would think it would be fine

but code size optimization is actually so hard ong

[real sierra]

small spoiler

in the Move struct, the origin and target squares are stored as BitBoards

where typedef uint64_t BitBoard

[sage crag]

bitboard with what info

[real sierra]

bit flip 1 = something there

[real sierra]

depends which bitboard it is

[real sierra]

they store piece locations, theres one per color-type combo

each bit corresponds to a board square

in the move struct, bitboard just store square from and square to

[dense kestrel]

what if you gave us an unsigned _BitInt(256) var;

[real sierra]

idk how go from this to piece type without just pulling all piece bitboards and checking

[dense kestrel]

instead for the board

so you can encode the actual piece as well

[real sierra]

im not using c23 im pretty sure

[dense kestrel]

unfort

[sage crag]

-std=c23

it so easy

shimo do you know about __COUNTER__

[real sierra]

token counter made 4 years ago

might be uncounted tokens in c23

[sage crag]

there are none

that im aware of

[real sierra]

maybe ill add the flag then

regardless, not switching bitboards from uint64_t

more helpful internally for them to be separate

[dense kestrel]

hmm

would it be that bad with 14 bitboards?

[real sierra]

14?

[dense kestrel]

pawns, knights, bishops, queens, kings, rooks, position occupancy

okay 13

[real sierra]

nono

there are only 12

what the bitboard stores is their occupancy

[dense kestrel]

wow

i see now

[real sierra]

the black pawn bitboard has bits flipped for every square containing a black pawn

very nice internally, lots of fast ops since its all bitwise integer math

unfun on token limit

[real sierra]

i do now

isnt that op

not that op actually

i was gonna use it in a for loop def but

actually no still op

[sage crag]

#define concat_inner(a, b) a##b
#define concat(a, b) concat_inner(a, b)
#define erf concat(erf, __COUNTER__)

erf // erf0
erf // erf1
erf // erf2

[dense kestrel]

#define c __COUNTER__

[real sierra]

#define for_(l) for_v(i ## __COUNTER__, l)
#define for_v(var, l) for(int var=0;var<l;var++)

[sage crag]

efine

[dense kestrel]

wait what are the time limits for the computation of the bots

[real sierra]

1 minute chess game

bots manage their time management

[dense kestrel]

https://tenor.com/view/bocchi-the-rock-bocchi-ryo-yamada-30000-usd-shocked-gif-3946323750587392724

wow i forgot the header of that

[tender river]

i saw this van gogh painting and thought "ai art" because of the palette

[sage crag]

im not sure this works how you think it does

[real sierra]

why not

OH

i redefined for