00:23 <+bridge> <teero777> @ryozuki what cpu do you have? xd
00:27 <+bridge> <Ewan> R9 7900X3D
07:54 <+bridge> <♂S1mple♂> What it's going to do?
08:31 <+bridge> <ryozuki> morning
08:50 <+bridge> <teero777> gm
08:50 <+bridge> <milkeeycat> Morning GMT+3 enjoyers and others
09:36 <+bridge> <meloƞ> Morning systemd enjoyers and others
09:36 <+bridge> <ryozuki> hi after gym
09:36 <+bridge> <ryozuki> now rust
09:36 <+bridge> <ryozuki> :gigachad:
09:39 <+bridge> <meloƞ> Did you think about what you wanna do as a package Manager yet ? :gigachad:
09:47 <+bridge> <ryozuki> https://iroh.computer/
09:47 <+bridge> <ryozuki> https://iroh.computer/docs
09:48 <+bridge> <ryozuki> hmm no but ill do something basic first
09:48 <+bridge> <ryozuki> git based
09:48 <+bridge> <meloƞ> :owo:
09:51 <+bridge> <meloƞ> I thought about doing Something similar once with the declarative Part of nixOS for different Versions to Upgrade/downgrade and ease of linking different Versions But still easy to use (without having to manually specify each Version in a File)
09:53 <+bridge> <meloƞ> Package Change python@2.7
09:53 <+bridge> <meloƞ> 
09:53 <+bridge> <meloƞ> `you are using python3.0, are you Sure you want to Switch to Python Version 2.7? This will keep python3.0 installed But remove its symlink`
09:53 <+bridge> <meloƞ> 
09:53 <+bridge> <meloƞ> Something Like that
09:54 <+bridge> <ryozuki> thats pyenv
09:54 <+bridge> <AssassinTee> It'll generate automappers for your tileset, you just need to configure each tile, but they're preconfigured. With the current state I can create an automapper for all of grass_main in under a minute
09:54 <+bridge> <meloƞ> Yesyesyes, now do that for clang where you want clang17 But clang-format 10 without having to manually Download Compile and symlink/move it to /User/bin
09:55 <+bridge> <meloƞ> I would want IT to Work with Git Tags aswell as releases
10:01 <+bridge> <♂S1mple♂> That's interesting
10:02 <+bridge> <teero777> first thing i see xd
10:02 <+bridge> <teero777> https://cdn.discordapp.com/attachments/293493549758939136/1240937483081089127/image.png?ex=66486097&is=66470f17&hm=5073930c743da78139b67ab63a02e0f489db22c9a7f43515e6daa2fd69e9b2a0&
11:44 <+bridge> <.233447> heyy im playing ddnet and im near in the end this part you can finish it in solo but someone hooks me and go me back and i have been 20m near to finish and he dont let me i do vote kicks but all press f4 can you kick him plsss
13:18 <+bridge> <ryozuki> https://cdn.discordapp.com/attachments/293493549758939136/1240986824986460160/image.png?ex=66488e8b&is=66473d0b&hm=f3c05337a7b9abad96c9eb1482b6e0c502d3804dabafd4014913c5fb2da8c1b4&
13:20 <+bridge> <ryozuki> https://doc.rust-lang.org/std/option/index.html#representation
13:51 <+bridge> <chillerdragon> That is amazing! I wanted to have such tool for a while now. But was too lazy to build it. Seems quite complex.
14:06 <+bridge> <chillerdragon> abuse
14:43 <+bridge> <milkeeycat> *bans chiller*
15:12 <+bridge> <Ewan> you’re banned
15:13 <+bridge> <Ewan> looks like ipfs
15:13 <+bridge> <Ewan> with better marketing
15:26 <+bridge> <ryozuki> https://blog.rust-lang.org/2024/05/17/enabling-rust-lld-on-linux.html
15:30 <+bridge> <Jupstar ✪> sadly i already use mold
15:32 <+bridge> <ryozuki> but good default for ppl who didnt change it
15:33 <+bridge> <Jupstar ✪> true, i just need zen5
15:34 <+bridge> <ryozuki> @jupeyy_keks why does the C ABI require 16 byte alignment for int128
15:34 <+bridge> <ryozuki> when the cpu only needs 8 for efficient load/stores
15:34 <+bridge> <ryozuki> <a:AngryAwooGlitch:451548046950531083>
15:35 <+bridge> <Jupstar ✪> bcs c must also run on my GPU
15:35 <+bridge> <Jupstar ✪> 😏
15:36 <+bridge> <Jupstar ✪> tbh abi stability is such an impossible topic
15:36 <+bridge> <Jupstar ✪> but at least for debug mode i'd also like to see it on rust
15:36 <+bridge> <Jupstar ✪> simply bcs of hot reloading puposes
15:36 <+bridge> <Jupstar ✪> for release it should ofc prefer max perf
15:38 <+bridge> <Jupstar ✪> i'd already be happy if it's stable between two invocations of the same compiler
15:38 <+bridge> <Jupstar ✪> and per arch
15:40 <+bridge> <Ewan> :f3:
15:40 <+bridge> <learath2> ackshully C does not define an ABI 🤓
15:41 <+bridge> <Ewan> i was about to say it
15:41 <+bridge> <Ewan> but 👽
15:41 <+bridge> <Ewan> where is the C language
15:41 <+bridge> <Ewan> find it
15:42 <+bridge> <ryozuki> so what does it?
15:42 <+bridge> <ryozuki> the c call convention?
15:42 <+bridge> <learath2> C doesn't define a calling convention either
15:43 <+bridge> <ryozuki> so its gcc and clang?
15:43 <+bridge> <learath2> it's the OS
15:43 <+bridge> <Jupstar ✪> i guess u simply mean alignment requirements
15:43 <+bridge> <ryozuki> x86 needs just 8 byte for efficient load store of a i128
15:43 <+bridge> <ryozuki> but to interface with C u need to use 16
15:44 <+bridge> <Ewan> says who
15:44 <+bridge> <learath2> https://groups.google.com/g/x86-64-abi ask here y
15:44 <+bridge> <ryozuki> In C and C++, on x86_64 Linux, alignof(__int128) is equal to 16. However, in Rust, align_of::<u128>() is equal to 8.
15:44 <+bridge> <ryozuki> 
15:44 <+bridge> <ryozuki> C++: https://gcc.godbolt.org/z/YAq1XC
15:44 <+bridge> <ryozuki> Rust: https://gcc.godbolt.org/z/QvZeqK
15:44 <+bridge> <ryozuki> 
15:44 <+bridge> <ryozuki> This will cause subtle UB if you ever try to use i128s across FFI boundaries; for example:
15:44 <+bridge> <ryozuki> 
15:44 <+bridge> <ryozuki> C++: https://gcc.godbolt.org/z/PrtHlp
15:44 <+bridge> <ryozuki> Rust: https://gcc.godbolt.org/z/_SdVqD
15:44 <+bridge> <ryozuki> this was fixed in rust 1.77
15:44 <+bridge> <ryozuki> to be 16 too
15:45 <+bridge> <Ewan> :rplace:
15:45 <+bridge> <Jupstar ✪> does c even have a 128 bit integer type standarized?
15:45 <+bridge> <Ewan> No
15:45 <+bridge> <Ewan> maybe in modern C
15:45 <+bridge> <Ewan> idk actually
15:45 <+bridge> <Ewan> but i think it’s extensions
15:45 <+bridge> <learath2> Ah, I do have an idea why they do 16 byte alignment
15:47 <+bridge> <ryozuki> simd?
15:47 <+bridge> <ryozuki> C23 defines a int with arbitrary size i think
15:47 <+bridge> <ryozuki> https://blog.tal.bi/posts/c23-bitint/
15:48 <+bridge> <ryozuki> ```c
15:48 <+bridge> <ryozuki> constexpr size_t K = /* compile time value */;
15:48 <+bridge> <ryozuki> typedef _BitInt(K) color;
15:48 <+bridge> <ryozuki> 
15:48 <+bridge> <ryozuki> struct RGB {
15:48 <+bridge> <ryozuki>     color R;
15:48 <+bridge> <ryozuki>     color G;
15:48 <+bridge> <ryozuki>     color B;
15:48 <+bridge> <ryozuki> }; /* sizeof(RGB) == 3 * K */
15:48 <+bridge> <ryozuki> ```
15:48 <+bridge> <Ewan> wild
15:50 <+bridge> <learath2> this was my guess but I couldn't find a source for it
15:50 <+bridge> <ryozuki> simd requires 16byte align
15:51 <+bridge> <learath2> I meant as in a source that states that that is indeed the reason
15:52 <+bridge> <Jupstar ✪> friendly reminder:
15:52 <+bridge> <Jupstar ✪> https://cdn.discordapp.com/attachments/293493549758939136/1213230303875964940/image.png?ex=66486f05&is=66471d85&hm=b35a80f2bead998590f38ed8a3a92282acbb0d2844e4d7d291c203ec456c0521&
15:52 <+bridge> <Jupstar ✪> that is what i call an alignment
15:53 <+bridge> <Jupstar ✪> 16 bytes ,pff
15:53 <+bridge> <ryozuki> whats that from
15:53 <+bridge> <ryozuki> gpu?
15:53 <+bridge> <Jupstar ✪> yes
15:53 <+bridge> <Jupstar ✪> allocating memory for a texture
15:54 <+bridge> <ryozuki> why that alignment?
15:54 <+bridge> <ryozuki> is that the bus transfer alignment or smth
15:55 <+bridge> <ryozuki> explain gpu expert
15:55 <+bridge> <Jupstar ✪> well i guess it has to do with such things yeah, but why exactly i dunno.
15:55 <+bridge> <Jupstar ✪> 
15:55 <+bridge> <Jupstar ✪> could also simply be that when 2000 shader cores read from memory it sucks latency wise if they are too close
15:55 <+bridge> <Jupstar ✪> guess u'll need to ask chairn
15:56 <+bridge> <Jupstar ✪> i wonder anyway when GPUs actually cache memory vs when not
15:56 <+bridge> <Jupstar ✪> it's different to CPUs
15:56 <+bridge> <Jupstar ✪> time to sneak into nvidia corp
16:02 <+bridge> <Jupstar ✪> @ryozuki do you often have to interact with low level stuff like alignment when working on your compiler? or does llvm abstract most stuff away and u just sometimes hit those?
16:06 <+bridge> <ryozuki> you pass a data layout to llvm
16:06 <+bridge> <ryozuki> and it handles those in store,loads and allocas
16:06 <+bridge> <ryozuki> but the hack here is that when we want to call a function of the program from jit or as a shared lib we need to care
16:06 <+bridge> <ryozuki> xd
16:07 <+bridge> <ryozuki> but the hack here is that when we want to call a function of the program from jit or as a lib we need to care
16:19 <+bridge> <ryozuki> friday
16:19 <+bridge> <ryozuki> rust
16:19 <+bridge> <ryozuki> @jupeyy_keks
16:21 <+bridge> <Jupstar ✪> i updated quinn and rustls and have a bug now
16:21 <+bridge> <Jupstar ✪> epyc friday rust moment
16:22 <+bridge> <learath2> I did dig around a bit out of curiosity. Intels optimization guide for Intel 64 bit processors suggests that instructions that move data operate best with destinations that are aligned to the natural alignment of the type
16:23 <+bridge> <learath2> Also, store to load forwards for 128-bit integers don't stall the pipeline if the destination of the load is aligned to the natural size
16:27 <+bridge> <learath2> Finally I think it also allows compilers to optimize using SIMD registers to move 128 bit values more efficiently
16:28 <+bridge> <learath2> http://www.intel.com/content/dam/www/public/us/en/documents/manuals/64-ia-32-architectures-optimization-manual.pdf here is what I was digging around it, mainly section 3.6.4, observe that the table only guarantees non stalled 128 bit store to load forwarding if both the store it naturally aligned and the load is from a dqword aligned address (also the natural alignment)
16:33 <+bridge> <ryozuki> https://increment.com/software-architecture/exit-the-haunted-forest/
16:34 <+bridge> <ryozuki> interesting, thanks
18:48 <+bridge> <ryozuki> dead chat
18:50 <+bridge> <ryozuki> https://www.osnews.com/story/139444/gentoo-bands-use-of-ai-tools/
18:50 <+bridge> <ryozuki> did we have trouble with ai prs and commits?
18:50 <+bridge> <meloƞ> u
18:50 <+bridge> <ryozuki> we should ban them to make a marketing  pr post
18:50 <+bridge> <ryozuki> :gigachad: :poggers2:
18:51 <+bridge> <ryozuki> @scrumplex did u have issues with ai prs
18:51 <+bridge> <meloƞ> imagine beeing called "osnews" and having a typo in your url
18:51 <+bridge> <meloƞ> gentoo bands? what kind of music they playing!
18:51 <+bridge> <ryozuki> xd
18:52 <+bridge> <ryozuki> i hate mobile
18:52 <+bridge> <ryozuki> https://www.reddit.com/r/programming/s/7ZezDaY9Ne
18:54 <+bridge> <Jupstar ✪> "Gentoo Council" lmao
18:54 <+bridge> <Jupstar ✪> 100 users have a council
18:54 <+bridge> <Jupstar ✪> to feel better
18:54 <+bridge> <meloƞ> i like this one:
18:54 <+bridge> <meloƞ> ```
18:54 <+bridge> <meloƞ> This is a disturbing trend. The AI kids believe they can automate software engineering with AI chatbots yet they not even know what the software development process of software is. And they are very confident of what they don't have experience about
18:54 <+bridge> <meloƞ> 
18:54 <+bridge> <meloƞ> A call it the new cargo cult programming
18:54 <+bridge> <meloƞ> ```
18:55 <+bridge> <learath2> don't be jealous you don't have a cool council making your decisions
18:55 <+bridge> <Jupstar ✪> the problem is, aren't gentoo ppl also rather script kiddies than actual software architects?
18:55 <+bridge> <meloƞ> oh god what have you done, ryo will launch all his fishies at once
18:55 <+bridge> <Jupstar ✪> i mean they don't develop the software themself do they?
18:56 <+bridge> <learath2> They do develop some software, mostly distro management stuff like the package manager
18:56 <+bridge> <learath2> Also in the past they maintained a fork of eudev when eudev took the systemdpill, and ofc the most important, they maintain OpenRC
18:57 <+bridge> <learath2> Well technically nowadays openrc has split off a bit more, but it's still mostly the same people
18:58 <+bridge> <fokkonaut> ping
18:58 <+bridge> <learath2> There is also the linux kernel patchset they maintain for gentoo
18:58 <+bridge> <fokkonaut> @melon
18:58 <+bridge> <fokkonaut> @blaiszephyr
18:58 <+bridge> <learath2> Are you having a stroke?
18:58 <+bridge> <fokkonaut> Kinda
18:58 <+bridge> <fokkonaut> Sorry to interrupt haha
18:58 <+bridge> <learath2> rip, you'll be missed
18:59 <+bridge> <fokkonaut> oh, that's very kind
18:59 <+bridge> <meloƞ> !ban fokkonaut 0
18:59 <+bridge> <fokkonaut> ❤️
18:59 <+bridge> <meloƞ> i wanna know if NixOS will follow the same path in declining PR's that used AI
19:00 <+bridge> <meloƞ> espacially because NixOS changes daily, whatever the AI spits out, is 99% wrong :D
19:03 <+bridge> <meloƞ> i find it so interesting that we're all playing the same game and contributed to such  - but everyone has his own preferences and does things differently, we have ~
19:03 <+bridge> <meloƞ> i find it so interesting that we're all playing the same game and contributed to such  - but everyone has his own preferences and does things differently, we have ~10 devs in here and every single one of them codes differently
19:03 <+bridge> <meloƞ> i find it so interesting that we're all playing the same game and contributed to such  - but everyone has his own preferences and does things differently, we have ~10 devs in here and every single one of them codes differently - their environment, their toolchain, OS - it all differs to some point and i find that quite interesting
19:04 <+bridge> <meloƞ> i find it so interesting that we're all playing the same game and contributed to such  - but everyone has his own preferences and does things differently, we have ~10 people in here activly chatting  and every single one of them codes differently - their environment, their toolchain, OS - it all differs to some point and i find that quite interesting
19:04 <+bridge> <meloƞ> i find it so interesting that we're all playing the same game and contributed to such  - but everyone has his own preferences and does things differently, we have ~10 people in here activly chatting  and every single one codes differently. Their environment, their toolchain, OS - it all differs at some point and i find that quite interesting.
19:06 <+bridge> <learath2> diversity is our power
19:08 <+bridge> <fokkonaut> I'm still using windows and visual studio 😤
19:09 <+bridge> <fokkonaut> xDD
19:09 <+bridge> <learath2> That's okay, we still like you as is
19:09 <+bridge> <meloƞ> yeah.. haha **hides knive** we respect you totally!!
19:10 <+bridge> <fokkonaut> 🔫
19:13 <+bridge> <fokkonaut> I think that kind of diversity can be applied to all human beings
19:16 <+bridge> <scrumplex> nope :o
19:18 <+bridge> <ryozuki> https://tenor.com/view/developers-gif-4458491
19:18 <+bridge> <ryozuki> @blaiszephyr right now
19:20 <+bridge> <ryozuki> this channel is also a rust sanctuary
19:20 <+bridge> <ryozuki> church*
19:20 <+bridge> <ryozuki> in rust we trust
19:20 <+bridge> <meloƞ> https://tenor.com/view/dsmp-gif-27256181
19:24 <+bridge> <egyt> :^(
19:24 <+bridge> <egyt> https://cdn.discordapp.com/attachments/293493549758939136/1241078905524195328/6490a98ad967077d.png?ex=6648e44d&is=664792cd&hm=f0d6aeff806cc193663e9c009127848a5a02816cd128123d89d1dda68a20bad3&
19:25 <+bridge> <ryozuki> @blaiszephyr we also discuss compilers, C stuff, abis, the meaning of life and more
19:25 <+bridge> <meloƞ> !!
19:26 <+bridge> <ryozuki> will ur kid learn rust? 🧐
19:26 <+bridge> <meloƞ> at the age of 6 she will learn assembly, at the age of 8 she will learn how to build her own phone
19:26 <+bridge> <ryozuki> good
19:26 <+bridge> <meloƞ> rust will be learned to write the phone's backend
19:27 <+bridge> <ryozuki> make sure to make her wear the appropiate programming socks
19:27 <+bridge> <learath2> I'm always here to provide any knowledge you need  from the old world
19:27 <+bridge> <meloƞ> i will buy her some!
19:28 <+bridge> <ryozuki> is it old if its needed?
19:28 <+bridge> <ryozuki> learn the sys v abi by memory
19:28 <+bridge> <ryozuki> the abi wizard
19:28 <+bridge> <learath2> It's not needed. You must be doing things wrong. Have you considered just using python?
19:28 <+bridge> <ryozuki> :justatest:
19:29 <+bridge> <learath2> This is a bit of a useless exercise imo. Atleast the details. I know the outline of it and where to look when needed instead
19:29 <+bridge> <milkeeycat> Is there any haskell enjoyers?
19:29 <+bridge> <ryozuki> no
19:29 <+bridge> <ryozuki> use elixir
19:33 <+bridge> <learath2> It's more like opt-in dynamic typing
19:33 <+bridge> <learath2> Most of everything I write is statically typed
20:09 <+bridge> <ryozuki> :NekoDrink: