๐๐๐๐๐๐๐๐
27.8K posts

๐๐๐๐๐๐๐๐
@noctice007
ู ู ุตุจูุงู ุฌููู





@svpino AI can't build a competent data/video/audio compression algorithm. Those people are still hand crafting ASM to optimize their inner loops. Same with CUDA compute kernels. LLM written CUDA is super slow. LLMs try to optimize CUDA like it was CPU C++ code.




"simple software" is a fairy tale every engineer likes to tell themselves. In reality, simple software is either dysfunctional, or satisfactory to at most 5 people on the entire planet. You simply cannot make software simple, because software is not a single math equation you apply to everything equally - the range of functionality and usecases of (realistically) any important piece of software is so great the code cannot be "simple". This entire idea on linux stems from dwm, (in my opinion) one of the stupidest marketing tricks that worked on dumb linux users. DWM is a barely functional, glorified lua script for barebones tiling, that only works because it runs on top of a fucking million line of code behemoth that is Xorg which does everything for it. DWM accomplishes nothing of value and that's why it can be 2k LOC. It's even more hilarious once you look at the wayland spiritual successor - dwl. It's purely dysfunctional, even with the 80k-or-so LOC wlroots (plus deps) powering it. Once you go past "I need to open 2 terminals" everything starts falling apart. The idea of "less lines of code = faster and more readable code bro" is something I'd expect a first year CS student to say, not an actual experienced programmer. The linux kernel is 50 million lines or whatnot and I don't see you complaining that AMD just added 10k more lines because a new gpu came out and they need to add driver support for it. No, patching code for functionality is terrible and is a braindead method of adding features that doesn't even work in most cases. No, more features do not make your computer slower. If you don't use it, the code doesn't fuckin run. You will never notice that 4ns branch ran 2x a minute dear linux user. No, more LOC doesn't mean slower. Hell, in some cases, more LOC means faster, because we can do intelligent things and not the fast-and-dirty methods. But that's just an opinion, of course. What do I know.


ูุงู ุงูููุฏู ููู ุจุฑุฏ: ุฅูู ุฃุฌุฏ ูู ููุงู ุงูุนุฑุจ ุญุดููุง! ููุงู ุงูู ุจุฑุฏ: ุฃูู ูุฌุฏุช ุฐููุ ููุงู: ูููููู ุนุจุฏุงููู ูุงุฆู ุซู ูููููู: ุฅู ุนุจุฏ ุงููู ูุงุฆู ุซู ูููููู ุฅู ุนุจุฏุงููู ููุงุฆู ุ ูุงูุฃููุงุธ ู ูุฑุฑุฉ ูุงูู ุนูู ูุงุญุฏ ููุงู ุงูู ุจุฑุฏ: ุจู ุงูู ุนุงูู ู ุฎุชููุฉ ูุงูุฃูู ุฅุฎุจุงุฑ ุนู ููุงู ู ูุงูุซุงูู ุฌูุงุจ ุนู ุณุคุงู ูุงูุซุงูุซ ุฑุฏู ุนูู ู ููุฑ.








>Linux is C >Git is C >Python interpreter (CPython) is C >Ruby interpreter (MRI) is C >PostgreSQL is C >SQLite is C >Redis is C >MySQL is C++ >MongoDB is C++ >Unreal Engine is C++ >Chrome is C++ >Firefox is C++ >Windows kernel is C >macOS kernel (XNU) is C >Photoshop is C++ >VMware is C++ >TensorFlow (Core) is C++ -still, you are not convinced to learn C and C++



Rust lost the race to low level programming the moment it decided to not make raw pointers ergonomic. For high level systems programming it's great, but for building the abstractions needed for such systems it simply sucks This one's a nice read as well zackoverflow.dev/writing/unsafeโฆ









