Thread

1. nylonstrung · 1790921796 · focus · HN ↗

   At this point I very much question that handwritten assembly can be more performant than what LLVM-based langs can produce at the highest optimization settings. Compilers are very good at optimization

   Particularly when the "handwritten" code in question is vibecoded, like this is

   1. hackermanai · 1790924141 · focus · HN ↗

      It should be more performant if actually handwritten I guess. But just lowering the implementation like this will make it faster than many other apps seemingly for free, but not having the abstractions from e.g. C or C++ will likely make it unmaintainable long term.
   2. tbrockman · 1790926090 · focus · HN ↗

      LLVM is bounded by what its IR can express, and what it can express is constrained by what is portable and what it can prove. There is a reason `ffmpeg`, OpenSSL, OpenBLAS, glibc's string routines, LuaJIT's interpreter, etc. all still leverage hand-written assembly.

      As for the editor, even if it isn't necessarily optimal assembly (I have no clue), a reduced feature set and relying on OS APIs instead of frameworks will still probably result in something which feels faster than Electron programs like VSCode.

   3. Someone · 1790954635 · focus · HN ↗

      > Particularly when the "handwritten" code in question is vibecoded, like this is

      Also, “For Apple silicon, a build-time translator converts that core to AArch64” likely doesn’t help the quality of the ARM code.

Unofficial Hacker News client; not affiliated with Y Combinator.