28
oraro
4y

Everyone's a pro ninja 10x whatever until you ask them to create multi-core optimized code

Comments
  • 12
    My last job our code base scaled linearly to 48+ physical cores... I doubled its throughput. Beg to differ!!!
  • 3
    @atheist your a real 48x engineer
  • 0
    Laughs in haskell
  • 1
    @atheist a 10x unicorn ninja rockstar from space worthy of the name
  • 1
    this is actually a huge problem. At my school there was only one class that taught multi threaded programming, and it was NOT mandatory. So i can definitely see why this is the case a lot of the time. it's unfortunate.
  • 1
    Python developers can cry in corner
  • 2
    It becomes even more fun when you throw in thread priorities. I mean real time priorities, not the Linux style.
  • 0
  • 2
    Well, at school I was the king of performance & multithreaded C (perfect 100% cpu utilisation on 20 cores 🤤)

    My real magic trick was to not use any kind of mutex (or atomics) 😆

    I only started using them because Rust forces me to 😆
  • 0
    @oraro or you know, an sde at Amazon, close enough? My point was more some of us have actually had to do that kind of high performance multi-threading.
  • 0
    @NoToJavaScript
    Laughs in custom logic.
  • 1
    @MagicSowap
    Teach me master.
  • 1
    @Gregozor2121 I guess there are cases where you need to do it manually. But I would argue 80% of the time it is better to use something what already exists
  • 0
    @chowdercake 😭😭😭
  • 0
    HA
    fork()
    HA HA
    fork()
    HA HA HA HA
  • 0
    With Go there are quite some constructs that help with that. Not mucking with treads but just focus on parallelization and either using mutexes or channels to do it safely. No more idle core.
Add Comment