Skip to content
DevMeme
3099 of 7590
A Creative, if Destructive, Approach to Runtime Optimization
Interviews Post #3415 · source on Telegram

A Creative, if Destructive, Approach to Runtime Optimization

Description

This is a two-part meme format contrasting a technical interview question with a candidate's absurd solution. The top section, labeled 'Interviewer:', asks, 'Can you improve the runtime of this function?'. The middle section, labeled 'Me:', displays a Python code snippet from a file named 'shitpost.py'. The code imports 'Thread', starts a given 'my_function' in a separate thread, and then enters a 'while' loop that calls 'time.sleep(-1)'. The bottom section shows the interviewer's reaction using the 'Outstanding move' meme, which features a man next to a chessboard, sarcastically applauding the move. The technical joke is that passing a negative number to 'time.sleep()' in Python raises an immediate 'ValueError', crashing the main thread. The function's runtime isn't improved at all; the program just terminates instantly, making it seem like the task is done. For experienced engineers, this is a multi-layered joke about interview pressure, lateral thinking, and the critical difference between actual performance optimization and simply making a program stop

Categories

Interviews Performance Languages Bugs DevCommunities

Common Tags

Interview Python PerformanceOptimization CodingHumor TechMeme

Comments

25
Anonymous ★ Top Pick My function's runtime is now O(ε), where ε is the time it takes the Python interpreter to process a fatal error. Your move, interviewer
  1. Anonymous ★ Top Pick

    My function's runtime is now O(ε), where ε is the time it takes the Python interpreter to process a fatal error. Your move, interviewer

  2. Anonymous

    Spawn the work in a thread, call sleep(-1), and watch Prometheus report P99=0 - turns out the fastest algorithm is convincing the metrics pipeline the request never existed

  3. Anonymous

    Ah yes, the classic 'negative sleep' optimization - because if sleeping for positive time makes things slower, surely sleeping for negative time will make them faster! Next up: improving database performance by setting the connection timeout to imaginary numbers

  4. Anonymous

    When asked to optimize runtime, this candidate discovered that negative sleep() is the ultimate performance hack - it's so fast, it travels backward in time! Who needs thread.join() when you can busy-wait with undefined behavior? The interviewer's chess analogy is apt: this move is so brilliant, it's like sacrificing your queen, both rooks, and your dignity in a single turn. At least the filename 'shitpost.py' sets appropriate expectations for code review

  5. Anonymous

    Asked to "improve runtime," I threaded it and called sleep(-1); the GIL plus a ValueError delivered the only true O(1): instant crash

  6. Anonymous

    Optimized the function to O(0): spawn a thread, return immediately - throughput unchanged, GIL unimpressed, but the KPI says “outstanding”

  7. Anonymous

    Shaves runtime to O(n/cores + eternal polling) - because proper Thread.join() is for architects, not interview gladiators

  8. @ZgGPuo8dZef58K6hxxGVj3Z2 5y

    Lol

  9. @nuntikov 5y

    what

    1. @ZgGPuo8dZef58K6hxxGVj3Z2 5y

      He makes it run on a different thread

  10. @nuntikov 5y

    -1?

  11. @bezuhten 5y

    Can u explain?

  12. @nuntikov 5y

    why not j0in?

  13. @t02x2 5y

    Threads will only use one CPU on your machine with python. Won’t be faster. Idea is funny though -> Use Multiprocessing

    1. @viktorrozenko 5y

      wait really? so Python only ever works on one CPU even when I make new Threads?

      1. @RiedleroD 5y

        there are ways to circumvent that, but it's quite complicated

        1. @viktorrozenko 5y

          might that be why Python concurrency sucks so much with its restrictions on net-card access only from the main thread and stuff like that?

          1. @RiedleroD 5y

            idk, I haven't actually done much research. I just know that python really doesn't want to be responsible for race conditions

      2. @t02x2 5y

        GIL ist the problem and has it's reasons https://realpython.com/python-gil/

      3. @slnt_opp 5y

        Yeah, it's same for many such languages, ruby too, because GIL That's why if it's about web apps one usually makes workers and forking🤷‍♂️

        1. @viktorrozenko 5y

          I just write Go instead, lol

          1. @slnt_opp 5y

            Me too, but simple stuff, like poll or smth - Django / Flask :)

            1. @viktorrozenko 5y

              Try FastAPI

  14. @t02x2 5y

    its very easy. Just use multiprocessing to start a new process. A python Interpreter can only be executed on one cpu. A new process is executed on a new cpu :)

  15. @Dexconv 5y

    laughs in go func() peasants

More like this

More like this in Interviews →
Interviews Interview Python
Jeff Dean, MapReduce Inventor, Asks 'What Is DSA Hard?'

Jeff Dean, MapReduce Inventor, Asks 'What Is DSA Hard?'

Interviews
10 YoE Engineer's Sarcastic Applause for Uni Data Structure Trivia

10 YoE Engineer's Sarcastic Applause for Uni Data Structure Trivia

Interviews
Vibe Coder Confesses JavaScript Framework Knowledge Is Just 'Regular' at Interview

Vibe Coder Confesses JavaScript Framework Knowledge Is Just 'Regular' at Interview

Frameworks
Use J and K for navigation