Ask HN: What industries need bare metal hand tuned performance optimizations?

4 points by k0k0r0 2 years ago | 5 comments
I know that they do that a lot in high-frequency trading and in the gaming industry. Both of them are not an option to me. Of course, this is done in academia, but that's not "really" an industry.

I suspect there is a lot of real world simulation, especially of some physics (and graphics, but that is less interesting to me), where such optimizations are needed. But as far as I understand, in general such help is not really appreciated there. The cost of to other programmers to deal with necessary changes existing code, as well as the fact that hardware and compute time are cheaper than programmers, seems to make things difficult. I seriously consider learning to write python interfaces to do "magic" to python projects, but I don't know whether there actually is a market for that.

What are your experiences? Is anybody here paid to speed things up?

Edit: Typos

  • samtho 2 years ago
    Embedded development but specifically low-powered embedded development. Not everything is wifi and cell network, sometimes you have devices that wake up, broadcast a signal over LoRa or Zigbee and return to sleep. Specialized batteries can last 10+ years. You see some of this stuff in smart lighting remotes and sensors.

    However much of this was done in the interest of as few cpu cycles as possible rather than raw speed.

    • emrah 2 years ago
      Although it's not a purely software industry, embedded development has a lot of opportunity for bare metal hand tuned optimizations
      • monero-xmr 2 years ago
        Cryptocurrency requires money to pay for transactions. The most performant pieces of smart contracts by the best firms are written in assembly, as this saves money. The crypto industry has been the primary financial motivator for advancement in formal verification technology over the past few years.
        • k0k0r0 2 years ago
          I really appreciate the reply and found another no-option. I would like to do something less... "abstract". I would definitly agree that crypto has meaningful usecases, but I guess it is not so easy to find those that are meaningful to me.