[userbinator]

...or "instant"(!) messaging applications taking gigabytes of memory and a full CPU core, and yet still can barely keep up with how fast a human can type.

I've often complained out loud with coworkers, while waiting for some horrible webapp to do its thing: "This computer can execute over a billion instructions every second. How many instructions does it take to render some formatted text!?!?"

Related: https://news.ycombinator.com/item?id=16001407

[marcosdumay]

For the likes of it, 15e9.

While throughput is reasonably easy to optimize for, for latency you will havSoftware latency is a hard to optimize target. Throughput is much easiere to fight against each abstraction layer on your code. And that includes layers bolted on your OS and hardware.