[BubRoss]

I try to remind people whenever I can that Knuth was talking about noodling loops in miniscule way. Between optimizing compilers, out of order super scalar CPUs and very different performance characteristics of modern CPUs, what he was talking about basically doesn't exist anymore.

[ska]

I don't think this is a helpful/accurate view. At a high level the type of optimization activity Knuth was talking about is alive and well, although the details of what people spend that time on has sometimes shifted.

I agree this quote is often abused but the fundamental idea behind it is intact and important: Sure, if you don't at least thing architecturally about performance early on as your problem domain reveals itself, you can make some poor decisions with long reaching performance implications. But on the other hand, if you spend a bunch of time tuning code when you don't know what the use will look like that time can be a dead loss.

This latter point was what Knuth was referring to - and in 2020 teams are still prematurely optimizing; i suspect about as much as they were back then.

[gameswithgo]

This idea is important when trying to finish an actual product of some kind, but when kids (or adults!) are learning, let them fiddle with the loops! Let them learn some intricacies and encourage the curiosity.

[ska]

Nothing wrong with fiddling with loops either, in that context.

I was objecting to the idea that premature optimization isn't a real problem anymore because technology. That's just not true.

[BubRoss]

No one is saying that. Knuth was frustrated with people wasting time on micro optimizations. Those don't really exist in the same form any more. Architecture is far more important to the speed of software and that does need to be dealt with up front.

The problem is when people YOLO their way through a program thinking optimization is for suckers because of an outdated quote from a different context.