I used to make 3D engines and asset pipelines professionally. High performance really is critical in both. So, I’d have to check myself frequently:

> If I optimize this, how much time is it going to save the whole world? What will be the sum of time saved by everyone who will ever run this code? Is that less than the time required of me to do the optimization? Then, don’t do it!

That question was relevant quite often.

The big optimizations though really mattered. And, required planning before initial implementation to set the system up to be optimizable.