When to do optimization?

[VorfeedCanal]

@2e71828

For larger programs, an iterative development process certainly helps: Instead of writing the whole program before measuring and optimizing, write one modular part at a time.

Well, that's reasonable. And doesn't contradict things which I'm talking about.

@2e71828

It wouldn’t survive for 5 minutes.

Seriously? I think you don't estimate (or maybe overestimate?) humans. I have seen way to many people who interpret this

@2e71828

Write for clarity first, and don't manually optimize until you've verified that you can do a better job than the compiler

in the following way:

1. Write the code which is clear and simple.
2. Show it to the others. Get approval and bonus.
3. Deploy/release it. Switch to other project if you want more bonuses.
4. When people complains about slowness and/or your bills for hosting grow to high — start thinking about optimization process. Usually that's problem for someone else than who did #1, #2 and #3.

Note that they perform your algorithm perfectly: till stage #4 they haven't verified that they can do a better job than the compiler. Because they never tried to optimize anything.

How can [u8; 1048576] be reject with such approach in 5 minutes — I have no idea.

@2e71828

Then, measure what you’ve written.

When should you do that? That's the question. Way too often that step is pushed for after delpoyment/release stage. That is how we end up with JavaScript which doesn't have integers or with Windows Update which needs three weeks to update Windows and other such things.

@2e71828

Then, refactor [u8; 1048576] into Rc<[u8; 1048576]> or Vec<u8> or something when it proves to be a problem.

Means when Windows Update starts taking weeks? Are sure you would be able to replace [u8; 1048576] with Rc<[u8; 1048576]> or Vec<u8> at this point?