Performance: Prefer prefix operator for non-primitive types #8579

No description provided.

Do you have some numbers? It would be nice to see some benchmarks added first and then comparing ++them with++ 👍

utACK fa18923, but I really doubt this will be measurable.

This was suggested in #7656#discussion-diff-56476755, but apparently for base58 this has 0 effect, the bench binaries are identical. :laughing:

Some other hunks did not produce identical binaries, so in theory there could be a performance difference. Though, I don't think the performance gain matters in those other places.

Not sure about specific cases here but in general performance gain could be quite significant: http://stackoverflow.com/a/9519095 Also it might simply be a good practice to use prefix operators by default and postfix only when needed imo.

In practice I expect this to have very little effect for containers with 'simple' iterators (where the iterator essentially consists of a single pointer). This is the case for std::vector, std::list, std::set, std::map, std::multimap (but not for std::vector<bool> or std::deque). In this case, technically a copy is made before returning, but the compiler will just optimize it away if it is not used.

Thank you for testing this Marco!

Some other hunks did not produce identical binaries, so in theory there could be a performance difference. Though, I don't think the performance gain matters in those other places.

Even is changing it retroactively is not deemed worth it, let's try to encourage ++x iso x++ usage in new code, that's pretty uncontroversial.

Even is changing it retroactively is not deemed worth it, let's try to encourage ++x iso x++ usage in new code, that's pretty uncontroversial.

Agree. Let's add that to doc/developer-notes.md.