On Thu Jul 11 13:20:38 2024 +0000, Tatsuyuki Ishi wrote:
Whether it makes a practical difference is the biggest question that needs to be answered about this change. Microbenchmarks and others are only useful to explain the practical metric, they are not justifications on their own. Hence I'm suggesting to prioritize having a way to measure practical impact, then focus on everything else. If you want help with measurements, how about providing the game name (and optionally information about setup, so people can replicate as much as possible) that you tested with? Also, I'm not sure dismissing everything reasonable as obvious is a constructive way to discuss. Academism also doesn't go without measurements and practical discussions.
Before a MR can be approved and merged, the maintainers need to agree that (1) this is a problem worth solving (2) this is a good approach to said problem (3) the code looks good (4) tests pass (whether CI, manual, or both).
If the answer to 1 or 2 turns out to be no, then most effort spent on 3 and 4 will become a waste of time. That's why we're focusing on the former two first.
It's clear that your expectations are different.