Very true! Hence the part where I'm thinking out loud regarding how the gc ordering of data after a copy/compact (in the one generation gc) impacts locality. And I suppose this also applies to more modern gc setups like multiple generations or parallel collection. Is there some notion of data locality that we can formalize and analyze for at least simplified versions of modern memory hierarchy? How well do standard gc algorithms perform when analyzed thusly? How efficient can each style of gc algorithm be if it's designed to optimize locality?

It clearly has some relation to optimal caching optimization problems, but with more explicit interdependencies. It's also seemingly both an online and offline problem in terms of how to analyze it, but I need to think on it more