According to this comment: https://old.reddit.com/r/rust/comments/p0ul6b/when_zero_cost_abstractions_arent_zero_cost/h8b7b2w/ https://old.reddit.com/r/rust/comm

According to this comment: https://old.reddit.com/r/rust/comments/p0ul6b/when_zero_cost_abstractions_arent_zero_cost/h8b7b2w/ https://old.reddit.com/r/rust/comments/p0ul6b/when_zero_cost... , `rustc`/`llvm` is indeed able to optimize the wrapped example to a big `memset`. Sure in this specific case, it is different from the first one which delayed the zeroing, but that is not unintelligently `clone` many times as claimed in the article.