Yeah, “also” a performance optimization. It’s also just legacy. We’ve always done it that way so we still do it that way for ABI compat and because it’s hard t

Yeah, “also” a performance optimization. It’s also just legacy. We’ve always done it that way so we still do it that way for ABI compat and because it’s hard to find a compiler that does it any other way. Imagine if the story was: “you totally can have a bounds on your ptrs if you pass a compiler flag and accept perf cost”. I bet some of us would find that useful.