I think that it is good to say that the goto-considered-evil movement was mainly about unchecked goto-statements that jumped to other functions (or procedures back then). And that was very dangerous as there was a very big change you left half a stackframe in memory behind, and that in a language with 60's/70's memory management.

And while I wouldn’t recommend goto for any high-level business logic, it is indeed very true that many high-performance algorithms require jumps to labels to get to most performance out of the machine. And it is kinda of the same story for named continues and break statements.