Computed gotos, TCO, threaded interpreters: experiments and findings

My post isn’t arguing for computed gotos in Rust.

It was an exploration of what assembly rustc can generate and whether there are still benefits these days from gotos, since the topic has come up on occasion in this forum and urlo.

I like learning by doing and so my conclusion from benchmarking is definitely “it depends.” In the case of Rust, I don’t see any value in building gotos into the language, excepting the possible case of become but that’s a different story.

Thanks for the explanation of how clang/llvm works in this case, that is really interesting. However, I don’t see at all how your conclusions, however true, derive from that explanation…