BCS: > I read that as, "the DMD backend *can't* do tail call elimination" because, > given how long Walter has been writing compilers, if it could, I'd be > surprised > if Walter wouldn't have added it yet. It may be a cases of "ya can't get > that from here" or some such that would requiter a major rewrite for a > relatively > minor gain.
I have already answered this in a recent post, but the short answer is: there are many ways to "cheat" at this; that is to perform a limited form of this optimization even if a backend isn't able to perform it. So what you say doesn't hold. Bye, bearophile
