It is misleading to suggest that an implementation of TCO follows directly from the instruction set architecture: it depends even more on what calling conventio

It is misleading to suggest that an implementation of TCO follows directly from the instruction set architecture: it depends even more on what calling convention is followed [1]. gcc supports several calling conventions for x86 [2]. In the Scheme world, often a hybrid stack-heap calling convention called Cheney on the MTA is used in which all calls are tail calls [3]. [1]: https://en.m.wikipedia.org/wiki/Calling_convention https://en.m.wikipedia.org/wiki/Calling_convention [2]: http://redstack.net/blog/x86-calling-conventions.html http://redstack.net/blog/x86-calling-conventions.html [3]: https://news.ycombinator.com/item?id=12776198 https://news.ycombinator.com/item?id=12776198