And once you start reading the docs on the more efficient but far less consistent 64-bit calling convention, you may find yourself choosing words to describe it other than the "improved" that this author opted for.
I take it you're a fan of the plan9 calling convention (f. ex.: all arguments and return value(s) are on the stack)?
Curiously, it doesn't actually appear all that ineffecient. Go uses it AFAIK. I wonder whether anyone has studied it. I also wonder whether gccgo uses that convention or defaults to SysV x64.
I didn't know it was a System V thing. Thank you for cluing me in!
I don't think of it in terms of like-vs-dislike. My observation is that it's a difficult thing to get right without a compiler, and thus avoided for introductory material.
As far as I am aware, storing values in a register requires fewer instructions. However, I have never personally confirmed the performance difference of this calling convention.
Handling of all calling conventions is one of the many things that I am personally much happier leaving up to GCC and LLVM in practice.
Curiously, it doesn't actually appear all that ineffecient. Go uses it AFAIK. I wonder whether anyone has studied it. I also wonder whether gccgo uses that convention or defaults to SysV x64.