Comment by alfiedotwtf
5 years ago
I’ve never thought of tail calls in depth before, so quick question about them - the point here I’m getting is to minimise the pushing/popping of registers to the stack from `call` By using a `jmp` instead. That’s all good... but what stack frame does the jumped procedure use? I’m assuming a new stack frame is still created but just without caller’s pushed registers?
There's always some stack frame sitting there; typically the one that called the main loop routine in the first place. At some point a RET instruction will be executed, and control will return to that stack frame's return point. The idea of tail calls is to avoid creating a new stack frame when it doesn't add any value, but stack frames still exist--they have to exist when the last thing that happens in a routine is ordinary instruction processing rather than a CALL.