Some architectures write the return address into a register and not onto the stack. The calling function has to preserve it's return address somehow, yes, but t

Some architectures write the return address into a register and not onto the stack. The calling function has to preserve it's return address somehow, yes, but that isn't done for you (well, a compiler does it for you, but not an assembler). So if all parameters are passed via registers and the function does not need more memory than the registers provide for the calculation, then this kind of function does not need to access the stack at all.