Each time a subroutine (or function) is called/referenced, some of the arguments may be put on the stack, and space on the stack committed to them. Furtner, each subroutine or function will itself consume some stack for local variables. When the routine executes a RETURN or end statement, the stack is released and restored to the state prior to the call.
Thus, you can see that it is not the number of subroutines that are called that matters, but the length of the chain of calls. In other words, if your code calls 99 subroutine each of which needs 128 bytes of stack, one after the other, the stack needed is still 128 bytes. However, if you call A, and A calls B, ..., S calls T, then the stack size needed is 24 X 128 bytes. In a realistic case, you will need to add the individual stack needs of the routines in the call chain, since they will not all be the same.
A simple way of finding the stack needs is to compile with more than sufficient stack allocated, fire up the program in the debugger, and record the value of the stack pointer just before the call to A and just before the exit from T, take the difference and add an allowance for safety.
