1. Routine A branches with link (BL) to routine B. This saves its return address, the address of the next instruction, into R14 and then jumps to the routine B’s first instruction.
2. Routine B is going to call another subroutine, routine C. Were it to immediately call routine C using BL then its existing stored R14 would be overwritten. So it stores its R14 to the stack to preserve it.
3. It BL’s to routine C. (R14 ← return address, PC ← Routine C).
4. Routine C does its work, then returns using MOV pc,r14. Its return value is passed back in R0. R1 is corrupted. (R0..R3 are allowed to be corrupted in AAPCS).
5. We return to Routine B after the BL instruction.
6. It restores R14 from the stack.
7. A return is effected and we end up at the instruction after the original BL.

Routine C does not call any subroutines; it is known as a leaf routine. Leaf routines can dispense with the overhead of storing their return address on the stack.