ARM compilers insert a prologue at function entry and an epilogue at exit to preserve CPU state. The AAPCS dictates which registers must stay intact, how the stack aligns, and how arguments travel. Without these routines, a call could overwrite the caller’s context, breaking reliability in embedded systems every time.

Take the sample `compute_sum()` on an STM32F407: the prologue pushes r7 and lr, reserves 24 bytes, and sets a frame pointer. Because the function calls `add()`, it is a non‑leaf and must restore lr in the epilogue. A leaf can skip this overhead for simple functions in production code today.

Interrupt handlers and RTOS context switches use a similar stack‑saving dance, but hardware pushes a subset of registers automatically. Developers often write naked functions to bypass compiler prologues when hand‑crafting low‑level routines. Understanding these mechanics lets engineers debug stack corruption and optimize memory usage in mission‑critical firmware for embedded systems.