Exception entry and exit sequence in ARM Architecture - Time & Space Complexity
When the processor handles an exception, it runs a special sequence of steps to save its state and jump to the exception handler.
We want to understand how the time taken by this sequence changes as the system or input changes.
Analyze the time complexity of the following ARM exception entry and exit sequence.
SUB SP, SP, #32 ; Make space on stack
STMFD SP!, {R0-R3, LR} ; Save registers
LDR R0, =ExceptionHandler
BX R0 ; Branch to handler
; Exception exit sequence
LDMFD SP!, {R0-R3, LR} ; Restore registers
ADD SP, SP, #32 ; Restore stack pointer
BX LR ; Return from exception
This code saves some registers and stack space on exception entry, then restores them on exit.
Look for repeated or costly steps in the sequence.
- Primary operation: Saving and restoring multiple registers using STMFD and LDMFD instructions.
- How many times: Each register saved/restored counts as one operation, done once per exception.
The time depends on how many registers are saved and restored.
| Input Size (number of registers) | Approx. Operations |
|---|---|
| 4 | 10 (5 saved + 5 restored) |
| 8 | 18 (9 saved + 9 restored) |
| 16 | 34 (17 saved + 17 restored) |
Pattern observation: The time grows linearly with the number of registers saved and restored.
Time Complexity: O(n)
This means the time to enter and exit an exception grows directly with how many registers are saved and restored.
[X] Wrong: "The exception entry time is always constant regardless of saved registers."
[OK] Correct: Saving more registers means more instructions run, so the time increases with the number of registers.
Understanding how exception handling time grows helps you write efficient low-level code and reason about system responsiveness.
"What if the exception sequence saved only the minimum registers needed? How would that affect the time complexity?"