Editing Loop unrolling (section)

==Disadvantages==
* Increased Code Size: Unrolling increases the number of instructions, leading to larger program binaries. 
** Higher Storage Requirements: The expanded code takes up more memory, which can be problematic for [[Microcontroller|microcontrollers]] or embedded systems with limited storage.
** [[Instruction cache|Instruction Cache]] Pressure: The unrolled loop consumes more space in the instruction cache. If it exceeds the cache size, frequent cache misses can occur, which can cause severe performance degradation due to costly memory accesses.
* Reduced Code [[Readability#Readability in computer programming|Readability]]: If loop unrolling is done manually instead of by an optimizing compiler, the code can become harder to understand and maintain.
* Conflict with Function [[Inlining]]: When the loop body contains function calls, unrolling may prevent inlining due to excessive code expansion, leading to a trade-off between these two optimizations.
* Increased [[Register (computer)|Register]] Pressure: On hardware that relies on software [[Pipelining (computing)|pipelining]] for performance (e.g., systems without register renaming or with [[Superscalar execution|in-order superscalar execution]]), unrolling may require additional registers to store temporary variables across iterations, limiting register reuse.<ref>{{cite journal |author=Sarkar, Vivek |title=Optimized Unrolling of Nested Loops |journal=International Journal of Parallel Programming |volume=29 |issue=5 |pages=545–581 |year=2001 |doi=10.1023/A:1012246031671 |s2cid=3353104 }}</ref>
* Branch Prediction: Modern CPUs use [[Branch predictor|branch prediction]] to try to guess which way a branch will go. If the prediction is correct, the CPU can continue executing instructions without waiting for the branch to resolve. However, if the prediction is incorrect, the CPU has to flush the pipeline and start executing the correct instructions, which can be a performance penalty. Loop unrolling can increase the number of branches in the code, which could lead to more branch mispredictions and lower performance. <ref>Adam Horvath [http://blog.teamleadnet.com/2012/02/code-unwinding-performance-is-far-away.html "Code unwinding - performance is far away"]</ref>