Consider the optimization of inlining: replacing a function…
Consider the optimization of inlining: replacing a function call by expanding the called function’s body into the caller. Here is a simple example in Tiger IR to illustrate its behavior. Without inlining start_function square int square(int x) int-list: x float-list:square: ret = x * x; return ret;end_function square// somewhere else in the program: b = a * 2; r = call square(b); if a < r goto label1; With inlining, the code at the end becomes b = a * 2; r = b * b; // this line changed if a < r goto label1; Assuming no other optimizations occur, what is the performance benefit of inlining a function call? If other optimizations occur, can inlining provide further benefits? If so, describe how. Describe a way in which inlining may worsen performance. Based on parts (1) – (3), give a heuristic that an optimizer could use to decide if an individual function call should be inlined – i.e., the benefit of inlining it is likely to exceed the drawback. Are there cases where a function call cannot be inlined because doing so would cause incorrect or erroneous behavior? Briefly explain why or why not. (15 pts)