EXERCISES

1. The GNU gcc compiler provides a number of options to control the nature of optimization that is attempted on a given source code. Study the man-page of gcc to get acquainted with the optimization options available. Remember that some of the options are specific to a particular processor. Try to test all possible situations where various techniques of optimization are applicable. At the minimum, compile the following C programs with a command like gcc <optimization options> −S mytest.c, using O1, O2, O3, Os and “unwinding loops” options, to obtain the assembly language output.

   (a)  int main(){
   	int i, j, k;
   	k = 5;
   	i = (k + 5) * (k + 5);
   	j = k + 5;
   	printf(“%d %d %d”, i, j, k);
        }
   (b)  int main(){
   	int i, j, k;
   	for(i = 0; i < 4; i++){
   	    j = 9;
   	    k = i + 4;
   	}
        }
   (c)  int main(){
   	int i, j, k;
   	for(i = 0; i < 4; i++){
   	    j = 9;
   	    k = i + 4;
   	printf(“%d %d %d”, i, j, k); 
   	}
        }
   

2. Suresh wrote an optimizer for his compiler which did constant folding and propagation and transformed

        i = 0;
   LOOP: i = i + 1;
         …   …
         if(i < 10) goto LOOP;

   into:

        i = 0;
   LOOP: i = 1;
         …   …
         if(1 < 10) goto LOOP;
   

   What is wrong, if any, with this optimization? What is the basic problem?
3. Apply as many optimization methods as are applicable to the following 4-tuple IR code:

   1:  SUB  4    2  T1
   2:  DIV  T1   2  T2
   3:  MUL  a   T2  T3
   4:  MUL  T3  T1  T4
   5:  ADD  T4   b  T5
   6:  MUL  T3  T1  T6
   7:  ADD  T6   b  T7
   8:  MUL  T5  T7  c