I'm currently preparing for an exam in compiler construction. The course is mainly about optimization in the backend of the compiler. For the optimization we have to do data flow analysis by hand. In one example I do not understand how gen and kill sets are determined. My solutions differ a lot although I stuck to the transfer functions for reaching definitions. The analysis is structured by base blocks.
Code:
1: i=2;
L1:
2: if( i <= 220 ) goto L4;
L2:
3: t4 = i*8;
4: t5 = base() + t4;
5: t6 = t4 + j;
6: j = j+1;
7: if( t6 >= 120 ) goto L3;
8: t5 = t6 + t4;
L3:
9: i = t5 + 1;
10: goto L1;
L4:
11: return i;
Base Blocks:
- BB1 = {1}
- me: gen={1} kill={2,3}
- solution: gen={1} kill={2,3}
- BB2 = {2}
- me: gen={} kill={}
- solution: gen={2} kill={}
- BB3 = {3, 4, 5, 6, 7}
- me: gen={3,4,5,6} kill={4,5,6,8,9}
- solution: gen={3,4,7} kill={5,6,8,9} ???
- BB4 = {8}
- me: gen={8} kill={4,9}
- solution: gen={8} kill={9}
- BB5 = {9, 10}
- me: gen={9} kill={3}
- solution: gen={9} kill={2,3}
- BB6 = {11}
- me: gen={} kill={}
- solution: gen={} kill={}
The gen and kill set are used to compute the in and out set. In and out set are then used to compute reaching definitions.
Can somebody explain to me why the sets differ by so much?
