2. (6 points - Correctness) More Hazards. Given the following program: loop: lw $t2, 4($t0) ori $t2, $t2, 8 SW $t2, 0($t0) addi $t0, $t0, 4 bne $t0, $t1, loop add $t4, $t3, $t5 (a) Rearrange and/or modify (do not remove instructions, the number of instructions should remain the same) the instructions in the given program to increase performance Assume a 5-stage pipeline processor with forwarding wherever possible. Also, assume that branches are resolved in Decode, branches are always predicted not-taken, and no architectural branch delay slot. Make sure that you maintain the functional correctness of the program after your changes. (b) Compare the average steady state CPI of the 'loop', before (given program) and after the modifications (part (a)). Assume that branches are resolved in Decode, branches are always predicted not-taken, and no branch delay slot. If necessary, consider the case when the branch was actually taken 100 times before it left the loop. 2. (6 points - Correctness) More Hazards. Given the following program: loop: lw $t2, 4($t0) ori $t2, $t2, 8 SW $t2, 0($t0) addi $t0, $t0, 4 bne $t0, $t1, loop add $t4, $t3, $t5 (a) Rearrange and/or modify (do not remove instructions, the number of instructions should remain the same) the instructions in the given program to increase performance Assume a 5-stage pipeline processor with forwarding wherever possible. Also, assume that branches are resolved in Decode, branches are always predicted not-taken, and no architectural branch delay slot. Make sure that you maintain the functional correctness of the program after your changes. (b) Compare the average steady state CPI of the 'loop', before (given program) and after the modifications (part (a)). Assume that branches are resolved in Decode, branches are always predicted not-taken, and no branch delay slot. If necessary, consider the case when the branch was actually taken 100 times before it left the loop