Problem #2 (30 points)
Suppose the DLX architecture is modified so that each instruction's execution takes 7 clock cycles. The execute phase is Ex1, EX2. The MEM phase for data memory takes two cycles, M1, M2. This means that the ALU is not internally pipelined, and the memory is not internally pipelined. In other words, there can be one instruction using the ALU or the memory.
IF ID EX1 EX2 MEM1 MEM2 WB
Repeat problem 1 with the above 7-stage pipeline. Indicate the comparators needed to detect the stall if one exists, and the comparators needed for forwarding as well.
1) ADD R7, R4, R3
LW R8, R7(40)
MUL R6, R7, R5
LW R2, R8(40)
SUB R6, R4, R2
SUB R7, R3, R4
LW R8, R740
SW R7(40), R8
ADD R3, R5, R7
LW R5, R3(30)
3) ADDI R4, R3, #5
SUB R8, R3, R5
SW R8(40, R7)
ADD R3, R5, R8
LW R5, R3(30)