You'll hand in this assignment on paper. However, we request that you use your favorite word processor, as hand-written assignments are difficult to read. Please put the section name (AA or AB) on your assignment, along with your name.

1. Consider executing the following code on the pipelined datapath shown in the book in Figure 6.46 (new book) or Figure 6.48 (old book).

lw  $3, 0($5)
lw  $4, 4($5)
add $7, $7, $3
add $8, $8, $4
add $10, $7, $8
sw  $6, 0($5)

For this code fragment, rewrite the code to minimize performance - that is, reorder the instructions so that this sequence takes the most clock cycles to execute in our pipelined datapath while still obtaining the same result.

2. Consider executing the following code on the pipelined datapath shown in the book in Figure 6.46 (new book) or Figure 6.48 (old book).

add $1, $2, $3
add $3, $4, $2
sub $5, $2, $4
and $1, $6, $3
or  $5, $7, $4

• In the fifth cycle of execution, which registers are being read and which are being written?
• Explain what the forwarding unit does in the fifth cycle of execution. If any comparisons are done, explain what values are used in those comparisons and what the results of the comparisons are.
• Explain what the hazard detection unit does in the fifth cycle of execution. Again, explain the values used in any comparisons and the results of those comparisons.

3. Consider the following code fragment:

      lw   $t1, 0($a0)
      addu $v0, $v0, $t1
      lw   $t1, 4($a0)
      sub  $t1, $t2, $t1
      addu $v0, $v0, $t1
      bne  $t1, $t2, lab
      add  $t1, $t1, $t3
      add  $t1, $t1, $t4
lab:  sw   $t1, 4($a0)

• How many cycles would it take to complete the above code fragment on the MIPS pipeline when the branch is taken, and how many when the branch is not taken? You should assume that the forwarding and hazard detection hardware are implemented. For control hazards, you should assume that the branch decision and target are computed in the MEM stage, and you should also assume that the "assume branch not taken" optimization has been implemented.
• Now assume that all the forwarding hardware in the pipeline is removed. Now how many cycles does the code fragment take to complete? Again, give one answer for when the branch is taken, and one for when it isn't.

4. Following the rules specified in the calling conventions handout, write the procedure prologue for each of the following procedures. When deciding on the framesize, use the minimum required space.

• A procedure named proc1. Proc1 is a leaf procedure that uses registers $s1 and $s2, and does not need any local or temporary stack storage.
• A procedure named proc2. Proc2 is a procedure that calls 2 other procedures, one of which takes 2 arguments, the other takes 3 arguments. Proc2 doesn't use any s-registers, nor does it have any local or temporary stack requirements.
• A procedure named proc3. Proc3 is a procedure that calls 3 other procedures (A, B, and C). A takes no arguments, B takes 6 arguments, and C takes 2 arguments. Proc3 uses registers $s0, $s2, and $s5. Proc3 needs to save three of its temporary registers on the stack.
• A procedure named proc4. Proc4 is a leaf procedure, uses $s0 and $s1, and has 2 local variables - an integer, and an array of 5 characters.

5. Suppose you are implementing a procedure that takes 7 integer arguments, and suppose that the framesize of your procedure is 44 bytes. Assume that the caller has placed those 7 arguments according to the calling conventions document, and that your code fragments come after the procedure prologue (so the stack pointer has already been adjusted).

• Write a MIPS code fragment that calculates the sum of all 7 arguments and leaves the result in register $t0.
• Suppose you need to save registers $a2 and $a3 on the stack, because you're about to call another procedure that use $a2 and $a3 for something else. Write the MIPS store instructions that save $a2 and $a3 on the stack.