CS2100

NATIONAL UNIVERSITY OF SINGAPORE

CS2100 – COMPUTER ORGANISATION

(Semester 2: AY2016/17)

Time Allowed: 2 Hours

INSTRUCTIONS TO CANDIDATES

1. This assessment paper consists of SIX (6) questions and comprises TWELVE (12) printed
pages.
2. This is a CLOSED BOOK assessment. One handwritten double-sided A4 reference sheet is
allowed. Calculators are not allowed.
3. Answer all questions and write your answers in the ANSWER BOOKLET provided.
4. Fill in your Student Number with a pen clearly on your ANSWER BOOKLET.
5. You may use pencil to write your answers.
6. Page 9 contains the MIPS Reference Data sheet and page 10 contains the MIPS Datapath.
7. Pages 11 and 12 are for your rough work.
8. You are to submit only the ANSWER BOOKLET and no other document.

Page 1 of 12
CS2100

1. [10 marks]
A theme park offers locker rental to its visitors. To use a locker, a visitor deposits 4
tokens, one at a time, into the locker’s token slot.
Design a sequential circuit with states ABC for the locker’s door using a D flip-flop for A,
a T flip-flop for B, and a JK flip-flop for C. The circuit consists of 5 states representing the
number of tokens a visitor has deposited: 0, 1, 2, 3 and 4 (or, in binary, ABC = 000, 001,
010, 011 and 100). The visitor can deposit only one token at a time. When the circuit
reaches the final state 4, it remains in state 4 even if the visitor continues to put tokens
into the slot.
Let the external input t denotes a token.
a. Complete the given state diagram on the Answer Booklet. The state values are
shown in decimal. The value on the arrow represents t. [2 marks]

0 1 2 3 4

b. Write the simplified SOP expressions for the flip-flop inputs. [8 marks]

2. [10 marks]
a. Given the following Boolean function:
F(A,B,C,D) = m(1, 4, 5, 6, 7, 13)
You are to implement F using a single 2-bit magnitude comparator with no
additional logic gates. Note that complemented literals are not available. [5 marks]

b. Given the following Boolean function:

G(A,B,C,D) = m(2, 11)
You are to implement G using a single 2×4 decoder with one-enable and active high
outputs, and one 2-input exclusive-OR gate. Note that complemented literals are
not available. [5 marks]

Page 2 of 12
 CS2100

3. [10 marks]
Study the following MIPS program. Arrays A and B are integer arrays.

# $s0 contains the starting address of array A

# $s1 contains the starting address of array B
add $s2, $s0, $zero #inst 1
add $s3, $s1, $zero #inst 2
L1: lw $t0, 0($s2) #inst 3
lw $t1, 0($s3) #inst 4
bne $t0, $zero, L2 #inst 5
beq $t1, $zero, done #inst 6
L2: slt $t2, $t0, $t1 #inst 7
beq $t2, $zero, L3 #inst 8
sw $t0, 0($s3) #inst 9
sw $t1, 0($s2) #inst 10
L3: addi $s2, $s2, 4 #inst 11
addi $s3, $s3, 4 #inst 12
j L1 #inst 13
done:

a. Give the instruction encoding in hexadecimal for instruction 1 (add $s2, $s0, $zero).
The opcode for add is 0 and the funct value for add is 0x20. [2 marks]

b. Give the instruction encoding in hexadecimal for instruction 6 (beq $t1, $zero,
done). The opcode for beq is 0x04. [2 marks]

c. If instruction 13 (j L1) is at memory address 0xE0480030, give the instruction

encoding in hexadecimal for instruction 13. The opcode for jump is 0x02.
[2 marks]

d. The following are the initial values of the array elements in arrays A and B.

Array A: 3 -1 7 0 2 -5 9 0 -9 1

Array B: 2 5 7 3 6 0 8 0 -3 2

Fill in the final values of the elements. [4 marks]

Page 3 of 12
CS2100

4. [35 marks]
Zephyr is a 32-bit stack-based processor with the specifications shown below. In this
table, registers “x” and “y” serve as placeholders for actual general purpose registers
$1, $2, …, $8, the capital letter “V” refers to a single variable, the capital letter “A”
refers to the first element of an array, and the small letter “c” refers to a constant. The
capital letters “X” and “Y” refer to the top-most and second elements of the stack
respectively. All constants and displacements are 2’s complement signed values.

Addressing Architecture: Stack based.

Number of General Purpose Registers: Eight ($1, $2, …, $8)
Special registers: Stack pointer ($sp)
Program counter ($pc)
Instruction formats: Fixed length 32-bit instructions
Arithmetic instructions: ADD: X and Y are popped off the stack,
X+Y are pushed back onto the stack.

Arithmetic instructions. X is the top-most SUB: X and Y are popped off the stack,
operand on the stack, Y is the next X-Y is pushed back onto the stack.
operand in the stack. MUL: X and Y are popped off the stack,
X*Y is pushed back onto the stack.
DIV: X and Y are popped off the stack.
X/Y is pushed back onto the stack.

Stack instructions: PUSHI c: Push immediate value c onto

the stack.
PUSH $y: Push register y onto the stack.
Stack manipulation instructions. Special
register $sp points to the top-most POP $y: Pop topmost item on the stack
element of the stack. Stacks are assumed into register y.
to be arbitrarily large, while popping an ZERO: Reset $sp to bottom of stack.
empty stack will cause an error, but we
WILL NOT consider that here. We will
assume that the stack is never empty nor
full.

Page 4 of 12
 CS2100

4. (continued)
Load/Store Instructions: LW $y, $x: Load 32-bit word stored in
address pointed to by register x into
register y.
These are load and store instructions
that get data from memory to registers SW $y, $x: Store 32-bit word in register
y, into the address pointed to by register
and vice versa.
x.
LB $y, $x: Load a single byte stored in the
All addresses are byte addresses. address indicated by register x, into the
lowest (bits 7-0) bits of register y.
SB $y, $x: Store the lowest 8-bits (bits 7-
0) of register y into the byte address
indicated by register x.
LDI $y, c: Store immediate constant c
into register y.
LDI $y, V: Store address of variable V into
register y.
LDI $y, A: Store base address of array A
into register y.
INCW $y: Register y is incremented by 4.
DECW $y: Register y is decremented by
4.
INC $y: Register y is incremented by 1.
DEC $y: Register y is decremented by 1.

B-type instructions: BEQ $x, $y, displ: Jump to address

($pc+4) + 4*displ if register x == register
y.
These are compare and branch
BNE $x, $y, displ: Jump to address
instructions.
($pc+4) + 4*displ if register x != register y
BLT $x, $y, displ: Jump to address
($pc+4) + 4*displ if x < y
BGT $x, $y, displ: Jump to address
($pc+4) + 4*displ if x>y

Page 5 of 12
CS2100

4. (continued)
a. Using the instruction set given above, write the Zephyr assembly language
equivalent of this program. Ensure that your code is properly commented. [5 marks]

for (i=0; i<5; i++)

if (x[i] < 3)
x[i] = x[i] + 5;

All offsets in Zephyr are expressed as 16-bit word addresses, while registers are
expressed as 3-bit register numbers (0002=$1, 0012=$2, …, 1112=$8). Similarly, all
constants in Zephyr are 16-bit long.
There are six classes of instructions:
A: No operands (e.g. ADD)
B: One register operand (e.g. PUSH $1)
C: One constant operand (e.g. PUSHI c)
D: One register and one constant operand (e.g. LDI $1, c)
E: Two registers (e.g. LW $1, $2)
F: Two registers and a displacement (e.g. BEQ $1, $2, displ)

b. Sketch the instruction formats for all 6 classes, assuming that all 32 bits of a Zephyr
instruction word are utilized fully, and that we maximize the number of opcode bits
possible each time. [12 marks]

c. If we utilize an expanding opcode scheme for Zephyr, what is the maximum number
of opcodes possible, assuming that there are at least one instruction in each class?
Show your working and reasoning process. Where convenient you may leave your
answers in terms of powers of 2. [5 marks]

d. What is the minimum number of opcodes possible, assuming that there are at least
one instruction in each class? Show your working and reasoning process. Again,
where convenient you may leave your answers in terms of powers of 2. [5 marks]

e. What is the furthest forward distance that you can branch to, in the BEQ, BNE, BLT
and BGT instructions? Express your answer in number of instructions. [2 marks]

f. Suppose that we have an array A of words (i.e. we access elements of A one word at
a time). What is the maximum size of A, expressed in words? [3 marks]

g. Suppose again that we have an array B of bytes (i.e. we access elements of B one
byte at a time). What is the maximum size of B, expressed in bytes? [3 marks]

Page 6 of 12
 CS2100

5. [15 marks]
In this question we want to modify the (non-pipelined) MIPS datapath to support two
new instructions: BLT and BGT – “branch on less than” and “branch on greater than”.
The BLT and BGT instructions are shown below:
BLT:
0x08 rs rt displ

BGT:
0x12 rs rt displ

a. The ALU for the MIPS processor is shown below as a single block with two 32-bit
inputs, and one 32-bit output. Show, by adding AT MOST ONE 32-input logic gate
and any additional wires, how to generate the IsZero and IsNegative signals. The
IsZero signal is 1 when ALUIn1 – ALUIn2 is zero, and the IsNegative signal is 1 when
ALUIn1 – ALUIn2 is negative. [4 marks]

32
ALUIn1

32

ALUOut

32
ALUIn2

ALUControl

b. The CONTROL unit in the datapath must now generate BranchLess and
BranchGreater signals from the instruction bits. Sketch the combinational circuits to
generate these signals. [5 marks]

c. For your convenience the MIPS datapath is shown in page 10. Sketch the
combinational logic circuit needed to generate the PCSrc control signal to support
the BEQ, BLT and BGT instructions. [6 marks]

Page 7 of 12
CS2100

6. [20 marks]
Suppose we have a cache that has an access time of 5ns, and a main memory with an
access time of 80ns.
a. What is the memory access time when you have a cache hit? [2 marks]

b. What is the memory access time when you have a cache miss? [3 marks]

c. You run some benchmarks on your system and find that 10,000 accesses take a
total of 70 microseconds (1 microsecond = 1000 nanoseconds). What is the miss
rate of your cache? [5 marks]

Your cache is implemented as a 4-way set associative write-back cache totaling 64KB.
Each cache block holds 8 words of 4 bytes each. CPU addresses are 32 bits long.
d. How many bits per set do you require to store the tags? [4 marks]

e. Assuming that the 64KB of cache refers purely to “usable cache” – i.e. cache that
is used only to store data or instructions, and not overheads like tag bits, what is
the total amount of static RAM that you require to implement this cache?
[6 marks]

~~ END OF PAPER ~~~

(The next two pages contain the MIPS Reference Data sheet and the MIPS Datapath.)

Page 8 of 12
 CS2100

Page 9 of 12
 CS2100

Page 10 of 12
 CS2100

(This page is for your rough work.)

Page 11 of 12
CS2100

(This page is for your rough work.)

Page 12 of 12