Design of a Simple CPU

11-instruction CPU with 212 (4096) 16-bit RAM.

Illustrates how a general-purpose computer is
assembled from gates and registers.
The design is simplified to make basic concepts
clear.
CPU design can be very complex.
Since the CPU is the heart of every computer,
emphasis is put on making it fast.
November 2013 Simple CPU Design 1
 The Register Set

A CPU is a sequential circuit

• repeatedly reads and executes an instruction from its
memory
• fetch-and-execute cycle

A machine language program is a set of

instructions drawn from the CPU instruction set
• Each instruction consists of two parts: an opcode and
an address

November 2013 Simple CPU Design 2

 11 instructions require a 4-bit opcode.
12 bits of the 16-bit word remain for addressing
212 = 4096 16-bit words of the RAM.
The CPU has 8 special registers:
• 16-bit accumulator (AC)
• 12-bit program counter (PC)
• 4-bit opcode (OPC)
• 12-bit memory address register (MAR)
• 16-bit memory data register (MDR)
• 16-bit input register (INR)
• 16-bit output register (OUTR)
• 1-bit halt register (HLT)

November 2013 Simple CPU Design 3

 The Registers Data Transfer in the CPU

November 2013 Simple CPU Design 4

 The Fetch-and-Execute Cycle

Fetch-and-execute cycle has two portions

Fetch portion is always the same:
• instruction, whose address is in the PC, is fetched into
the MDR
• opcode portion of this register is copied into the OPC

At this point the action of the CPU

diverges, based on the instruction in OPC

November 2013 Simple CPU Design 5

Suppose OPC is a load accumulator instruction.
• the action required is to copy the word specified by the
address part of the instruction into the accumulator
• load accumulator is decomposed into 8
microinstructions executed in 6 microcycles

November 2013 Simple CPU Design 6

 The Instruction Set

November 2013 Simple CPU Design 7

Other operations can be performed using this set:
• subtraction is realized with ADD, CMA, and 2’s comp
• multiplication is possible through the use of CIL and
ADD
• multiple CILs can be used to rotate right one place,
hence division is also possible
• AND and CMA are complete Boolean system, hence
every Boolean function can be realized by this machine
if it is designed to address enough memory locations

November 2013 Simple CPU Design 8

Direct memory instruction
• Addresses refer directly to memory locations containing
the operands (data) on which the program operates

Indirect memory instructions

• Instructions in which an address is interpreted as the
address at which to find the address containing the
needed operand
• CPU does two memory fetches
• the first to find the address of the operand
• the second to find the operand itself

November 2013 Simple CPU Design 9

 Assembly-Language Programming

Instructions stored in memory consist of 16 bits.

• the first 4 are opcode and the last 12 are address

It is hard to interpret such machine-language

statements.
• mnemonic opcodes and assembly languages are
devised

Assembler is a program translating an assembly

language program into a machine language.

November 2013 Simple CPU Design 10

 Example: a Boot Program

Assembly-language program to boot up a

computer is reading another program provided
through the input port and stores the new program
(16-bit words) in memory just above Boot.

When reading is finished (determined by 0 word),

control is transferred to the new program by
jumping to the first location above Boot.

November 2013 Simple CPU Design 11

0
1
2
3
4
5
6
7
8
9

November 2013 Simple CPU Design 12

 The 1st instruction
following the ORG
statement reads the
input register value
into the accumulator.

The 2nd instruction jumps to address 10 if AC==0,

indicating that the last word of the program being
loaded has been written into the memory.
The next executed instruction is in address 10; control
is transferred to the second program.
November 2013 Simple CPU Design 13
 If AC!=0, its value is
stored (indirectly) at
an address (value)
stored at ADDR_2
memory address.

On the first execution of this command the value stored

at memory address ADDR_2 is 10. The content of the
accumulator is therefore stored at location 10.
The next 3 steps increment by 1 the value of ADDR_2
by placing its contents in the accumulator, adding the
value in memory address ADDR_3 (=1) to it and storing
the new value into memory address ADDR_2.
November 2013 Simple CPU Design 14
Finally, the accumulator
is zeroed and a JZ
instruction is used to
return to address
ADDR_1 (=0), the first
address of the boot
program.

The indirect STA instruction is not supported by the simple

computer, but can be simulated by a self-modifying
subprogram.
Though self-modifying programs are considered a bad
programming practice, it illustrates the power of self-
modification and the advantage of having indirection in
the instruction set of a computer.
November 2013 Simple CPU Design 15
 Timing and Control
We need to design circuits controlling the combining and
movement of data.
Assignment notation is introduced.

November 2013 Simple CPU Design 16

 Register transfer notation uses these assignment
operations as well as timing information to break down a
machine-level instruction into microinstructions that are
executed in successive microcycles.

November 2013 Simple CPU Design 17

 microinstructions for the execute portion

November 2013 Simple CPU Design 18

November 2013 Simple CPU Design 19
November 2013 Simple CPU Design 20
November 2013 Simple CPU Design 21
 Generating Control Signals

Define control variables that control the movement of

data between registers or combine the contents of two
registers and assign the result to another register.

Associate a control variable L(A,B) if a microinstruction

results in the movement of data from register B to
register A, denoted A ← B.

L(OPC,MDR) = t3 : OPC is loaded with the contents

of MDR when t3 = 1
November 2013 Simple CPU Design 22
 Associate a control variable L(A,B ⊙ C) if a
microinstruction results in the combination of the
contents of registers B and C with the operation ⊙
and the assignment of the result to register A,
denoted A ← B ⊙ C.

L(AC, AC+MDR) = cADD ∧ t6 : The contents of AC are

added to those of MDR and copied into AC when
cADD ∧ t6 = 1.

November 2013 Simple CPU Design 23

 Control of Accumulator
Group together all the Derive control variables
microinstructions
affecting accumulator

November 2013 Simple CPU Design 24

 Hardware Implementation

LSB: 0 MSB: 15
16 7:1 MUXes

MUX 7:1 MUX 7:1

16-bit ALU ADD AND NOT

16-bit registers
LSB: 0

14
MSB: 15

November 2013 Simple CPU Design 25

 Execution of

LSB: 0 MSB: 15

MUX 7:1 MUX 7:1

ADD AND NOT

LSB: 0

14
MSB: 15

November 2013 Simple CPU Design 26

 Execution of AC  Shift(AC)

LSB: 0 MSB: 15

MUX 7:1 MUX 7:1

ADD AND NOT

LSB: 0

14
MSB: 15

November 2013 Simple CPU Design 27

Group together all
the microinstructions
affecting MAR

Derive control variables

November 2013 Simple CPU Design 28

Group together all
the microinstructions
affecting MDR

Derive control variables

Quiz: What and how many MUXes are required? Ans: 16 MUXes of 2:1

November 2013 Simple CPU Design 29

 Group together all
the microinstructions
affecting a given
register

Derive control variables

November 2013 Simple CPU Design 30

 Group together all
the microinstructions
affecting a given
register

Derive control variables

November 2013 Simple CPU Design 31