In

this lecture, we will examine how the CPU or microprocessor actually execute
programs – the internal working of a computer.

1
In the last lecture we have looked at the overall structure of a computer, which has
three main components: CPU, memory and IO, all connected together via three
buses: address bus, data bus and control bus.
In this lecture we will look at the CPU in some detail.
A CPU has a number of modules inside. These include a arithmetic and logic unit, a
control unit and at least one, but usually many more, registers.
Remember registers are just a bunch of D-FFs.
Modern computers and microprocessors are based on the von Neumann model
which uses memory to store both the instruction codes (i.e. the program) and the
data.
It is important to remember and understand that if you could open the top of a
memory chip and read the values stored inside the chip, you will see ‘1’s and ‘0’s.
What they mean depends on where they are stored and the context. You cannot
tell by just looking at memory which word is an instruction word, and which word is
data. For that, we need some other information.

2
If you look inside a simple CPU, you are going to find these modules:
PC = Program Counter – it stores the address of the NEXT instruction to be
executed
IR = Instruction Register – It stores the current instruction binary code (called
machine code) to be executed
ALU = Arithmetic/Logic Unit – It performs the actual arithmetic or logical
operations
ACC = Accumulator (or result register) – It stores the temporary data or result
In modern CPUs, there are many other modules. For example, instead of having only
one ACC, there could be many more temporary registers. For example, the ARM
processor in the Pyboard has 16 registers (including one for the PC). In addition to
the ALU, there is also a floating point unit (FPU) and even an additional
computational engine known as Adaptive Real-Time Accelerator (ART).
In order for you to understand how a CPU does its job, let us now create one that is
very simple and only having the few modules shown above.

3
This simple CPU uses instruction code that has 16-bits.
We assume that there is only 8k byte of memory and the processor is a 16-bit
processor meaning that it process data in 16-bit words. (ESP32 normally process
data in 32-bits.) Therefore the processor memory is organised as 4k x 16 bits. 4k
words require 12 address bits.
Shown here is the format of the instruction. The top 4 bits are used to specify the
operation to be performed. This is known as the opcode.
With 4-bit opcode, we could specify 16 operations, but we will only create 8
operations with opcode going from 0 to 7.
The remaining lower significant 12 bits in the instruction word are used to specify
the memory address for the operation if required. We will see what that means in
the next few slides.

4
Here is our simple instruction set with only 8 instructions. Here are the explanations
for all 8 instructions. These instructions are shown as three letter code (known as
mnemonics) to indicate what each instruction is suppose to do. This is known as
Assembly Language program. However, the CPU can only understand binary code.
Therefore shown here are the opcodes for each instructions forming the 4 most-
significant bits (MSBs). Then each instruction is followed by a 12-bit memory
address S. The binary form of the instruction is known as Machine Code program.
LDA S - Load Accumulator with the content at memory location S
STO S - Store Accumulator value to memory location S
ADD S - Add the data in memory location S to ACC and store the result back to ACC
SUB S - Subtract the data in memory location S from ACC and store the result back
to ACC
JMP S - Jump to instruction at memory address S
JGE S - Jump to instruction at memory address S if the previous ALU operation is
greater than or equal to zero
JNE S - Same as before but for the case where the previous ALU operation does
not equal to zero
STP - Stop the processor. Since this instruction does not require any memory
address, it only uses the top 4-bits of the instruction word

5
Now let us assume that at memory location 000 (hex) to 003, there are already
stored four instructions. Furthermore, we also assume that in memory locations
02E and 02F (in hex again) stores two numbers AAAA and 1111 to be added
together. The result of this addition, which is BBBB, will be stored back to memory
location 030.
Let us now see how these instructions are executed by the CPU, one instruction at a
time.

6
Firstly, we assume that we start from PC = 000, i.e. on reset, the program counter is
zeroed.
This means that the PC is pointing to address 000 in memory where the first
instruction code is stored. It is 002E, which is LDA 02E instruction, meaning that it
should read the 16-bit data from memory location 02E and stores this in the
accumulator.
So in the first clock cycle, the PC is pushed out via the address register onto the
address bus. A memory read operation is performed and the instruction code 002E
is fetched and put into the instruction register IR. This is the instruction fetch cycle.
The opcode part (top 4-bits) of the instruction is passed through the instruction
decoder unit and the CPU is ready to “execute” the opcode by performing a data-
read-from-memory operation.
This operation is done in the second clock cycle.
The bottom 12-bit of the instruction stored in IR, 02E, is now pushed to the address
register then onto the address bus. The data stored at location 02E (which is AAAA)
is read from this memory location and stored in ACC.
Now something else ALWAYS happens. After the PC is used to fetch an instruction,
its value is ALWAYS automatically incremented. In other words, the PC always
counts up once it is used (hence the name program counter). This is because we
want it to always point to the next instruction.

7
The second instruction is to add the previously fetch data to the data stored in
memory location 02F.
Again in the first clock cycle, the PC value is send to the address bus, and the
instruction code 202F is fetched and stored in IR as shown.
On the next cycle, the opcode (2) is decoded by the CPU and force an add operation
to be performed. This involves the CPU sending the lower 12-bit of the instruction
code (02F) to the address bus and performs a memory read. The data from memory
02F (i.e. 1111) is send, together with the value stored in ACC, to the ALU and the
opcode tells the ALU to do an addition. The result BBBB is stored back to ACC.

8
In this instruction, the contents of ACC is stored to memory location 030.
Finally, the last instruction is to tell the CPU to stop. This instruction does not involve any
memory read AFTER fetching the instruction. It therefore only use the top four bit of the
instruction for opcode, and it only takes one clock cycle. This is shown on the next slide.
To summarise
The operation of most processors are governed by a clock signal. For this simple CPU, we
assume that:
1.  The number of clock cycles taken by an instruction is the same as the number of
memory access it makes.
2.  LDA, STO, ADD, SUB therefore takes 2 clock cycles each: one to fetch (and decode)
the instruction, a second to fetch (and operate on) the data
3.  JMP, JGE, JNE, STP only need one memory read and therefore can be executed in one
clock cycle.
4.  Program counter (PC) - its content is incremented every time it is used (i.e. it also
points to the next instruction).
5.  The processor must start from a known state. Therefore, there is always a reset
signal to initialise the processor on power-up.
6.  Assume MU0 will always reset to start execution from address 00016.

9
8.  Microprocessors performs operations depending on instruction codes stored
in memory
9.  Instruction usually has two parts:
•  Opcode - determines what is to be done
•  Operand - specifies where/what is the data
10.  Memory contains both program and data. A peek into memory will tell you
very little except a bunch of ‘1’s and ‘0’s
11.  Program area and data area in memory are usually well separated
12.  ALU is responsible for arithmetic and logic operations
13.  There is always at least one register known as accumulator where the result
from ALU is stored
14.  There is usually one or more general purpose register for storing results or
memory addresses temporarily
15.  Fetching data from inside the CPU is much faster than from external memory

10
This video explains how a CPU or a microprocessor is made. You can find this video
on youtube:

https://www.youtube.com/watch?v=qm67wbB5GmI

11