Registers | ACCA4021

Overview

A register is a tiny storage location inside the CPU. This worksheet explains why registers exist and introduces the five special-purpose registers every Von Neumann CPU uses to fetch and run instructions: the PC, MAR, MDR, CIR and ACC. In the final task you take the place of the control unit and move values between registers by hand. Work through the tasks in order, and try each question before checking it.

Learning Objectives

By the end of this lab you will be able to:

Explain what a register is and why registers are faster than main memory
State the full name and purpose of the PC, MAR, MDR, CIR and ACC
Distinguish registers that hold addresses from registers that hold data or instructions
Use register transfer notation to describe moving a value between registers

Concept

Main memory is large but, compared with the CPU, slow. Every trip along the bus to memory costs time. A register holds a single value, often just one word of 8, 32 or 64 bits, but it is built into the CPU itself, right next to the ALU and the control unit, so reading or writing it is almost instant.

Some registers are general-purpose, and programs can use them for anything. Others are special-purpose, and each has one fixed job in fetching and running instructions. This worksheet is about the special-purpose ones.

Moving a value between registers is written in register transfer notation. The arrow points to the register being written, and square brackets mean "the contents of".

MAR ← [PC]      copy the contents of the PC into the MAR

Tasks

Task 1 - Why registers are fast

The chart below compares how long it takes the CPU to reach a value in different kinds of storage. The scale is logarithmic, so each grid line is 10 times further than the last. Switch to the human scale view to get a feel for the differences.

Why are registers faster to reach than main memory?

A. They use a different kind of binary B. They are inside the CPU, so no trip along the bus to memory is needed C. They hold more data than memory D. They are made of hard disk

Registers are part of the CPU’s own circuitry. Reaching main memory means sending a request along the bus and waiting for the reply.

If registers are so fast, why does a computer not just have millions of them instead of main memory?

A. Registers cannot store binary B. Registers take up space and power on the CPU chip itself, so only a small number fit C. Registers lose their contents every clock tick D. The ALU can only use main memory

Every register is built into the CPU. Space on the CPU chip is limited and expensive, so CPUs have a small number of registers backed by much larger, slower memory.

Task 2 - The special-purpose registers

Register	Full name	What it holds
PC	Program Counter	The address of the next instruction to be fetched
MAR	Memory Address Register	The address of the memory location about to be read or written
MDR	Memory Data Register (also called the MBR, Memory Buffer Register)	The value just read from memory, or about to be written to it
CIR	Current Instruction Register	The instruction currently being decoded and executed
ACC	Accumulator	The result of the most recent calculation by the ALU

The CPU also has a status register that holds the ALU’s flags, such as Z (zero) and V (overflow), which you met in the ALU worksheet.

The diagram below shows a simple CPU part-way through running a program. Click each register to see what it does and which paths it uses to move values around. Then try the two Show buttons.

Task 3 - Addresses or data?

Use the Show buttons in Task 2 to see which registers hold addresses and which hold data or instructions. In this CPU, memory has 16 locations, so an address needs 4 bits, and each memory location holds 8 bits.

The two registers that hold addresses are the and the .

In this CPU the PC is bits wide, because it only ever holds an address.

In this CPU the MDR is bits wide, because it holds a whole memory location.

The diagram shows the CIR holding 0011 0110, which is the instruction ADD .

Why must the MAR be exactly as wide as the address bus?

Every bit of the MAR drives one wire of the address bus. If the MAR were narrower, some address wires would never be used and part of memory could not be reached. If it were wider, the extra bits would have no wires to go out on.

Why does the CPU need both a PC and a MAR? Could the PC be connected straight to the address bus instead?

The address bus is not only used to fetch instructions. When an instruction reads or writes data, the address of the data has to go out on the bus while the PC keeps pointing at the next instruction. Having a separate MAR means the PC never has to be overwritten just to reach a piece of data.

Task 4 - You are the control unit

Each button below performs one register transfer. Complete the three challenges by choosing the right transfers in the right order. Try to use the fewest transfers you can. If you get stuck, press Start again.

Transfer	Meaning
`MAR ← [PC]`	Copy the PC into the MAR
`MDR ← [Memory[MAR]]`	Read the memory location whose address is in the MAR, and put its contents in the MDR
`PC ← [PC] + 1`	Add 1 to the PC
`CIR ← [MDR]`	Copy the MDR into the CIR
`MAR ← operand`	Copy the address part of the instruction in the CIR into the MAR
`ACC ← [MDR]`	Copy the MDR into the accumulator

To fetch an instruction, the first transfer is MAR ← [].

After MDR ← [Memory[MAR]], the instruction is copied into the so it can be decoded.

To carry out LOAD 5, the CPU copies the operand into the MAR, reads memory into the MDR, then copies the MDR into the .

In challenge 1, does it matter whether you do PC ← [PC] + 1 before or after MDR ← [Memory[MAR]]? Explain your answer.

No. Once MAR ← [PC] has been done, the MAR already holds the address being fetched, so changing the PC does not affect the read. It only matters that the PC is incremented after its value has been copied into the MAR. If the PC were incremented first, the CPU would fetch from the wrong address.