The heart of a computer is the central processing unit or CPU. This device contains all the circuitry that the computer needs to manipulate data and execute instructions. The CPU is amazingly small given the immense amount of circuitry it contains. We have already seen that the circuits of a computer are made of gates. Gates, however are also made of another tiny component called a transistor, and a modern CPU has millions and millions of transistors in its circuitry. The image to the right [Intel 2000] shows just how compact a CPU can be. The CPU is a Pentium® III processor for mobile PCs.

The CPU is composed of five basic components: general purpose registers, special  purpose registers, buses, the ALU, and the Control Unit. Each of these components are pictured in the diagram below. The diagram shows a top view of a simple CPU with 16 bytes of RAM. To better understand the basic components of the CPU, we will consider each one in detail.

• General Purpose Registers: these components are general use memory storage in the CPU that can be accessed very fast (as compared to RAM); they are created from combining latches with a decoder. The latches create circuitry that can remember while the decoder creates a way for individual memory locations to be selected.
  
  
• Special Purpose Registers: these components are dedicated memory storage in the CPU that can be accessed very fast. Three SP registers are shown: the Instruction Register (IR), the Program Counter (PC), and the Accumulator.
  
  
• Buses: these components are the information highway for the CPU. Buses are bundles of tiny wires that carry data between components. The three most important buses are the address, the data, and the control buses.
  
  
• ALU: this component is the number cruncher of the CPU. The Arithmetic / Logic Unit performs all the mathematical calculations of the CPU. It is composed of complex circuitry similar to the adder presented in the previous lesson. The ALU, however, can add, subtract, multiply, divide, and perform a host of other calculations on binary numbers.
  
  
• Control Unit: this component is responsible for directing the flow of instructions and data within the CPU. The Control Unit is actually built of many other selection circuits such as decoders and multiplexors. In the diagram above, the Decoder and the Multiplexor compose the Control Unit.

In order for a CPU to accomplish meaningful work, it must have two inputs: instructions and data. Instructions tell the CPU what actions need to be performed on the data. We have already seen how data is represented in the computer, but how do we represent instructions? The answer is that we represent instructions with binary codes just like data. In fact, the CPU makes no distinction about the whether it is storing instructions or data in RAM. This concept is called the stored-program concept. Brookshear [1997] explains:

"Early computing devices were not known for their flexibility, as the program that each device executed tended to be built into the control unit as a part of the machine...One approach used to gain flexibility in early electronic computers was to design the control units so they could be conveniently rewired. A breakthrough came with the realization that the program, just like data, can be coded and stored in main memory. If the control unit is designed to extract the program from memory, decode the instructions, and execute them, a computer's program can be changed merely by changing the contents of the computer's memory instead of rewiring the control unit. This stored-program concept has become the standard approach used today. To apply it, a machine is designed to recognize certain bit patterns as representing certain instructions. This collection of instructions along with the coding system is called the machine-language because it defines the means by which we communicate algorithms to the machine."

Thus both inputs to the CPU are stored in memory, and the CPU functions by following a cycle of fetching an instruction (and incrementing the program counter), decoding it, and executing it. This process is known as the fetch-decode-execute cycle. The cycle begins when an instruction is transferred from a memory address (identified by the program counter) to the instruction register (IR) along the data bus. In the IR, the unique bit patterns that make up a machine-language instruction are extracted and sent to the Decoder. This component is responsible for the second step of the cycle, that is, recognizing which operation the bit pattern represents and activating the correct circuitry to perform the operation. Sometimes this involves reading data from memory, storing data in memory, or activating the ALU to perform a mathematical operation. Once the operation is performed, the cycle begins again with the next instruction. The CPU always knows where to find the next instruction because the Program Counter (PC) holds the address of the next instruction to be executed. Each time an instruction is fetched, the program counter is advanced to the next instruction memory address.

Each machine instruction is composed of two parts: the op-code and the operands. According to Brookshear [1997], "the bit pattern appearing in the op-code field indicates which of the elementary operations, such as STORE or JUMP, is requested by the instruction. The bit patterns found in the operands field field provide more detailed information about the operation specified by the op-code. For example, in the case of a STORE operation, the information in the operand field indicates which register contains the data to be stored and which memory cell is to receive the data." The image to the right shows the format of an instruction for our CPU. The first three bits represent the op-code and the final six bits represent the operand. The middle bit distinguishes between operands that are memory addresses and operands that are numbers. When the bit is set to '1', the operand represents a number. A simple set of machine instructions for our CPU are listed in the table below. Notice that all the op-codes are given an English mnemonic to simplify programming. Together these mnemonics are called an assembly language. Programs written in assembly language must be converted to their binary representation before the CPU can understand them. This usually done by another program called an assembler, hence the name.