The Digital Computer

The American Heritage Dictionary defines a computer as

Historical Development

The ENIAC at the Ballistic Research Lab in Philadelphia Pennsylvania, circa 1940s/1950s.

Development on the devices that led to the modern definition of a computer began in the 1940s as part of the war effort. Early electronic computing devices, such as the ENIAC (Electronic Numerical Integrator and Computer), were programmed by connecting wires to various ports or plugs. There was no concept of stored instructions. To change the actions of the computer (i.e. reprogram the computer), it had to be rewired for the new task, which could be very tedious.[1]

John W. Mauchly and J. Presper Eckert, the inventors of the ENIAC, proposed a design for a new device that could store the instructions to be executed instead of having to rewire the computer for each new task. They were not, however, able to publish their idea for the EDVAC (Electronic Discrete Variable Automatic Computer) due to the war and their involvement in the topic secret ENIAC project. The initial design for the EDVAC was published instead by John von Neumann (pronounced von noy-man), who was involved in the ENIAC project, but did not have the same restrictions.[2][3]

The EDVAC, circa 1940s/1950s.

The EDVAC was ultimately constructed and delivered to the U.S. military in late 1949. The completed computer contained an ultrasonic serial memory, which had a capacity equivalent to 5.5K bytes, and a magnetic tape reader-recorder. It also contained other units similar to those used in modern computers. These included a control unit, a computational unit to perform arithmetic operations, and a dispatcher unit that received instructions from the control and memory units and directed them to other units.

Although the design of the EDVAC was the first proposed stored-program computer, the British completed construction on the actual first two stored-program computers based on the design published by von Neumann. The Manchester Baby was built at the University of Manchester in June 1948. It served as a testbed device that ultimately led to the Manchester Mark I, which was placed into regular service in April 1949. The EDSAC (Electronic delay storage automatic calculator) was constructed at the University of Cambridge Mathematical Laboratory in May 1949.[2]

The Manchester Mark I, circa 1940s/1950s.

The EDSAC, circa 1940s/1950s.

The architecture described by von Neumann in 1945 (in his famous paper, First Draft of a Report on the EDVAC), which led to these early computers, is today known as the von Neumann architecture. All modern stored-program computers are based on this architecture.

Components of a Computer

Modern stored-program digital computers include at least three characteristics of the von Neumann architecture:

• Comprised of three hardware components as illustrated by the schematic diagram in Figure 1: (1) a processor that contains a control unit, an arithmetic logic unit, and registers; (2) a main memory system; and (3) an Input/Output system.
• Has the ability to execute or carry out sequential instruction processing. That is, execute one instruction and then move to the next instruction without intervention.
• Contains a single path or bus between the processor and main memory and the processor and the input/output system. By sharing a single bus, fetching an instruction from memory can not be performed at the same time as a data operation. This limitation is known as the von Neumann bottleneck.

Figure 1. Basic components of a modern computer.

Special Topic

The Modern PC Architecture

The schematic diagram presented earlier is a simplified version that illustrates the components and the basic connections between the components. Modern personal computers use a slightly modified architecture as illustrated below: Some of the operations that would have been performed by earlier processors was moved to other controllers. The southbridge (or I/O controller hub) handles most of the computers's I/O functions and corresponds to the I/O System of the original design. The northbridge, is a memory controller that was introduced to manage access to the main memory. In this newer design, the processor does not connect directly to the main memory or the I/O system. Instead, the processor connects directly to the northbridge and the northbridge connects to the other two components. Thus, it acts like a gateway to the processor. The result is a modern design that allows for faster processing. I/O operations, which can occur at any time, are relatively slow compared to the arithmetic and logic operations performed by the ALU. For example, consider the use of a network adapter and the arrival of a message from the network. In the original design, the processor would have to pause the operations it was performing and turn its attention to the network adpater in order to receive the message. That in turn reduces the computational power of the processor. With the use of the memory controller, I/O devices can access main memory directly without having to go through the processor. In our network example, messages that arrive from the network are written directly into a pre-defined area of memory and the operating system (OS) is notified via a signal. This allows the OS (and in turn the processor) to handle the input on its own schedule. A similar technique is used for output as well. When you write to a file on disk, for example, the processor does not have to handle the actual output. Instead, the data is written to a pre-defined location or buffer in memory and the OS driver of the disk drive is notified via a digital signal. This instructs the device to read the data directly from memory and and write it to the disk.

Processor

Traditionally, we refer to the processor as the central processing unit or CPU for short. The term central was introduced early in the history of digital computers when most of the processing functions were centralized on a single component or unit. Modern systems can contain single processors with multiple cores or contain multiple processors with multiple cores. Thus, the term processor is more commonly used today to reflect this change in architecture, but the term CPU is still widely used.

The processor contains the digital circuitry that executes the instructions of a program. The individual instructions perform simple operations, which typically include those for arithmetic, logic, and memory access. In later chapters, you will learn about many of the instructions used by the MIPS processor and see how some of them are implemented in the hardware. As indicated earlier, the processor is comprised of several common components as described below.

The arithmetic-logic unit (ALU) performs all arithmetic, logic, and bit-wise operations, which comprise the bulk of the instructions contained in a computer program. Originally, processors could only carry out integer operations. Later, a separate component, known as a floating-point unit (FPU), was included for floating-point operations. Both of these units will be discussed in more detail in later chapters.

The processor contains a small number of high-speed storage elements called registers, which are used to store temporary values. The registers are wired directly to the ALU and provide fast access as instructions are being executed. Modern processors also contain several special registers used in the execution of instructions. The instruction register (IR) holds the instruction currently being executed and the program counter (PC) register keeps track of the next instruction to be executed.

The control unit coordinates the activities of the other units. It decodes the instructions and sends signals to direct other units in carrying out or executing the instruction. While it is easier to view the control unit as a single well-defined component, in modern computers, the digital circuitry of the controller is distributed throughout the processor and computer.

Memory System

The program instructions and data are stored in the memory unit. A computer's memory is constructed using semiconductor storage cells, each of which is capable of storing a single bit of information. These storage cells are grouped into larger fixed sizes, typically consisting of eight bits, that are identified by a distinct numerical address. Access to the memory unit is managed through the system bus, which is a digital communications channel that connects the various components of a computer. Accessing data in memory is many times slower than the direct access to the registers on a processor, but memory is much cheaper to construct than are the registers. Thus, the amount of memory in a computer is far greater than the number of available registers.

In modern computers, programs are typically stored on a secondary storage device such as a hard drive or some type of compact disk. To be executed, however, a program must reside in the computer's memory unit, which is commonly referred to as primary storage or the main memory. As the program is executed, the processor fetches the instructions, one at time from memory and carries out the appropriate action. We will explore the memory unit and secondary storage in later chapters.

Input/Output System

Computer programs typically follow a common process: receive or input data, process it, then output the results. The input/output system manages the various input/output (I/O) devices that are used to provide input to a program and to receive the output or results from the program. Today, there are a wide variety of I/O devices. Some provide input only, such as keyboards, mice, and joysticks; some only accept output from a program, such as video terminals and printers; while others provide input and accept output, such as disk drives and network adapters. In modern computers, there is a separate controller used to manage the I/O system, which frees up the main controller and processor to perform their main tasks.

Special Topic

A Look Inside

Early computers were comprised of multiple large units that were physically separated due to the use of vacuum tubes and the excessive heat they produced. These computers required large rooms to house all of the various components. With the invention of the integrated circuit or microchip, the size of computers began to shrink and their speed greatly increased. Integrated circuits are constructed using transistors (transfer resistor), which are far smaller than vacuum tubes. Early integrated circuits included dozens of transistors on a single chip. As manufacturing and chip technology advanced, the number of transistors packed onto a single small chip greatly increased. Today, microchips can contain 100,000 or more transistors. It is this advance in microchip technology that allowed for the introduction of personal computers and the more modern mobile devices. Modern computers are still constructed using similar components as earlier computers, but their size has greatly reduced. Today, the components of a computer are integrated onto a motherboard that is contained within a single physical structure. As illustrated in the image below, the system bus and other communication channels are "printed" onto the board and the various components are physically connected to them. The size of the motherboard can vary depending on how the computer will be used and the number and type of peripheral devices supported. Identifying the various components can be difficult, even if you are familar with the components and their functions. Most motherboards that support the Intel architecture, like the one illustrated above, have a similar design and include similar components. The construction of an actual computer is far more complex than the simple schematic design presented earlier.

The Fetch-Execute Cycle

The construction and implementation of a processor can be rather complex, but the actual operation of a processor is rather simple. During the entire time a processor is on (i.e. has electricity running through it), it performs a simple routine called the fetch-execute cycle, which consists of four steps:

• fetch the next instruction
• decode the instruction
• execute the instruction
• repeat

When a program is executed, the machine code instructions are loaded into the computer's memory and the processor is directed to begin executing the first instruction. The first step is to fetch or read the instruction from memory. The fetched instruction is then decoded to determine what operation is to be performed and what data will be used. Next, the instruction is executed or carried out using the logic circuits of the processor. After the operation completes, the entire cycle repeats again with the next instruction.

As you can see, the high-level view of the operation of a processor is rather simple. The actual details, however, are a bit more complex. In the first part of this book, you will learn to write simple programs in assembly language that are assembled into machine code and executed on a MIPS processor. Later in book, you will get an up close view of the internals of a processor and learn how the processor is constructed and how instructions are actually "executed."

Example

To illustrate the fetch-execute cycle, suppose we have a simple program that evaluates the mathematical expression

which is done using two machine code instructions

When this program is executed, the first instruction is fetched from memory and then decoded by the control unit. The control unit will discover that this instruction is a multiply operation. The equivalent instruction in assembly is

Machine code instructions are very basic and perform a single operation. Here, the value in register a1, which holds the value from variable b, is multiplied by 4 and the result is stored in register v0. After this instruction is complete, the cycle repeats with the next instruction.

After the next instruction is fetched and decoded, the control unit will discover that it is an add operation, the assembly equivalent of

When executed, the instruction will add the value in register a0, which holds the value from variable a and adds it to the value in register v0. The result is stored back into register a0.