The von Neumann model

In the earliest electronic computing machines, programming was synonymous with connecting wires to plugs. No layered architecture existed, so programming a computer was as much of a feat of electrical engineering as it was an exercise in algorithm design. Before their work on the ENIAC was complete, John W. Mauchly and J. Presper Eckert conceived of an easier way to change the behavior of their calculating machine. They reckoned that memory devices, in the form of mercury delay lines, could provide a way to store program instructions. This would forever end the tedium of rewiring the systeme each time it had a new problem to solve, or an old one to debug. Mauchly and Eckert documented theiri idea proposing it as the foundation for their next computer, the EDVAC. Unfortunately, while they werei involvedin the top secret ENIAC project during World War II, Mauchly and Eckert could not immediately publish their insight.

No such proscriptions, however, applied to a number of people working at the periphery of the ENIACproject. One of these people was a famous Hungarian mathematician named John von Neumann ( pronounced von noy-man). After reading Mauchly and Eckert’s proposal for the EDVAC, von Neumann published and publicized the idea. So effective was he in the delivery of this concept that history has credited him with its invention. All stored-program computers have come to be known as von Neumann systems using the von Neumann architecture. Although we are compelled by tradition to say that stored- program computers use the von Neumann architecture, we shall not do so without paying proper tribute to its true inventors: John W. Mauchly and J. Presper Eckert.

Today’s version of the stored-program machine architecture satisfies at least the following

characteristics:

•  Consists of three hardware systems: A      central processing unit (CPU) with a control unit, an arithmetic logic unit (ALU), registers (small storage areas), and a program counter; a main memory system which holds programs that control the computer’s operation; and an I/O system.

•  Capacity to carry out sequential                    instruction processing.

• Contains a single path, either physically      or logically, between the main memory        system and the control unit of the CPU,      forcing alternation of instruction and          execution cycles. This single path is            oftenr referredto as the von Neumann  bottleneck.

Figure  shows how these features work together in modern computer systems. Notice that the systems shownin the figure passes all of its I/O through the arithmetic logic unit (actually, it passes through the accumulator which is part of the ALU). This architecture runs programs in what is known as the von Neumann execution cycle (also called the fetch-decode-execute cycle), which describes how the machine works. One iteration of the cycle is as follows:

1. The control unit fetches the next program instruction from the memory, using the program counter to determine where the instruction is located.

2. The instruction is decoded into a language the ALU can understand.

3. Any data operands required to execute the instruction are fetched from memory and placed in registers in the CPU.

4. The ALU executes the instruction and places the results in registers or memory.

The ideas present in the von Neumann architecture have been extended so that programs and data stored in a slow-to-access storage medium, such as a hard disk, can be copied to a fast-access, volatiles storagemedium such as RAM prior to execution. This architecture has also been streamlined into what is currently called the system bus model, which is shown in Figure 1.6. The data bus moves data from main memory to the CPU registers (and vice versa). The address bus holds the address of the data that the data bus is currently accessing. The control bus carries the necessary control signals that specify how the information transfer is to take place.

Other enhancements to the von Neumann architecture include using index registers for addressing, adding floating-point data, using interrupts and asynchronous I/O, adding virtual memory, and adding general registers.