VON NEUMANN ARCHITECTURE: ORIGIN, MODEL, HOW IT WORKS - TECHNOLOGY - 2024

The von Neumann architecture is a theoretical design so that a computer can have a program stored internally, serving as the basis for almost all computers that are currently made.

A von Neumann machine consists of a central processing unit, which includes an arithmetic logic unit and a control unit, plus a main memory, secondary storage, and input / output devices.

Source: David strigoi - Own work, Public Domain, commons.wikimedia.org

This architecture assumes that every computation extracts data from memory, processes it, and then sends it back to memory.

In a von Neumann architecture, the same memory and the same bus are used to store both the data and the instructions that execute a program.

Architecture improvement

Because data and program memory cannot be accessed at the same time, the von Neumann architecture is prone to bottlenecks and computer performance weakening. This is what is known as the von Neumann bottleneck, where power, performance and cost are affected.

One of the changes made involved rethinking how much data actually needed to be sent to memory and how much could be stored locally.

In this way, instead of having to send everything to memory, multiple caches and proxy caches can reduce the flow of data from processor chips to different devices.

Origin

In 1945, after World War II, two scientists independently raised how to build a more malleable computer. One of them was the mathematician Alan Turing and the other was the equally talented scientist John Von Neumann.

British Alan Turing had been involved in cracking the Enigma code in Bletchley Park, using the 'Colossus' computer. On the other hand, the American John Von Neumann had been working on the Manhattan Project to build the first atomic bomb, which required a great deal of manual calculations.

Until then, wartime computers were more or less "programmed" by reconnecting the entire machine in order to carry out a different task. For example, the first computer called ENIAC took three weeks to reconnect to do a different calculation.

The new concept consisted in that in a memory not only the data had to be stored, but also the program that processed that data should be stored in the same memory.

This internally stored program architecture is commonly known as the 'Von Neumann' architecture.

This novel idea meant that a computer with this architecture would be much easier to reprogram. Indeed, the program itself would be treated the same as the data.

Model

The main foundation of the Von Neumann model is the thought that the program is stored internally in a machine. The memory unit contains the data and also the program code. The architecture design consists of:

Source: From UserJaimeGallego - This file is derived from Von Neumann Architecture.svg, CC BY-SA 3.0, commons.wikimedia.org

- Central Processing Unit (CPU)

It is the digital circuit that is responsible for executing the instructions of a program. It is also called a processor. The CPU contains the ALU, the control unit, and a set of registers.

Logical arithmetic unit

This part of the architecture is only involved in performing arithmetic and logical operations on the data.

The usual calculations of adding, multiplying, dividing and subtracting will be available, but data comparisons such as 'greater than', 'less than', 'equal to' will also be available.

Control unit

It controls the operation of the computer's ALU, memory, and input / output devices, instructing them how to act on the instructions in the program you just read from memory.

The control unit will manage the process of moving data and programs to and from memory. It will also take care of executing the program's instructions, one at a time or sequentially. This includes the idea of ​​a register to hold the intermediate values.

Records

They are high-speed storage areas on the CPU. All data must be stored in a register before it can be processed.

The memory address register contains the memory location of the data to be accessed. The memory data register contains the data that is transferred to memory.

- Memory

The computer will have memory that can hold data, as well as the program that processes that data. In modern computers this memory is the RAM or main memory. This memory is fast and accessible directly by the CPU.

RAM is divided into cells. Each cell consists of an address and its content. The address will uniquely identify each location in memory.

- Entrance exit

This architecture allows to capture the idea that a person needs to interact with the machine, through the input-output devices.

- Bus

Information must flow between the different parts of the computer. In a computer with the von Neumann architecture, information is transmitted from one device to another along a bus, connecting all the CPU units to main memory.

The address bus carries the addresses of data, but not data, between the processor and memory.

The data bus carries data between the processor, memory, and input-output devices.

How does the von Neumann architecture work?

The relevant principle of the von Neumann architecture is that both data and instructions are stored in memory and treated the same, which means that instructions and data are directional.

It works using four simple steps: find, decode, execute, store, called the "Machine Cycle."

The instructions are obtained by the CPU from memory. The CPU then decodes and executes these instructions. The result is stored back in memory after the instruction execution cycle completes.

Search for

In this step the instructions are obtained from RAM and cached for access by the control unit.

Decode

The control unit decodes the instructions in such a way that the logical arithmetic unit can understand them, and then sends them to the logical arithmetic unit.

Run

The arithmetic logical unit executes the instructions and sends the result back to the cache.

To stock

Once the program counter indicates stop, the final result is downloaded to main memory.

Bottleneck

If a Von Neumann machine wishes to perform an operation with data in memory, it has to be transferred via the bus to the CPU. After performing the calculation, the result needs to be moved to memory via the same bus.

The Von Neumann bottleneck occurs when data being entered or removed from memory must linger while the current memory operation completes.

That is, if the processor has just completed a calculation and is ready to perform the next one, it has to write the finished calculation, which occupies the bus, into memory before it can retrieve new data from memory, which also use the same bus.

This bottleneck has been getting worse over time, because the microprocessors have increased their speed and on the other hand the memory has not advanced so quickly.

Advantage

- The control unit retrieves data and instructions in the same way from memory. Therefore, the design and development of the control unit is simplified, being cheaper and faster.

- Data from input / output devices and main memory is retrieved in the same way.

- The organization of the memory is carried out by the programmers, which allows to use all the memory capacity.

- Managing a single memory block is simpler and easier to achieve.

- The design of the microcontroller chip is much simpler, since only one memory will be accessed. The most important thing about the microcontroller is access to RAM and in the von Neumann architecture this can be used both to store data and to store program instructions.

Development of operating systems

The main advantage of having the same memory for programs and data is that programs can be processed as if they were data. In other words, you can write programs whose data is other programs.

A program whose data is another program is nothing more than an operating system. In fact, if programs and data were not allowed in the same memory space, as is the case with the von Neumann architecture, operating systems could never have been developed.

Disadvantages

Although the advantages far outweigh the disadvantages, the problem is that there is only one bus connecting the memory to the processor, so only one instruction or data item can be fetched at a time.

This means that the processor may have to wait longer for data or instructions to arrive. This is known as the von Neumann bottleneck. As the CPU is much faster than the data bus, this means that it is often idle.

- Due to the sequential processing of the instructions, the parallel implementation of the program is not allowed.

- By sharing memory, there is a risk that one instruction will be written over another due to an error in the program, causing the system to crash.

- Some defective programs cannot free memory when they are done with it, which could cause the computer to freeze due to insufficient memory.

- Data and instructions share the same data bus, although the speed at which each must be retrieved is usually very different.

References

1. Semiconductor Engineering (2019). Von Neumann Architecture. Taken from: semiengineering.com
2. Scott Thornton (2018). What's the difference between Von-Neumann and Harvard architectures? Microcontroller Tips. Taken from: microcontrollertips.com.
3. Teach ICT (2019). The Von Neumann machine. Taken from: teach-ict.com.
4. Computer Science (2019). Von Neumann Architecture. Taken from: computerscience.gcse.guru.
5. Learn IT with Mr C (2019). The Von Neumann machine. Taken from: learnitwithmrc.co.uk.
6. Solid State Media (2017). How do computers work? The Von Neumann Architecture. Taken from: solidstateblog.com.