What is Harvard Architecture?
Harvard architecture is a type of computer architecture that evolved from the Von Neumann model. It was first developed in the 1940s, and has since become widely used in the computer industry. The main difference between the two architectures is that the Harvard model has two separate memory spaces: one for instructions, and one for data. This makes it more efficient and allows for better control over the execution of instructions.
How Harvard Architecture Works
Harvard architecture, unlike Von Neumann architectures, separates instruction and data memory, accessing both at the same time. Instructions and data are held in different memory sections, however, the memory controller manages them both simultaneously. The Harvard architecture enables both instruction and data access in parallel via the program counter, instruction register, instruction decoder, and data register, each likely being accessed and managed by the memory controller.
Depending on what type of processor is used, either an ASIC or a programmable microcontroller, its instructions are held in either ROM/PROM or RAM. The instructions can be programmed external to the device, and integrated into the core’s memory. A Harvard architecture can also support multiple register banks and each can contain bits, bytes, words, and power-of-2 bytes, depending on the implementation of the architecture.
Benefits of Harvard Architecture
Harvard architecture is more complicated than Von Neumann architecture, however, the benefits more than outweigh the costs. The most obvious advantage of the Harvard architecture is its improved instruction performance as both data and instructions can be retrieved/executed in parallel rather than in serial. This significantly reduces wait- State operations and increases the CPU throughput.
Another advantage of the Harvard architecture is that since data and instruction are stored in separate memories, it is easier to protect the instruction memory from external threats. The instruction memory is usually implemented in ROM/PROM that cannot be modified, which limits malicious access to the system and virus attacks. Moreover, the instruction-memory protects the code and related routines from being overwritten when the program is being executed. Each instruction can read data right away, making it much faster than Von Neumann architecture, where programs need to access data first before moving onto the next instruction.
Disadvantages of the Harvard Architecture
However, the Harvard architecture has its own set of drawbacks. The architecture is mainly tailored for systems with specialized requirements, so cannot be used for any application. As the architecture has its own unique features, programming this architecture can be difficult, particularly when compared to the Von Neumann architecture that is functional in almost all applications. Further, development using the Harvard architecture is costlier than the Von Neumann architecture.
And, since the design of the Harvard architecture is more complex and the instructions are stored in static memory, adding more instructions requires considerable changes to the hardware, which can make the design process very troublesome.
Applications of Harvard Architecture
Since the Harvard architecture is an efficient way of dealing with large datasets, it is mainly associated with embedded systems research. NASA’s Curiosity Rover is a classic example of the Harvard architecture in use. It stores data collected by the sensors in different areas, is immune to data corruption, and allows Curiosity to receive instructions simultaneously.
Dave Johnson, a senior engineer specializing in power management and embedded systems, believes the Harvard architecture approach works best for electronic system designs, “The Harvard architecture is particularly well suited for FPGAs and microcontroller-based applications, especially where data needs to be processed at low clock speeds, and the hardware needs to rapidly respond to hardcoded rules,” says Johnson.
The Harvard architecture is prevalent in the security and military industries, where ‘time-critical’ decisions are made, surveillance and reconnaissance drones need to access data rapidly, and data safety is paramount.
Differences and Comparisons between Harvard and Von Neuman Architectures
The Harvard architecture and the Von Neuman architecture both store data and instructions, but there are several key differences between them. In the Von Neuman architecture, instructions and data are stored together, while in the Harvard architecture they are stored in separate memory areas. In the Von Neumann architecture, programs occupy the same address space as data, while the Harvard architecture keeps instructions and data in two different spaces.
In addition, the Von Neumann architecture uses an instruction cache to speed up fetch and delivery, while the Harvard architecture can access instructions and data, simultaneously. The Von Neumann architecture requires two instruction cycles to get an instruction and data, while the Harvard architecture only needs a single cycle.
Comparing Performance of Harvard and Von Neuman Architectures
In terms of performance, the Harvard architecture is considered the superior architecture for embedded systems design. This is because the Harvard architecture can process instructions in parallel, whereas the Von Neumann architecture processes them in a serial manner. Also, the Harvard architecture has a wider memory bandwidth, which makes it faster than the Von Neumann architecture.
Apart from superior performance, the Harvard architecture has better system security because instruction memory is separated from other areas and stored in read-only memory (ROM). This protects the code and related routines from being overwritten when a program is executed. Finally, instruction implementation is easier in the Harvard architecture than in the Von Neumann architecture.
Programming Languages Supported by Harvard Architecture
Due to its modern-day application of embedded systems, Harvard architecture is usually programmed with Assembly language. AsAssembly language is a low-level programming language, it provides more direct communication with the processor and memory, making it the ideal language to program devices operating on the Harvard architecture.
In addition, there are some modern-day programming languages that support Harvard architecture. For example, Java Card programming language takes advantage of the Harvard architecture for cryptographic applications and smart-cards.
Hardware Implications of Harvard Architecture
The fundamentals of the Harvard architecture memory setup require extra hardware systems to support it. However, designing these systems is quite intuitive and is not as complicated as some people think.
The memory layout is separated into instruction and data, meaning different bus systems and address width registers. Each demand having dedicated programmable logic for the discrimination of instructions and data for forwarding, as well as the address decode logic required for the core to access the correct part of the memory.
In addition, the Harvard architecture requires two separate address selection systems, two clock networks, and even two separate clocks. The dual clock networks allow instructions and data traffic to perform tasks at the same time, while two power nets are necessary so that each part of the memory can be powered without affecting the other.
Impact of Harvard Architecture on the Computer Industry
Since its first introduction, the Harvard architecture has revolutionized the computer industry in many ways. It is especially beneficial for handling complex data workloads, embedded systems, and military applications.
Due to the dual memory structure, instructions and data are available concurrently, making it much faster than the Von Neumann architecture, as instructions can be read as soon as they are loaded into the decoder. The Harvard architecture is usually implemented in single-chip microprocessors, allowing it to access large datasets with minimal processing power.
Coming to the security aspect, the Harvard architecture offers better protection to the instruction memory as it is usually implemented in ROM/PROM, which cannot be modified. This prevents malicious access to the system and virus attacks.
Conclusion
The Harvard architecture is a revolutionary advancement in computer architecture as it allows for faster and more secure access to data and instructions. While it is more complex than the Von Neumann architecture and without a doubt more costly to develop, the benefits of the Harvard architecture have revolutionized the computer industry.
From embedded systems in NASA’s Curiosity Rover, to large-scale military applications and surveillance drones, the Harvard architecture has forever changed what computers are capable of. With its dual memory layout, combined with the speed, security, and flexibility it offers, the Harvard architecture is here to stay.
