Classifications of Computer Architecture

Summary

This lecture delves into the classifications of computer architecture, focusing on two main models: Von Neumann and Harvard. Von Neumann architecture, also known as Princeton architecture, features a single path between the CPU and memory, leading to the Von Neumann bottleneck. In contrast, Harvard architecture separates instruction and data memory to overcome this limitation. The session also covers Flynn's taxonomy, categorizing architectures based on instruction and data streams, with SISD, SIMD, MISD, and MIMD types, providing a foundational understanding for modern processor design.

Takeaways

• 🏛️ Computer architecture is classified into two main categories: Von Neumann architecture and non-Von Neumann architecture.
• 🔍 Von Neumann machines have three basic hardware subsystems: CPU, main memory, and I/O system, and they operate as stored program computers with sequential instruction execution.
• 🔄 The Von Neumann bottleneck occurs when data and instructions share a single memory pathway, limiting simultaneous instruction and data processing.
• 🏫 The Harvard architecture, proposed by Harvard University, overcomes the Von Neumann bottleneck by using separate memory units for instructions and data, allowing parallel processing.
• 🛠️ Modified Harvard architecture combines elements of both Harvard and Von Neumann architectures, with a small, fast cache memory that can act as either depending on the access mode.
• 🤖 Flynn's taxonomy classifies computer architectures based on the number of instruction streams and data streams, such as SISD, SIMD, MISD, and MIMD.
• 🔢 SISD (Single Instruction Stream, Single Data Stream) is the category where Von Neumann architecture falls, with one CPU executing one instruction at a time.
• 🌟 SIMD (Single Instruction Stream, Multiple Data Stream) machines have multiple ALUs executing the same operation on different data sets in parallel, controlled by a single control unit.
• 🛑 MISD (Multiple Instruction Streams, Single Data Stream) is a theoretical category with no practical implementation, where different instructions operate on the same data item.
• 🔄 MIMD (Multiple Instruction Streams, Multiple Data Streams) refers to multiprocessors with independent processors, each executing different instructions on different data sets.
• 📚 Michael J. Flynn's classification system has been instrumental in the design and functionality of modern processors since its introduction in 1966.

Q & A

What are the two broad classifications of computer architecture discussed in the script?

The two broad classifications of computer architecture discussed are Von Neumann architecture and non-Von Neumann architecture.

What are the three basic hardware subsystems of a Von Neumann machine?

The three basic hardware subsystems of a Von Neumann machine are the CPU, the main memory, and the I/O system.

What is the main characteristic that defines a stored program computer?

A stored program computer is defined by its ability to hold the program controlling the computer operation in the main memory, allowing the computer to manipulate its own program as well as any other data stored inside the memory.

What is the 'Von Neumann bottleneck' and why does it occur?

The 'Von Neumann bottleneck' refers to the limitation where a processor cannot simultaneously read an instruction and operate on data because both instructions and data are fetched over the same path. This occurs in systems with pure Von Neumann architecture where instructions and data are stored in the same memory unit.

What is the Harvard architecture and how does it differ from the Von Neumann architecture?

The Harvard architecture is a model that uses separate memory units for storing instructions and data items. This allows the processor to read an instruction and perform data memory access simultaneously, unlike the Von Neumann architecture where both operations share a single memory pathway.

What is the Modified Harvard architecture and how does it combine features of both Harvard and Von Neumann architectures?

The Modified Harvard architecture is similar to the Harvard architecture but it relaxes the strict division of instruction memory and data memory. It includes a small, fast memory storage called caching. When the processor executes from the cache, it acts like a pure Harvard architecture, and when accessing from the backing memory, it acts like a pure Von Neumann machine.

What does SISD stand for in computer architecture and what are its characteristics?

SISD stands for Single Instruction Stream, Single Data Stream. Computers in this category have one CPU that executes one instruction at a time and fetches and stores one item of data at a time.

What is SIMD and how does it differ from SISD?

SIMD stands for Single Instruction Stream, Multiple Data Stream. Unlike SISD, SIMD machines have a control unit that operates like a Von Neumann machine executing a single instruction stream, but they have more than one ALU, allowing the execution of the same operation on different sets of data items in lock steps.

What is the difference between MISD and MIMD in Flynn's taxonomy?

MISD stands for Multiple Instruction Streams, Single Data Stream, where theoretically different programs would execute on the same data items. MIMD stands for Multiple Instruction Streams, Multiple Data Streams, where each processor can execute a different set of instructions on its own set of data, making it suitable for parallel processing.

Who proposed the classification model used to categorize computer architectures by the number of processors, programs, and memory structures?

The classification model was proposed by Michael J. Flynn in 1966, and it is commonly known as Flynn's taxonomy.

What is the significance of Flynn's taxonomy in the design of modern processors?

Flynn's taxonomy is significant as it has been used as a tool in the design of modern processors and their functionalities, helping to understand and categorize different types of computer architectures based on their instruction and data handling capabilities.

Outlines

00:00

💾 Introduction to Computer Architecture Classifications

The video script begins with an introduction to the topic of computer architecture classifications. It explains that computer architecture can be broadly divided into two categories: Von Neumann and non-Von Neumann architectures. Von Neumann architecture, also known as Princeton architecture, is characterized by three basic hardware subsystems: the CPU, main memory, and I/O system. It is a stored-program computer where the program controlling the computer operation is stored in the main memory. This architecture is known for its sequential instruction execution and the single path between the main memory and the processor, which can lead to the Von Neumann bottleneck. The Harvard architecture, proposed by Harvard University, is a non-Von Neumann model that uses separate memory units for instructions and data, allowing for parallelism and overcoming the bottleneck issue. Modified Harvard architecture is also discussed, which combines elements of both Harvard and Princeton architectures, featuring a cache memory that can act as a pure Harvard architecture when executing from cache and as a Von Neumann machine when accessing from the backing memory. The script provides a detailed explanation of these classifications and their significance in modern processor design.

05:01

🔍 Flynn's Taxonomy of Computer Architectures

The second paragraph delves into the classification of computer architectures based on the number of instruction streams and data streams they handle. It introduces the concept of SISD (Single Instruction Stream, Single Data Stream), where computers like the Von Neumann architecture execute one instruction at a time and process one data item at a time. SIMD (Single Instruction Stream, Multiple Data Stream) is explained as architectures with a control unit that operates like a Von Neumann machine but with multiple ALUs executing the same operation on different data items in parallel. The paragraph also mentions MISD (Multiple Instruction Streams, Single Data Stream), which theoretically executes different programs on the same data but lacks practical implementation. Finally, MIMD (Multiple Instruction Streams, Multiple Data Streams), also known as multiprocessors, is described as architectures with more than one independent processor, each capable of executing different instructions on different data sets. This classification, proposed by Michael J. Flynn in 1966, is known as Flynn's taxonomy and has been instrumental in the design and functionality of modern processors.

Keywords

💡Computer Architecture

Computer architecture refers to the design and organization of computer components and systems. It is the fundamental concept that this video is built upon, as it discusses various classifications and models within this field. The script delves into how different architectures affect the operation and performance of computers.

💡Von Neumann Architecture

The Von Neumann architecture, also known as Princeton architecture, is a model where the CPU, main memory, and I/O system are the basic hardware subsystems, and it operates on a stored program concept. The video explains this architecture as having a single path between the main memory and the processor, which can lead to the Von Neumann bottleneck.

💡Harvard Architecture

Harvard architecture is another model that separates instruction memory from data memory, allowing for parallel processing of instructions and data. This separation overcomes the Von Neumann bottleneck by providing two distinct memory pathways, as mentioned in the script when discussing the need for this architecture.

💡Von Neumann Bottleneck

The Von Neumann bottleneck is a concept where the data and instructions share a common bus, leading to a limitation in processing speed since the CPU cannot read an instruction and operate on data simultaneously. The script uses this term to illustrate the limitations of the pure Von Neumann architecture.

💡Modified Harvard Architecture

Modified Harvard architecture is a hybrid model that combines elements of both Harvard and Von Neumann architectures. It includes a small, fast cache memory and a larger backing memory. The script explains that when the processor uses the cache, it operates like a Harvard architecture, and when accessing the backing memory, it behaves like a Von Neumann machine.

💡SISC (Single Instruction Stream, Single Data Stream)

SISC is a classification within Flynn's taxonomy, referring to computers that have one CPU executing one instruction at a time and processing one piece of data at a time. The script mentions that the Von Neumann architecture belongs to this category.

💡SIMD (Single Instruction Stream, Multiple Data Stream)

SIMD computers have a single instruction stream but can process multiple data streams simultaneously, often using multiple ALUs under the control of a single control unit. The script describes SIMD as machines where the same operation is performed on different data items in lock steps.

💡MISD (Multiple Instruction Streams, Single Data Stream)

MISD is a theoretical model where multiple instructions operate on a single data stream. The script notes that there is no practical implementation of this model, but it is part of the classification system proposed by Michael J. Flynn.

💡MIMD (Multiple Instruction Streams, Multiple Data Streams)

MIMD, also known as multiprocessors, are computers with more than one independent processor, each capable of executing different sets of instructions on different data sets. The script explains that this classification represents the most flexible and powerful type of computer architecture.

💡Flynn's Taxonomy

Flynn's Taxonomy is a classification system for computer architectures proposed by Michael J. Flynn in 1966. It categorizes architectures based on the number of instruction streams and data streams they can handle. The script uses this taxonomy to explain the different types of computer architectures, such as SISC, SIMD, MISD, and MIMD.

Highlights

Introduction to the lecture on classifications of computer architecture.

Broad classification of computer architecture into Von Neumann and non-Von Neumann architectures.

Characteristics of Von Neumann machines including three basic hardware subsystems and stored program concept.

Sequential instruction execution in Von Neumann architecture leading to a single path debate.

The concept of Princeton architecture as an evolution of Von Neumann's model.

Harvard architecture's introduction of separate memory units for instructions and data to overcome Von Neumann bottleneck.

Understanding the need for Harvard architecture due to limitations in pure Von Neumann architecture.

Modified Harvard architecture combining features of both Harvard and Princeton models with caching.

Flynn's taxonomy as a classification model based on the number of instruction streams and data streams.

SISD (Single Instruction Stream, Single Data Stream) classification and its relation to Von Neumann architecture.

SIMD (Single Instruction Stream, Multiple Data Stream) classification and its application in processor arrays.

MISD (Multiple Instruction Streams, Single Data Stream) as a theoretical model with no practical implementation.

MIMD (Multiple Instruction Streams, Multiple Data Streams) as a category for multiprocessors with independent processors.

Practical applications of modified Harvard architecture in modern processors with caching mechanisms.

Flynn's taxonomy's impact on the design and functionality of modern processors since 1966.

Conclusion of the lecture with a summary of the two popular classifications of computer architecture.

Appreciation for the audience's attention and an invitation to the next lecture.

Transcripts

[Music]

well hello everyone welcome to the

session and as promised in this

particular lecture we are going to dive

right into various classifications of

computer architecture so without any

further ado let's get to learning now

coming to computer architecture's

classifications there are various models

among them we will be learning about two

most popular approaches so let's start

with the first one computer architecture

is broadly classified into specifically

two categories one newman architecture

and non-von neumann architecture now one

newman machines have the following

characteristics it has three basic

hardware subsystems the cpu the main

memory and the i o system

moreover it must be a stored program

computer that means the main memory

system holds the program which controls

the computer operation and the computer

can manipulate its own program more or

less also it can do the same to any

other data stored inside the memory

furthermore it carries out instructions

sequentially so the cpu appears to

execute one operation at a time well the

major debate has always been about

having a single path between the main

memory and the processor proposed by

this very architecture now already in

our previous discussion we came to know

about von neumann's professorship at

princeton university and that is the

reason why this architecture was also

referred to as princeton architecture

now people from harvard university

proposed another model which adopted

most of the features proposed in the

princeton architecture except for the

fact they used two different memory

units and thus pioneered parallelism

this is popularly known as harvard

architecture let's try to understand the

need of it first of all systems having

pure von neumann architecture stores the

instructions and the data in the same

memory unit thus both the data and the

instructions are phased over the same

path

this means the processor can't

simultaneously read an instruction and

operate on data this phenomenon is known

as von neumann bottleneck and was first

officially discussed by the great john

backers inventor of backus normal form

for context free languages during his

1977 acm turing hour lecture next there

is modified harvard architecture which

cannot completely be classified as von

neumann architecture as it doesn't

strictly follow the von neumann

principles but it is more of a

combination between both the harvard and

the princeton architecture now let's

study them in a bit details

now in harvard architecture we have

separate memory units

one is for strictly storing the

instructions and the other for the data

items

this way the processor can both read an

instruction and perform data memory

access at the same time therefore

computers belonging to this architecture

family are faster because instruction

fetches and data accesses are not

competing for the single memory pathway

anymore now coming to modified harvard

architecture it is very much like the

harvard architecture itself however it

relaxes the strict division of

instruction memory and data memory here

the processor is accompanied by a small

yet fast memory storage called caching

when the processor is executing from the

cachet it acts as a pure harvard

architecture and when it is accessing

from the backing memory it acts like a

pure von neumann machine this

modification it wide spread in modern

processors so this was the first

approach of classification of computer

architecture

let's move on to the next one in this

model computer architectures are

classified by a variety of

characteristics including the number of

processors the number of programs they

can execute and the memory structures

being used the first group is sisd that

is single instruction stream single data

stream the von neumann architecture

belongs to this specific category sisd

computers have one cpu that executes one

instructions at a time hence single

instruction stream and fetches and

stores one item of data at a time hence

single data stream the next group is

simd that is single instruction stream

multiple data stream

processor array falls into this category

simd machines have a control unit that

operates like a von neumann machine that

executes a single instructions stream

but they have more than one alus the

control unit generates the control

signals for all the alus which execute

the same operation on different set of

data items generally in lock steps hence

multiple data streams

the next group of computers are misd

multiple instruction streams single data

stream logically machine of this class

would execute various different programs

on the same data items this group

doesn't have any particular practical

implementation nonetheless some machines

belonging to the next category can be

used in this manner all right the next

and the final type is mimd multiple

instruction streams multiple data

streams machines these are also known as

multiprocessors

these are comprised of more than one

independent processors and each one of

them can execute a different set of

instructions hence multiple instruction

streams on its own set of data hence

multiple data streams now this

classification was proposed by michael j

flynn in 1966

since then it has been used as a tool in

design of modern processors and their

functionalities that is why it is also

known as flynn's taxonomy

alright folks that was the two most

popular classifications of computer

architecture hope you liked it and

learned something new i'll see you guys

in the next one thank you all for

watching

[Applause]

[Music]

you