Chapter 3 (Hardware)

abstraction

abstraction

A method used to hide unneeded or complex details of a problem or idea

Found in lecture Chapter 3 - Hardware 1

von Neumann Architecture

Computer architecture with four subsystems:1) Memory2) Input/Output Devices (I/O)3) ALU4) ControlTwo characteristics:1) Instructions are stored like data2) Instructions are executed sequentially

Found in lecture Chapter 3 - Hardware 2

RAM

Random Access Memory.Memory is divided into cells. Each cell (aka minimum unit of access) is associated with an address. A standard cell width is 8 bits (1 byte). Must always access ENTIRE cell, even if you just want 1 bit.

Found in lecture Chapter 3 - Hardware 1 and zyBook Section 3.2

ROM

Read Only Memory. ROM holds important system instructions and data such as the START UP instructions for a computer (aka firmware). Usually info has been prerecorded during manufacture.

Found in lecture Chapter 3 - Hardware 1

volatile memory

Information disappears without power, e.g., RAM and cache (lose data on power loss)

Found in lecture Chapter 3 - Hardware 1 and zyBook Section 3.3

non-volatile memory

Information preserved without power, e.g., ROM, disc, flash

Found in lecture Chapter 3 - Hardware 1 and zyBook Section 3.3

cache

Stores DATA from a slower device to be accessed faster. On-chip memory used to decrease memory access times. Works because of locality.

• SMALL (on order of few megabytes) because it's expensive.

• FAST (5-10x faster than RAM)

We use cache to decrease memory access time.

Found in lecture Chapter 3 - Hardware 2 and zyBook Section 3.3

cache hit

When requested data is found in cache

Found in lecture Chapter 3 - Hardware 2

cache hit rate/ratio

Probability a data request will be a cache hit. In other words, the % of time that the info needed is in cache memory.

INCREASE cache hit ratio -> DECREASE overall memory access time.

• Good to know: Cache hit ratio CHANGES as system executes:

• Decreases as we move from one locality to another.

• Increases as we stay in the same locality (if cache is big enough to hold locality).

Found in lecture Chapter 3 - Hardware 2

cache miss

When requested data is not found in cache

Found in lecture Chapter 3 - Hardware 2

What can we do to increase cache hit ratio?

• Increase size of cache (can store more, small cache size can yield 75% hit ratio).

• Be smarter about what to put in the cache (Studies show 98% of process time is spent in localities. If OS can be smarter about what to put in cache -> increase hit ratio)

Found in lecture Chapter 3 - Hardware 2

memory cell

Fixed sized unit of memory. Standard size = 8 bits = 1 byte

address

Unique identifier for a memory cell.

Address space = MAX # of possible addresses.

Example question: Suppose RAM addresses are N bits. How many memory cells can theoretically exist?2^N.

If 16 bit ADDRESS, how many cells could exist in memory? 65,536 (64K).

Found in lecture Chapter 3 - Hardware 1

register

Stores VALUES that are BEING USED during processing.

Special high speed storage cell that holds something specific. Note: registers are not designed for general purposes like other types of memory (e.g., cache, ram)

Found in lecture Chapter 3 - Hardware 3

ALU

Arithmetic Logic Unit. Performs math and logical operations (e.g., addition, subtraction, comparisons, and, or, not)

Found in lecture Chapter 3 - Hardware 2 and zyBook Section 3.2

control unit

Component of processor that FETCES, DECODES, and EXECUTES instructions.

Found in lecture Chapter 3 - Hardware 2 and zyBook Section 3.2

processor

ALU + control unit

Found in lecture Chapter 3 - Hardware 2 and zyBook Section 3.2

interrupt

An input signal to the processor indicating an event that needs immediate attention. Alerts the processor and serves as a request for the processor to interrupt the currently executing code, so that the event can be processed in a timely manner.

Found in lecture Chapter 3 - Hardware 3

Moore's Law

Since the invention of the IC (integrated circuit) around 1960, IC's have doubled in circuit capacity about every 1.5 or 2 years.

Found in lecture Chapter 3 - Hardware 2 and zyBook Section 3.7

truth table

One way to describe the behavior of a combinational circuit, listing the output value for every possible combination of input values.

Found in lecture Chapter 3 - Circuits 1 and zyBook Section 3.1

transistor

a device with no mechanical/moving parts that can be OFF or ON

Found in lecture Chapter 3 - Circuits 2 and zyBook Section 3.7

tautology

A Boolean expression that is always true. Example: (A OR B) OR (NOT A)

Found in lecture Chapter 3 - Circuits 1

circuit

a collection of logic gates that transforms a set of binary inputs into a set of binary outputs.

Found in lecture Chapter 3 - Circuits 1 and zyBook Section 3.1

gate

an electronic device with 1+ transistors that operates on a set of binary inputs to produce a binary output.

Found in lecture Chapter 3 - Circuits 1 and zyBook Section 3.1

boolean expression

an expression that evaluates to either true or false

Found in lecture Chapter 3 - Circuits 1

boolean logic

An area of mathematics dealing with rules for manipulating true and false

Found in lecture Chapter 3 - Circuits 1

computer

a device that:

1. Takes input.

2. Processes the input in some way.

3. Produces output (a result).

Found in lecture Chapter 3 - Hardware 1

von Neumann Bottleneck

• As CPU speeds have increased, CPU idle time has increased (due to waiting for data to be fetched from memory).

• Needs significant changes to computer architecture to handle larger problems.

Found in lecture Chapter 3 - Hardware 2

Two key issues with von Neumann architectures and how we might solve them

TWO ISSUES:

1. Inability to place circuits closer together on a chip. Moore's Law is slowing down! Slow down in processor speed while our problems to be solve are getting larger!-> Solution: Parallel Processing (multi-processor). "If you can't build something to work twice as fast, build it to do two things at once. The results will be identical".

2. Slow speed to access memory.-> Solution: adding a cache helps. Adding multi-threading helps. Adding new types of RAM (e.g., DDR SDRAM) helps. But NONE solve the full problem. No matter how fast a given CPU can work, it is limited by the rate of memory transfer.

Found in lecture Chapter 3 - Hardware 2

Memory vs. Disc

Discs are SLOW (order of magnitude slower than memory).

Solid state disc (SSD) are more common for discs now:

• no moving part -> faster

• more expensive $$$

Why not use SSD as RAM?

• SSD are starting to get as fast as RAM, but SSDs wear out (would wear out fast if accessed it OFTEN, like we do with RAM).

• SSDs optimized to act as a disc. RAM optimized to act as memory.

Found in lecture Chapter 3 - Hardware 2

bootloader

Transfers OS from a pre-determined location into RAM. Jump to OS for system run.

Found in lecture Chapter 3 - Hardware 2

The 4 basic instruction types

Data transfer: moves data to/from memory and I/O devices.

Arithmetic: Calculate numerical operations. (e.g., add)

Comparison: Compare 2 values (e.g., ==, <, >, !=) and set flag appropriately.

Branch/Jump: control execution flow of program (e.g., if, loops). Changes value in program counter (PC) register.

Found in lecture Chapter 3 - Hardware 3

instruction set

Set of all operations that can be executed by a processor. Set of all implemented opcodes.

Found in lecture Chapter 3 - Hardware 3

address fields

Second part of machine instruction. Specifies which memory addresses are being operated on. (up to 3)

Found in lecture Chapter 3 - Hardware 3

Operation code (opcode)

Opcode. First part of machine language instruction. Specifies which operation needs to be carried out (e.g., addition, store)

• If 2-bit OPCODE, how many different instructions are possible? Four (01 Input, 11 Add, 10 Output, 00 Stop).

• If 8 bit OPCODE? 2^N = 2^8 = 256 different instructions are possible.

Found in lecture Chapter 3 - Hardware 3

machine language

Binary representation of program instructions. Each machine language instruction has 2 types of fields:

1. an opcode (specifies which operation needs to be carried out).

2. address fields (specifies what memory addresses are being operated on).

Found in lecture Chapter 3 - Hardware 3

execute

Control unit action: Execute instruction by issuing appropriate command

Found in lecture Chapter 3 - Hardware 3

decode

Control unit action: use OPCODE to determine what to do for an instruction.

Found in lecture Chapter 3 - Hardware 3

fetch

Control unit action: fetch the next instruction from memory at address stored in PC (program counter).

Fetch gets value at an address (nondestructive fetch)

• Value remains in memory

• Load MAR, Decode MAR (fetch address; place in MDR)

• return value in MDR

Found in lecture Chapter 3 - Hardware 3

MIPS

millions of instructions per second

Found in lecture Chapter 3 - Hardware 3

GIPS

billions of instructions per second

Found in lecture Chapter 3 - Hardware 3